Methods and compositions for diagnosing and monitoring transplant rejection

[0001] This application is a Continuation-in-Part application of Ser. No. 10/131,831, filed Apr. 24, 2002, now U.S. Pat. No. 7,026,121 which is a Continuation-in-Part application of Ser. No. 10/006,290 filed Oct. 22, 2001, which claims priority to U.S. provisional patent application No. 60/296,764 filed Jun. 8, 2001, all of which are hereby incorporated by reference in their entirety.

[0002] This invention is in the field of expression profiling. In particular, this invention is in the field of leukocyte expression profiling.

[0003] Many of the current shortcomings in diagnosis, prognosis, risk stratification and treatment of disease can be approached through the identification of the molecular mechanisms underlying a disease and through the discovery of nucleotide sequences (or sets of nucleotide sequences) whose expression patterns predict the occurrence or progression of disease states, or predict a patient's response to a particular therapeutic intervention. In particular, identification of nucleotide sequences and sets of nucleotide sequences with such predictive value from cells and tissues that are readily accessible would be extremely valuable. For example, peripheral blood is attainable from all patients and can easily be obtained at multiple time points at low cost. This is a desirable contrast to most other cell and tissue types, which are less readily accessible, or accessible only through invasive and aversive procedures. In addition, the various cell types present in circulating blood are ideal for expression profiling experiments as the many cell types in the blood specimen can be easily separated if desired prior to analysis of gene expression. While blood provides a very attractive substrate for the study of diseases using expression profiling techniques, and for the development of diagnostic technologies and the identification of therapeutic targets, the value of expression profiling in blood samples rests on the degree to which changes in gene expression in these cell types are associated with a predisposition to, and pathogenesis and progression of a disease.

[0004] There is an extensive literature supporting the role of leukocytes, e.g., T-and B-lymphocytes, monocytes and granulocytes, including neutrophils, in a wide range of disease processes, including such broad classes as cardiovascular diseases, inflammatory, autoimmune and rheumatic diseases, infectious diseases, transplant rejection, cancer and malignancy, and endocrine diseases. For example, among cardiovascular diseases, such commonly occurring diseases as atherosclerosis, restenosis, transplant vasculopathy and acute coronary syndromes all demonstrate significant T cell involvement (Smith-Norowitz et al. (1999) Clin Immunol 93:168–175; Jude et al. (1994) Circulation 90:1662–8; Belch et al. (1997) Circulation 95:2027–31). These diseases are now recognized as manifestations of chronic inflammatory disorders resulting from an ongoing response to an injury process in the arterial tree (Ross et al. (1999) Ann Thorac Surg 67:1428–33). Differential expression of lymphocyte, monocyte and neutrophil genes and their products has been demonstrated clearly in the literature. Particularly interesting are examples of differential expression in circulating cells of the immune system that demonstrate specificity for a particular disease, such as arteriosclerosis, as opposed to a generalized association with other inflammatory diseases, or for example, with unstable angina rather than quiescent coronary disease.

[0005] A number of individual genes, e.g., CD11b/CD18 (Kassirer et al. (1999) Am Heart J 138:555–9); leukocyte elastase (Amaro et al. (1995) Eur Heart J 16:615–22; and CD40L (Aukrust et al. (1999) Circulation 100:614–20) demonstrate some degree of sensitivity and specificity as markers of various vascular diseases. In addition, the identification of differentially expressed target and fingerprint genes isolated from purified populations of monocytes manipulated in various in vitro paradigms has been proposed for the diagnosis and monitoring of a range of cardiovascular diseases, see, e.g., U.S. Pat. Nos. 6,048,709; 6,087,477; 6,099,823; and 6,124,433 “COMPOSITIONS AND METHODS FOR THE TREATMENT AND DIAGNOSIS OF CARDIOVASCULAR DISEASE” to Falb (see also, WO 97/30065). Lockhart, in U.S. Pat. No. 6,033,860 “EXPRESSION PROFILES IN ADULT AND FETAL ORGANS” proposes the use of expression profiles for a subset of identified genes in the identification of tissue samples, and the monitoring of drug effects.

[0006] The accuracy of technologies based on expression profiling for the diagnosis, prognosis, and monitoring of disease would be dramatically increased if numerous differentially expressed nucleotide sequences, each with a measure of specificity for a disease in question, could be identified and assayed in a concerted manner. In order to achieve this improved accuracy, the appropriate sets of nucleotide sequences need to be identified and validated against numerous samples in combination with relevant clinical data. The present invention addresses these and other needs, and applies to any disease or disease state for which differential regulation of genes, or other nucleotide sequences, of peripheral blood can be demonstrated.

[0007] The present invention is thus directed to a system for detecting differential gene expression. In one format, the system has one or more isolated DNA molecules wherein each isolated DNA molecule detects expression of a gene selected from the group of genes corresponding to the oligonucleotides depicted in the Sequence Listing. It is understood that the DNA sequences and oligonucleotides of the invention may have slightly different sequences that those identified herein. Such sequence variations are understood to those of ordinary skill in the art to be variations in the sequence which do not significantly affect the ability of the sequences to detect gene expression.

[0008] The sequences encompassed by the invention have at least 40–50, 50–60, 70–80, 80–85, 85–90, 90–95 or 95–100% sequence identity to the sequences disclosed herein. In some embodiments, DNA molecules are less than about any of the following lengths (in bases or base pairs): 10,000; 5,000; 2500; 2000; 1500; 1250; 1000; 750; 500; 300; 250; 200; 175; 150; 125; 100; 75; 50; 25; 10. In some embodiments, DNA molecule is greater than about any of the following lengths (in bases or base pairs): 10; 15; 20; 25; 30; 40; 50; 60; 75; 100; 125; 150; 175; 200; 250; 300; 350; 400; 500; 750; 1000; 2000; 5000; 7500; 10000; 20000; 50000. Alternately, a DNA molecule can be any of a range of sizes having an upper limit of 10,000; 5,000; 2500; 2000; 1500; 1250; 1000; 750; 500; 300; 250; 200; 175; 150; 125; 100; 75; 50; 25; or 10 and an independently selected lower limit of 10; 15; 20; 25; 30; 40; 50; 60; 75; 100; 125; 150; 175; 200; 250; 300; 350; 400; 500; 750; 1000; 2000; 5000; 7500 wherein the lower limit is less than the upper limit.

[0009] The gene expression system may be a candidate library, a diagnostic agent, a diagnostic oligonucleotide set or a diagnostic probe set. The DNA molecules may be genomic DNA, protein nucleic acid (PNA), cDNA or synthetic oligonucleotides.

[0010] In one format, the gene expression system is immobilized on an array. The array may be a chip array, a plate array, a bead array, a pin array, a membrane array, a solid surface array, a liquid array, an oligonucleotide array, a polynucleotide array, a cDNA array, a microfilter plate, a membrane or a chip.

[0011] The present invention is further directed to a method of diagnosing or monitoring transplant rejection in a patient, comprising detecting the expression level of one or more genes or surrogates derived therefrom in the patient to diagnose or monitor transplant rejection in said patient wherein said one or more genes include a nucleotide sequence selected from SEQ ID NO:4; SEQ ID NO:26; SEQ ID NO:60; SEQ ID NO:130; SEQ ID NO:176; SEQ ID NO:184; SEQ ID NO:261; SEQ ID NO:707; SEQ ID NO:792; SEQ ID NO:841; SEQ ID NO:1024; SEQ ID NO:1128; SEQ ID NO:1140; SEQ ID NO:1333; SEQ ID NO:1345; SEQ ID NO:1435; SEQ ID NO:1749; SEQ ID NO:1778; SEQ ID NO:1956; SEQ ID NO:2086; SEQ ID NO:2228; SEQ ID NO:2518; SEQ ID NO:2519; SEQ ID NO:2770; SEQ ID NO:2801; SEQ ID NO:3134; SEQ ID NO:3263; SEQ ID NO:3842; SEQ ID NO:4092; SEQ ID NO:4191; SEQ ID NO:4460; SEQ ID NO:4515; SEQ ID NO:5108; SEQ ID NO:5280; SEQ ID NO:5573; SEQ ID NO:5673; SEQ ID NO:5834; SEQ ID NO:6091; SEQ ID NO:6112; SEQ ID NO:6221; SEQ ID NO:6309; SEQ ID NO:6347; SEQ ID NO:6514; SEQ ID NO:6573; SEQ ID NO:7094; SEQ ID NO:7199; SEQ ID NO:7481; SEQ ID NO:7482; SEQ ID NO:7605; SEQ ID NO:8076 and SEQ ID NO:8089.

[0012] The present invention is further directed to a method of diagnosing or monitoring transplant rejection, cardiac rejection, renal rejection, and rejection of any other organ or tissue, in a patient, comprising detecting the expression level of one or more genes in the patient to diagnose or monitor transplant rejection in the patient wherein the one or more genes include a nucleotide sequence selected from SEQ ID NO:3 SEQ ID NO:45 SEQ ID NO:90 SEQ ID NO:145 SEQ ID NO:157, SEQ ID NO:176, SEQ ID NO:180, SEQ ID NO:194, SEQ ID NO:229, SEQ ID NO:230, SEQ ID NO:248, SEQ ID NO:253, SEQ ID NO:258, SEQ ID NO:261, SEQ ID NO:262, SEQ ID NO:269, SEQ ID NO:272, SEQ ID NO:281, SEQ ID NO:500, SEQ ID NO:573, SEQ ID NO:792, SEQ ID NO:802, SEQ ID NO:812, SEQ ID NO:841, SEQ ID NO:873, SEQ ID NO:1154, SEQ ID NO:1295, SEQ ID NO:1398, SEQ ID NO:1540, SEQ ID NO:1675, SEQ ID NO:1696, SEQ ID NO:1718, SEQ ID NO:1754, SEQ ID NO:1771, SEQ ID NO:1778, SEQ ID NO:1825, SEQ ID NO:1854, SEQ ID NO:1907, SEQ ID NO:1944, SEQ ID NO:1956, SEQ ID NO:1975, SEQ ID NO:1977, SEQ ID NO:1978, SEQ ID NO:1984, SEQ ID NO:1991, SEQ ID NO:2005, SEQ ID NO:2006, SEQ ID NO:2067, SEQ ID NO:2086, SEQ ID NO:2148, SEQ ID NO:2246, SEQ ID NO:2268, SEQ ID NO:2287, SEQ ID NO:2429, SEQ ID NO:2518, SEQ ID NO:2519, SEQ ID NO:2563, SEQ ID NO:2639, SEQ ID NO:2691, SEQ ID NO:2712, SEQ ID NO:2786, SEQ ID NO:2855, SEQ ID NO:2894, SEQ ID NO:2947, SEQ ID NO:2964, SEQ ID NO:3034, SEQ ID NO:3081, SEQ ID NO:3087, SEQ ID NO:3096, SEQ ID NO:3105, SEQ ID NO:3305, SEQ ID NO:3374, SEQ ID NO:3510, SEQ ID NO:3541, SEQ ID NO:3580, SEQ ID NO:3707, SEQ ID NO:3709, SEQ ID NO:3720, SEQ ID NO:4113, SEQ ID NO:4132, SEQ ID NO:4281, SEQ ID NO:4361, SEQ ID NO:4365, SEQ ID NO:4399, SEQ ID NO:4460, SEQ ID NO:4504, SEQ ID NO:4516, SEQ ID NO:4517, SEQ ID NO:4540, SEQ ID NO:4565, SEQ ID NO:4570, SEQ ID NO:4578, SEQ ID NO:4580, SEQ ID NO:4604, SEQ ID NO:4606, SEQ ID NO:4758, SEQ ID NO:4761, SEQ ID NO:4789, SEQ ID NO:4817, SEQ ID NO:4946, SEQ ID NO:5032, SEQ ID NO:5057, SEQ ID NO:5280, SEQ ID NO:5350, SEQ ID NO:5353, SEQ ID NO:5437, SEQ ID NO:5847, SEQ ID NO:6028, SEQ ID NO:6030, SEQ ID NO:6091, SEQ ID NO:6207, SEQ ID NO:6308, SEQ ID NO:6421, SEQ ID NO:6514, SEQ ID NO:6573, SEQ ID NO:6890, SEQ ID NO:6907, SEQ ID NO:7016, SEQ ID NO:7094, SEQ ID NO:7274, SEQ ID NO:7293, SEQ ID NO:7298, SEQ ID NO:7481, SEQ ID NO:7482, SEQ ID NO:7605, SEQ ID NO:8076, SEQ ID NO:9402, SEQ ID NO:9403, SEQ ID NO:9404, SEQ ID NO:9405, SEQ ID NO:9406, SEQ ID NO:9407, and SEQ ID NO:9408.

[0013] The present invention is further directed to a method of diagnosing or monitoring transplant rejection, cardiac rejection, renal rejection, and rejection of any other organ or tissue, in a patient, comprising detecting the expression level of two or more genes in the patient to diagnose or monitor transplant rejection in the patient wherein the one or more genes include a nucleotide sequence selected from SEQ ID NO:3 SEQ ID NO:45 SEQ ID NO:90 SEQ ID NO:145 SEQ ID NO:157, SEQ ID NO:176, SEQ ID NO:180, SEQ ID NO:194, SEQ ID NO:229, SEQ ID NO:230, SEQ ID NO:248, SEQ ID NO:253, SEQ ID NO:258, SEQ ID NO:261, SEQ ID NO:262, SEQ ID NO:269, SEQ ID NO:272, SEQ ID NO:281, SEQ ID NO:500, SEQ ID NO:573, SEQ ID NO:792, SEQ ID NO:802, SEQ ID NO:812, SEQ ID NO:841, SEQ ID NO:873, SEQ ID NO:1154, SEQ ID NO:1295, SEQ ID NO:1398, SEQ ID NO:1540, SEQ ID NO:1675, SEQ ID NO:1696, SEQ ID NO:1718, SEQ ID NO:1754, SEQ ID NO:1771, SEQ ID NO:1778, SEQ ID NO:1825, SEQ ID NO:1854, SEQ ID NO:1907, SEQ ID NO:1944, SEQ ID NO:1956, SEQ ID NO:1975, SEQ ID NO:1977, SEQ ID NO:1978, SEQ ID NO:1984, SEQ ID NO:1991, SEQ ID NO:2005, SEQ ID NO:2006, SEQ ID NO:2067, SEQ ID NO:2086, SEQ ID NO:2148, SEQ ID NO:2246, SEQ ID NO:2268, SEQ ID NO:2287, SEQ ID NO:2429, SEQ ID NO:2518, SEQ ID NO:2519, SEQ ID NO:2563, SEQ ID NO:2639, SEQ ID NO:2691, SEQ ID NO:2712, SEQ ID NO:2786, SEQ ID NO:2855, SEQ ID NO:2894, SEQ ID NO:2947, SEQ ID NO:2964, SEQ ID NO:3034, SEQ ID NO:3081, SEQ ID NO:3087, SEQ ID NO:3096, SEQ ID NO:3105, SEQ ID NO:3305, SEQ ID NO:3374, SEQ ID NO:3510, SEQ ID NO:3541, SEQ ID NO:3580, SEQ ID NO:3707, SEQ ID NO:3709, SEQ ID NO:3720, SEQ ID NO:4113, SEQ ID NO:4132, SEQ ID NO:4281, SEQ ID NO:4361, SEQ ID NO:4365, SEQ ID NO:4399, SEQ ID NO:4460, SEQ ID NO:4504, SEQ ID NO:4516, SEQ ID NO:4517, SEQ ID NO:4540, SEQ ID NO:4565, SEQ ID NO:4570, SEQ ID NO:4578, SEQ ID NO:4580, SEQ ID NO:4604, SEQ ID NO:4606, SEQ ID NO:4758, SEQ ID NO:4761, SEQ ID NO:4789, SEQ ID NO:4817, SEQ ID NO:4946, SEQ ID NO:5032, SEQ ID NO:5057, SEQ ID NO:5280, SEQ ID NO:5350, SEQ ID NO:5353, SEQ ID NO:5437, SEQ ID NO:5847, SEQ ID NO:6028, SEQ ID NO:6030, SEQ ID NO:6091, SEQ ID NO:6207, SEQ ID NO:6308, SEQ ID NO:6421, SEQ ID NO:6514, SEQ ID NO:6573, SEQ ID NO:6890, SEQ ID NO:6907, SEQ ID NO:7016, SEQ ID NO:7094, SEQ ID NO:7274, SEQ ID NO:7293, SEQ ID NO:7298, SEQ ID NO:7481, SEQ ID NO:7482, SEQ ID NO:7605, SEQ ID NO:8076, SEQ ID NO:9402, SEQ ID NO:9403, SEQ ID NO:9404, SEQ ID NO:9405, SEQ ID NO:9406, SEQ ID NO:9407, and SEQ ID NO:9408.

[0014] The present invention is further directed to a method of diagnosing or monitoring transplant rejection in a patient, comprising detecting the expression level of one or more genes in the patient to diagnose or monitor transplant rejection in the patient wherein the one or more genes include a nucleotide sequence selected from SEQ ID NO:3, SEQ ID NO:145, SEQ ID NO:194, SEQ ID NO:229, SEQ ID NO:230, SEQ ID NO:253, SEQ ID NO:258, SEQ ID NO:261, SEQ ID NO:269, SEQ ID NO:281, SEQ ID NO:500, SEQ ID NO:573, SEQ ID NO:792, SEQ ID NO:802, SEQ ID NO:812, SEQ ID NO:841, SEQ ID NO:873, SEQ ID NO:1154, SEQ ID NO:1295, SEQ ID NO:1398, SEQ ID NO:1540, SEQ ID NO:1675, SEQ ID NO:1696, SEQ ID NO:1718, SEQ ID NO:1754, SEQ ID NO:1771, SEQ ID NO:1778, SEQ ID NO:1854, SEQ ID NO:1907, SEQ ID NO:1944, SEQ ID NO:1956, SEQ ID NO:1977, SEQ ID NO:1978, SEQ ID NO:1991, SEQ ID NO:2067, SEQ ID NO:2086, SEQ ID NO:2148, SEQ ID NO:2246, SEQ ID NO:2268, SEQ ID NO:2287, SEQ ID NO:2429, SEQ ID NO:2518, SEQ ID NO:2519, SEQ ID NO:2563, SEQ ID NO:2639, SEQ ID NO:2691, SEQ ID NO:2712, SEQ ID NO:2786, SEQ ID NO:2894, SEQ ID NO:2947, SEQ ID NO:2964, SEQ ID NO:3034, SEQ ID NO:3081, SEQ ID NO:3087, SEQ ID NO:3096, SEQ ID NO:3105, SEQ ID NO:3305, SEQ ID NO:3374, SEQ ID NO:3510, SEQ ID NO:3541, SEQ ID NO:3580, SEQ ID NO:3707, SEQ ID NO:3720, SEQ ID NO:4113, SEQ ID NO:4281, SEQ ID NO:4361, SEQ ID NO:4365, SEQ ID NO:4399, SEQ ID NO:4460, SEQ ID NO:4504, SEQ ID NO:4516, SEQ ID NO:4517, SEQ ID NO:4540, SEQ ID NO:4565, SEQ ID NO:4570, SEQ ID NO:4578, SEQ ID NO:4580, SEQ ID NO:4604, SEQ ID NO:4606, SEQ ID NO:4758, SEQ ID NO:4761, SEQ ID NO:4789, SEQ ID NO:4817, SEQ ID NO:4946, SEQ ID NO:5032, SEQ ID NO:5057, SEQ ID NO:5280, SEQ ID NO:5350, SEQ ID NO:5353, SEQ ID NO:5437, SEQ ID NO:5847, SEQ ID NO:6028, SEQ ID NO:6030, SEQ ID NO:6091, SEQ ID NO:6207, SEQ ID NO:6308, SEQ ID NO:6421, SEQ ID NO:6514, SEQ ID NO:6573, SEQ ID NO:6890, SEQ ID NO:6907, SEQ ID NO:7016, SEQ ID NO:7094, SEQ ID NO:7274, SEQ ID NO:7293, SEQ ID NO:7298, SEQ ID NO:7481, SEQ ID NO:7482, SEQ ID NO:7605, SEQ ID NO:8076, SEQ ID NO:9403, SEQ ID NO:9404, SEQ ID NO:9405, SEQ ID NO:9406, SEQ ID NO:9407, and SEQ ID NO:9408. The invention may further comprise detecting the expression level of one or more additional genes in the patient to diagnose or monitor transplant rejection in a patient, wherein the one or more additional genes include a nucleic acid sequence selected from SEQ ID NO:45, SEQ ID NO:90, SEQ ID NO:157, SEQ ID NO:176, SEQ ID NO:180, SEQ ID NO:248, SEQ ID NO:262, SEQ ID NO:272, SEQ ID NO:1825, SEQ ID NO:1975, SEQ ID NO:1984, SEQ ID NO:2005, SEQ ID NO:2006, SEQ ID NO:2855, SEQ ID NO:3709, SEQ ID NO:4132, and SEQ ID NO:9402.

[0015] The present invention is further directed to detecting the expression level of one or more proteins expressed by the one or more genes. The proteins may include an amino acid sequence selected from SEQ ID NO:9314, SEQ ID NO:9315, SEQ ID NO:9316, SEQ ID NO:9318, SEQ ID NO:9319, SEQ ID NO:9321, SEQ ID NO:9323, SEQ ID NO:9324, SEQ ID NO:9325, SEQ ID NO:9326, SEQ ID NO:9327, SEQ ID NO:9328, SEQ ID NO:9329, SEQ ID NO:9330, SEQ ID NO:9331, SEQ ID NO:9332, SEQ ID NO:9333, SEQ ID NO:9334, SEQ ID NO:9336, SEQ ID NO:9337, SEQ ID NO:9338, SEQ ID NO:9340, SEQ ID NO:9341, SEQ ID NO:9343, SEQ ID NO:9346, SEQ ID NO:9347, SEQ ID NO:9348, SEQ ID NO:9349, SEQ ID NO:9350, SEQ ID NO:9351, SEQ ID NO:9352, SEQ ID NO:9353, SEQ ID NO:9354, SEQ ID NO:9355, SEQ ID NO:9356, SEQ ID NO:9357, SEQ ID NO:9359, SEQ ID NO:9360, SEQ ID NO:9361, SEQ ID NO:9362, SEQ ID NO:9363, SEQ ID NO:9364, SEQ ID NO:9365, SEQ ID NO:9366, SEQ ID NO:9367, SEQ ID NO:9368, SEQ ID NO:9369, SEQ ID NO:9370, SEQ ID NO:9371, SEQ ID NO:9373, SEQ ID NO:9374, SEQ ID NO:9376, SEQ ID NO:9377, SEQ ID NO:9378, SEQ ID NO:9379, SEQ ID NO:9380, SEQ ID NO:9381, SEQ ID NO:9382, SEQ ID NO:9383, SEQ ID NO:9384, SEQ ID NO:9385, SEQ ID NO:9386, SEQ ID NO:9387, SEQ ID NO:9388, SEQ ID NO:9389, SEQ ID NO:9390, SEQ ID NO:9391, SEQ ID NO:9392, SEQ ID NO:9393, SEQ ID NO:9394, SEQ ID NO:9395, SEQ ID NO:9397, SEQ ID NO:9398, SEQ ID NO:9399, SEQ ID NO:9400, and SEQ ID NO:9401. The present invention may further comprise detecting the level of one or more proteins expressed by the one or more additional genes. The proteins may include an amino acid sequence selected from SEQ ID NO:9309, SEQ ID NO:9310, SEQ ID NO:9311, SEQ ID NO:9312, SEQ ID NO:9313, SEQ ID NO:9317, SEQ ID NO:9320, SEQ ID NO:9322, SEQ ID NO:9335, SEQ ID NO:9339, SEQ ID NO:9342, SEQ ID NO:9344, SEQ ID NO:9345, SEQ ID NO:9358, SEQ ID NO:9372, SEQ ID NO:9375, and SEQ ID NO:9396.

[0016] The present invention is further directed to a method of diagnosing or monitoring cardiac transplant rejection in a patient, comprising detecting the expression level of one or more genes in the patient to diagnose or monitor cardiac transplant rejection in the patient, wherein the one or more genes include a nucleotide sequence selected from SEQ ID NO:3, SEQ ID NO:145, SEQ ID NO:194, SEQ ID NO:229, SEQ ID NO:230, SEQ ID NO:253, SEQ ID NO:258, SEQ ID NO:261, SEQ ID NO:262, SEQ ID NO:269, SEQ ID NO:272, SEQ ID NO:281, SEQ ID NO:500, SEQ ID NO:573, SEQ ID NO:792, SEQ ID NO:802, SEQ ID NO:812, SEQ ID NO:841, SEQ ID NO:873, SEQ ID NO:1154, SEQ ID NO:1295, SEQ ID NO:1398, SEQ ID NO:1540, SEQ ID NO:1675, SEQ ID NO:1696, SEQ ID NO:1718, SEQ ID NO:1754, SEQ ID NO:1771, SEQ ID NO:1778, SEQ ID NO:1854, SEQ ID NO:1907, SEQ ID NO:1944, SEQ ID NO:1956, SEQ ID NO:1975, SEQ ID NO:1977, SEQ ID NO:1978, SEQ ID NO:1991, SEQ ID NO:2005, SEQ ID NO:2067, SEQ ID NO:2086, SEQ ID NO:2148, SEQ ID NO:2246, SEQ ID NO:2268, SEQ ID NO:2287, SEQ ID NO:2429, SEQ ID NO:2518, SEQ ID NO:2519, SEQ ID NO:2563, SEQ ID NO:2639, SEQ ID NO:2691, SEQ ID NO:2712, SEQ ID NO:2786, SEQ ID NO:2894, SEQ ID NO:2947, SEQ ID NO:2964, SEQ ID NO:3034, SEQ ID NO:3081, SEQ ID NO:3087, SEQ ID NO:3096, SEQ ID NO:3105, SEQ ID NO:3305, SEQ ID NO:3374, SEQ ID NO:3510, SEQ ID NO:3541, SEQ ID NO:3580, SEQ ID NO:3707, SEQ ID NO:3709, SEQ ID NO:3720, SEQ ID NO:4113, SEQ ID NO:4281, SEQ ID NO:4361, SEQ ID NO:4365, SEQ ID NO:4399, SEQ ID NO:4460, SEQ ID NO:4504, SEQ ID NO:4516, SEQ ID NO:4517, SEQ ID NO:4540, SEQ ID NO:4565, SEQ ID NO:4570, SEQ ID NO:4578, SEQ ID NO:4580, SEQ ID NO:4604, SEQ ID NO:4606, SEQ ID NO:4758, SEQ ID NO:4761, SEQ ID NO:4789, SEQ ID NO:4817, SEQ ID NO:4946, SEQ ID NO:5032, SEQ ID NO:5057, SEQ ID NO:5280, SEQ ID NO:5350, SEQ ID NO:5353, SEQ ID NO:5437, SEQ ID NO:5847, SEQ ID NO:6028, SEQ ID NO:6030, SEQ ID NO:6091, SEQ ID NO:6207, SEQ ID NO:6308, SEQ ID NO:6421, SEQ ID NO:6514, SEQ ID NO:6573, SEQ ID NO:6890, SEQ ID NO:6907, SEQ ID NO:7016, SEQ ID NO:7094, SEQ ID NO:7274, SEQ ID NO:7293, SEQ ID NO:7298, SEQ ID NO:7481, SEQ ID NO:7482, SEQ ID NO:7605, SEQ ID NO:8076, SEQ ID NO:9402, SEQ ID NO:9403, SEQ ID NO:9404, SEQ ID NO:9405, SEQ ID NO:9406, SEQ ID NO:9407, and SEQ ID NO:9408. The present invention may further comprise detecting the expression level of one or more additional genes in the patient to diagnose or monitor transplant rejection in a patient, wherein the one or more additional genes include a nucleotide sequence selected from SEQ ID NO:45, SEQ ID NO:90, SEQ ID NO:157, SEQ ID NO:176, SEQ ID NO:180, SEQ ID NO:248, SEQ ID NO:1825, SEQ ID NO:1984, SEQ ID NO:2006, SEQ ID NO:2855, and SEQ ID NO:4132.

[0017] The present invention is further directed detecting the expression level of one or more proteins expressed by the one or more genes. The proteins may include an amino acid sequence selected from SEQ ID NO:9314, SEQ ID NO:9315, SEQ ID NO:9316, SEQ ID NO:9318, SEQ ID NO:9319, SEQ ID NO:9320, SEQ ID NO:9321, SEQ ID NO:9322, SEQ ID NO:9323, SEQ ID NO:9324, SEQ ID NO:9325, SEQ ID NO:9326, SEQ ID NO:9327, SEQ ID NO:9328, SEQ ID NO:9329, SEQ ID NO:9330, SEQ ID NO:9331, SEQ ID NO:9332, SEQ ID NO:9333, SEQ ID NO:9334, SEQ ID NO:9336, SEQ ID NO:9337, SEQ ID NO:9338, SEQ ID NO:9339, SEQ ID NO:9340, SEQ ID NO:9341, SEQ ID NO:9343, SEQ ID NO:9344, SEQ ID NO:9346, SEQ ID NO:9347, SEQ ID NO:9348, SEQ ID NO:9349, SEQ ID NO:9350, SEQ ID NO:9351, SEQ ID NO:9352, SEQ ID NO:9353, SEQ ID NO:9354, SEQ ID NO:9355, SEQ ID NO:9356, SEQ ID NO:9357, SEQ ID NO:9359, SEQ ID NO:9360, SEQ ID NO:9361, SEQ ID NO:9362, SEQ ID NO:9363, SEQ ID NO:9364, SEQ ID NO:9365, SEQ ID NO:9366, SEQ ID NO:9367, SEQ ID NO:9368, SEQ ID NO:9369, SEQ ID NO:9370, SEQ ID NO:9371, SEQ ID NO:9372, SEQ ID NO:9373, SEQ ID NO:9374, SEQ ID NO:9376, SEQ ID NO:9377, SEQ ID NO:9378, SEQ ID NO:9379, SEQ ID NO:9380, SEQ ID NO:9381, SEQ ID NO:9382, SEQ ID NO:9383, SEQ ID NO:9384, SEQ ID NO:9385, SEQ ID NO:9386, SEQ ID NO:9387, SEQ ID NO:9388, SEQ ID NO:9389, SEQ ID NO:9390, SEQ ID NO:9391, SEQ ID NO:9392, SEQ ID NO:9393, SEQ ID NO:9394, SEQ ID NO:9395, SEQ ID NO:9396, SEQ ID NO:9397, SEQ ID NO:9398, SEQ ID NO:9399, SEQ ID NO:9400, and SEQ ID NO:9401. The present invention may further comprise detecting the level of one or more proteins expressed by the one or more additional genes. The proteins may include an amino acid sequence selected from SEQ ID NO:9309, SEQ ID NO:9310, SEQ ID NO:9311, SEQ ID NO:9312, SEQ ID NO:9313, SEQ ID NO:9317, SEQ ID NO:9335, SEQ ID NO:9342, SEQ ID NO:9345, SEQ ID NO:9358, and SEQ ID NO:9375.

[0018] The present invention is further directed to a method of diagnosing or monitoring kidney transplant rejection in a patient, comprising detecting the expression level of one or more genes in the patient to diagnose or monitor transplant rejection in the patient wherein the one or more genes include a nucleotide sequence selected from SEQ ID NO:3, SEQ ID NO:145, SEQ ID NO:180, SEQ ID NO:194, SEQ ID NO:229, SEQ ID NO:230, SEQ ID NO:248, SEQ ID NO:253, SEQ ID NO:258, SEQ ID NO:261, SEQ ID NO:269, SEQ ID NO:281, SEQ ID NO:500, SEQ ID NO:573, SEQ ID NO:792, SEQ ID NO:802, SEQ ID NO:812, SEQ ID NO:841, SEQ ID NO:873, SEQ ID NO:1154, SEQ ID NO:1295, SEQ ID NO:1398, SEQ ID NO:1540, SEQ ID NO:1675, SEQ ID NO:1696, SEQ ID NO:1718, SEQ ID NO:1754, SEQ ID NO:1771, SEQ ID NO:1778, SEQ ID NO:1854, SEQ ID NO:1907, SEQ ID NO:1944, SEQ ID NO:1956, SEQ ID NO:1977, SEQ ID NO:1978, SEQ ID NO:1991, SEQ ID NO:2067, SEQ ID NO:2086, SEQ ID NO:2148, SEQ ID NO:2246, SEQ ID NO:2268, SEQ ID NO:2287, SEQ ID NO:2429, SEQ ID NO:2518, SEQ ID NO:2519, SEQ ID NO:2563, SEQ ID NO:2639, SEQ ID NO:2691, SEQ ID NO:2712, SEQ ID NO:2786, SEQ ID NO:2855, SEQ ID NO:2894, SEQ ID NO:2947, SEQ ID NO:2964, SEQ ID NO:3034, SEQ ID NO:3081, SEQ ID NO:3087, SEQ ID NO:3096, SEQ ID NO:3105, SEQ ID NO:3305, SEQ ID NO:3374, SEQ ID NO:3510, SEQ ID NO:3541, SEQ ID NO:3580, SEQ ID NO:3707, SEQ ID NO:3720, SEQ ID NO:4113, SEQ ID NO:4281, SEQ ID NO:4361, SEQ ID NO:4365, SEQ ID NO:4399, SEQ ID NO:4460, SEQ ID NO:4504, SEQ ID NO:4516, SEQ ID NO:4517, SEQ ID NO:4540, SEQ ID NO:4565, SEQ ID NO:4570, SEQ ID NO:4578, SEQ ID NO:4580, SEQ ID NO:4604, SEQ ID NO:4606, SEQ ID NO:4758, SEQ ID NO:4761, SEQ ID NO:4789, SEQ ID NO:4817, SEQ ID NO:4946, SEQ ID NO:5032, SEQ ID NO:5057, SEQ ID NO:5280, SEQ ID NO:5350, SEQ ID NO:5353, SEQ ID NO:5437, SEQ ID NO:5847, SEQ ID NO:6028, SEQ ID NO:6030, SEQ ID NO:6091, SEQ ID NO:6207, SEQ ID NO:6308, SEQ ID NO:6421, SEQ ID NO:6514, SEQ ID NO:6573, SEQ ID NO:6890, SEQ ID NO:6907, SEQ ID NO:7016, SEQ ID NO:7094, SEQ ID NO:7274, SEQ ID NO:7293, SEQ ID NO:7298, SEQ ID NO:7481, SEQ ID NO:7482, SEQ ID NO:7605, SEQ ID NO:8076, SEQ ID NO:9403, SEQ ID NO:9404, SEQ ID NO:9405, SEQ ID NO:9406, SEQ ID NO:9407, and SEQ ID NO:9408. The present invention may further comprise detecting the expression level of one or more additional genes in the patient to diagnose or monitor transplant rejection in a patient, wherein the one or more additional genes include a nucleotide sequence selected from SEQ ID NO:45, SEQ ID NO:90, SEQ ID NO:157, SEQ ID NO:176, SEQ ID NO:262, SEQ ID NO:272, SEQ ID NO:1825, SEQ ID NO:1975, SEQ ID NO:1984, SEQ ID NO:2005, SEQ ID NO:2006, SEQ ID NO:3709, SEQ ID NO:4132, and SEQ ID NO:9402.

[0019] The present invention is further directed to detecting the expression level by measuring the level of one or more proteins expressed by the one or more genes. The proteins may include an amino acid sequence selected from SEQ ID NO:9314, SEQ ID NO:9315, SEQ ID NO:9316, SEQ ID NO:9313, SEQ ID NO:9317, SEQ ID NO:9318, SEQ ID NO:9319, SEQ ID NO:9321, SEQ ID NO:9323, SEQ ID NO:9324, SEQ ID NO:9325, SEQ ID NO:9326, SEQ ID NO:9327, SEQ ID NO:9328, SEQ ID NO:9329, SEQ ID NO:9330, SEQ ID NO:9331, SEQ ID NO:9332, SEQ ID NO:9333, SEQ ID NO:9334, SEQ ID NO:9336, SEQ ID NO:9337, SEQ ID NO:9338, SEQ ID NO:9340, SEQ ID NO:9341, SEQ ID NO:9343, SEQ ID NO:9346, SEQ ID NO:9347, SEQ ID NO:9348, SEQ ID NO:9349, SEQ ID NO:9350, SEQ ID NO:9351, SEQ ID NO:9352, SEQ ID NO:9353, SEQ ID NO:9354, SEQ ID NO:9355, SEQ ID NO:9356, SEQ ID NO:9357, SEQ ID NO:9358, SEQ ID NO:9359, SEQ ID NO:9360, SEQ ID NO:9361, SEQ ID NO:9362, SEQ ID NO:9363, SEQ ID NO:9364, SEQ ID NO:9365, SEQ ID NO:9366, SEQ ID NO:9367, SEQ ID NO:9368, SEQ ID NO:9369, SEQ ID NO:9370, SEQ ID NO:9371, SEQ ID NO:9373, SEQ ID NO:9374, SEQ ID NO:9376, SEQ ID NO:9377, SEQ ID NO:9378, SEQ ID NO:9379, SEQ ID NO:9380, SEQ ID NO:9381, SEQ ID NO:9382, SEQ ID NO:9383, SEQ ID NO:9384, SEQ ID NO:9385, SEQ ID NO:9386, SEQ ID NO:9387, SEQ ID NO:9388, SEQ ID NO:9389, SEQ ID NO:9390, SEQ ID NO:9391, SEQ ID NO:9392, SEQ ID NO:9393, SEQ ID NO:9394, SEQ ID NO:9395, SEQ ID NO:9397, SEQ ID NO:9398, SEQ ID NO:9399, SEQ ID NO:9400, and SEQ ID NO:9401. The present invention may further comprise detecting the level of one or more proteins expressed by the one or more additional genes. The proteins may include an amino acid sequence selected from SEQ ID NO:9309, SEQ ID NO:9310, SEQ ID NO:9311, SEQ ID NO:9312, SEQ ID NO:9320, SEQ ID NO:9322, SEQ ID NO:9335, SEQ ID NO:9339, SEQ ID NO:9342, SEQ ID NO:9344, SEQ ID NO:9345, SEQ ID NO:9372, SEQ ID NO:9375, and SEQ ID NO:9396.

[0020] According to the present invention, the proteins may be measured by any means known in the art. The proteins may be from any tissue or bodily fluid, including serum. According to the present invention, the protein may be a cell surface protein. According to the present invention, the protein may also be measured using a fluorescent activated cell sorter.

[0021] The present invention is further directed to oligonucleotide having the nucleotide sequence selected from SEQ ID NO:3 SEQ ID NO:45 SEQ ID NO:90 SEQ ID NO:145 SEQ ID NO:157, SEQ ID NO:176, SEQ ID NO:180, SEQ ID NO:194, SEQ ID NO:229, SEQ ID NO:230, SEQ ID NO:248, SEQ ID NO:253, SEQ ID NO:258, SEQ ID NO:261, SEQ ID NO:262, SEQ ID NO:269, SEQ ID NO:272, SEQ ID NO:281, SEQ ID NO:500, SEQ ID NO:573, SEQ ID NO:792, SEQ ID NO:802, SEQ ID NO:812, SEQ ID NO:841, SEQ ID NO:873, SEQ ID NO:1154, SEQ ID NO:1295, SEQ ID NO:1398, SEQ ID NO:1540, SEQ ID NO:1675, SEQ ID NO:1696, SEQ ID NO:1718, SEQ ID NO:1754, SEQ ID NO:1771, SEQ ID NO:1778, SEQ ID NO:1825, SEQ ID NO:1854, SEQ ID NO:1907, SEQ ID NO:1944, SEQ ID NO:1956, SEQ ID NO:1975, SEQ ID NO:1977, SEQ ID NO:1978, SEQ ID NO:1984, SEQ ID NO:1991, SEQ ID NO:2005, SEQ ID NO:2006, SEQ ID NO:2067, SEQ ID NO:2086, SEQ ID NO:2148, SEQ ID NO:2246, SEQ ID NO:2268, SEQ ID NO:2287, SEQ ID NO:2429, SEQ ID NO:2518, SEQ ID NO:2519, SEQ ID NO:2563, SEQ ID NO:2639, SEQ ID NO:2691, SEQ ID NO:2712, SEQ ID NO:2786, SEQ ID NO:2855, SEQ ID NO:2894, SEQ ID NO:2947, SEQ ID NO:2964, SEQ ID NO:3034, SEQ ID NO:3081, SEQ ID NO:3087, SEQ ID NO:3096, SEQ ID NO:3105, SEQ ID NO:3305, SEQ ID NO:3374, SEQ ID NO:3510, SEQ ID NO:3541, SEQ ID NO:3580, SEQ ID NO:3707, SEQ ID NO:3709, SEQ ID NO:3720, SEQ ID NO:4113, SEQ ID NO:4132, SEQ ID NO:4281, SEQ ID NO:4361, SEQ ID NO:4365, SEQ ID NO:4399, SEQ ID NO:4460, SEQ ID NO:4504, SEQ ID NO:4516, SEQ ID NO:4517, SEQ ID NO:4540, SEQ ID NO:4565, SEQ ID NO:4570, SEQ ID NO:4578, SEQ ID NO:4580, SEQ ID NO:4604, SEQ ID NO:4606, SEQ ID NO:4758, SEQ ID NO:4761, SEQ ID NO:4789, SEQ ID NO:4817, SEQ ID NO:4946, SEQ ID NO:5032, SEQ ID NO:5057, SEQ ID NO:5280, SEQ ID NO:5350, SEQ ID NO:5353, SEQ ID NO:5437, SEQ ID NO:5847, SEQ ID NO:6028, SEQ ID NO:6030, SEQ ID NO:6091, SEQ ID NO:6207, SEQ ID NO:6308, SEQ ID NO:6421, SEQ ID NO:6514, SEQ ID NO:6573, SEQ ID NO:6890, SEQ ID NO:6907, SEQ ID NO:7016, SEQ ID NO:7094, SEQ ID NO:7274, SEQ ID NO:7293, SEQ ID NO:7298, SEQ ID NO:7481, SEQ ID NO:7482, SEQ ID NO:7605, SEQ ID NO:8076, SEQ ID NO:9402, SEQ ID NO:9403, SEQ ID NO:9404, SEQ ID NO:9405, SEQ ID NO:9406, SEQ ID NO:9407, and SEQ ID NO:9408.

[0022] The present invention is further directed to oligonucleotides having the nucleotide sequence including SEQ ID NO:9409, SEQ ID NO:9410, SEQ ID NO:9411, SEQ ID NO:9412, SEQ ID NO:9413, SEQ ID NO:9414, SEQ ID NO:9415, SEQ ID NO:9416, SEQ ID NO:9417, SEQ ID NO:9418, SEQ ID NO:9419, SEQ ID NO:9420, SEQ ID NO:9421, SEQ ID NO:9422, SEQ ID NO:9423, SEQ ID NO:9424, SEQ ID NO:9425, SEQ ID NO:9426, SEQ ID NO:9427, SEQ ID NO:9428, SEQ ID NO:9429, SEQ ID NO:9430, SEQ ID NO:9431, SEQ ID NO:9432, SEQ ID NO:9433, SEQ ID NO:9434, SEQ ID NO:9435, SEQ ID NO:9436, SEQ ID NO:9437, SEQ ID NO:9438, SEQ ID NO:9439, SEQ ID NO:9440, SEQ ID NO:9441, SEQ ID NO:9442, SEQ ID NO:9443, SEQ ID NO:9444, SEQ ID NO:9445, SEQ ID NO:9446, SEQ ID NO:9447, SEQ ID NO:9448, SEQ ID NO:9449, SEQ ID NO:9450, SEQ ID NO:9451, SEQ ID NO:9452, SEQ ID NO:9453, SEQ ID NO:9454, SEQ ID NO:9455, SEQ ID NO:9456, SEQ ID NO:9457, SEQ ID NO:9458, SEQ ID NO:9459, SEQ ID NO:9460, SEQ ID NO:9461, SEQ ID NO:9462, SEQ ID NO:9463, SEQ ID NO:9464, SEQ ID NO:9465, SEQ ID NO:9466, SEQ ID NO:9467, SEQ ID NO:9468, SEQ ID NO:9469, SEQ ID NO:9470, SEQ ID NO:9471, SEQ ID NO:9472, SEQ ID NO:9473, SEQ ID NO:9474, SEQ ID NO:9475, SEQ ID NO:9476, SEQ ID NO:9477, SEQ ID NO:9478, SEQ ID NO:9479, SEQ ID NO:9480, SEQ ID NO:9481, SEQ ID NO:9482, SEQ ID NO:9483, SEQ ID NO:9484, SEQ ID NO:9485, SEQ ID NO:9486, SEQ ID NO:9487, SEQ ID NO:9488, SEQ ID NO:9489, SEQ ID NO:9490, SEQ ID NO:9491, SEQ ID NO:9492, SEQ ID NO:9493, SEQ ID NO:9494, SEQ ID NO:9495, SEQ ID NO:9496, SEQ ID NO:9497, SEQ ID NO:9498, SEQ ID NO:9499, SEQ ID NO:9500, SEQ ID NO:9501, SEQ ID NO:9502, SEQ ID NO:9503, SEQ ID NO:9504, SEQ ID NO:9505, SEQ ID NO:9506, SEQ ID NO:9507, SEQ ID NO:9508, SEQ ID NO:9509, SEQ ID NO:9510, SEQ ID NO:9511, SEQ ID NO:9512, SEQ ID NO:9513, SEQ ID NO:9514, SEQ ID NO:9515, SEQ ID NO:9516, SEQ ID NO:9517, SEQ ID NO:9518, SEQ ID NO:9519, SEQ ID NO:9520, SEQ ID NO:9521, SEQ ID NO:9522, SEQ ID NO:9523, SEQ ID NO:9524, SEQ ID NO:9525, SEQ ID NO:9526, SEQ ID NO:9527, SEQ ID NO:9528, SEQ ID NO:9529, SEQ ID NO:9530, SEQ ID NO:9531, SEQ ID NO:9532, SEQ ID NO:9533, SEQ ID NO:9534, SEQ ID NO:9535, SEQ ID NO:9536, SEQ ID NO:9537, SEQ ID NO:9538, SEQ ID NO:9539, SEQ ID NO:9540, SEQ ID NO:9541, SEQ ID NO:9542, SEQ ID NO:9543, SEQ ID NO:9544, SEQ ID NO:9545, SEQ ID NO:9546, SEQ ID NO:9547, SEQ ID NO:9548, SEQ ID NO:9549, SEQ ID NO:9550, SEQ ID NO:9551, SEQ ID NO:9552, SEQ ID NO:9553, SEQ ID NO:9554, SEQ ID NO:9555, SEQ ID NO:9556, SEQ ID NO:9557, SEQ ID NO:9558, SEQ ID NO:9559, SEQ ID NO:9560, SEQ ID NO:9561, SEQ ID NO:9562, SEQ ID NO:9563, SEQ ID NO:9564, SEQ ID NO:9565, SEQ ID NO:9566, SEQ ID NO:9567, SEQ ID NO:9568, SEQ ID NO:9569, SEQ ID NO:9570, SEQ ID NO:9571, SEQ ID NO:9572, SEQ ID NO:9573, SEQ ID NO:9574, SEQ ID NO:9575, SEQ ID NO:9576, SEQ ID NO:9577, SEQ ID NO:9578, SEQ ID NO:9579, SEQ ID NO:9580, SEQ ID NO:9581, SEQ ID NO:9582, SEQ ID NO:9583, SEQ ID NO:9584, SEQ ID NO:9585, SEQ ID NO:9586, SEQ ID NO:9587, SEQ ID NO:9588, SEQ ID NO:9589, SEQ ID NO:9590, SEQ ID NO:9591, SEQ ID NO:9592, SEQ ID NO:9593, SEQ ID NO:9594, SEQ ID NO:9595, SEQ ID NO:9596, SEQ ID NO:9597, SEQ ID NO:9598, SEQ ID NO:9599, SEQ ID NO:9600, SEQ ID NO:9601, SEQ ID NO:9602, SEQ ID NO:9603, SEQ ID NO:9604, SEQ ID NO:9605, SEQ ID NO:9606, SEQ ID NO:9607, SEQ ID NO:9608, SEQ ID NO:9609, SEQ ID NO:9610, SEQ ID NO:9611, SEQ ID NO:9612, SEQ ID NO:9613, SEQ ID NO:9614, SEQ ID NO:9615, SEQ ID NO:9616, SEQ ID NO:9617, SEQ ID NO:9618, SEQ ID NO:9619, SEQ ID NO:9620, SEQ ID NO:9621, SEQ ID NO:9622, SEQ ID NO:9623, SEQ ID NO:9624, SEQ ID NO:9625, SEQ ID NO:9626, SEQ ID NO:9627, SEQ ID NO:9628, SEQ ID NO:9629, SEQ ID NO:9630, SEQ ID NO:9631, SEQ ID NO:9632, SEQ ID NO:9633, SEQ ID NO:9634, SEQ ID NO:9635, SEQ ID NO:9636, SEQ ID NO:9637, SEQ ID NO:9638, SEQ ID NO:9639, SEQ ID NO:9640, SEQ ID NO:9641, SEQ ID NO:9642, SEQ ID NO:9643, SEQ ID NO:9644, SEQ ID NO:9645, SEQ ID NO:9646, SEQ ID NO:9647, SEQ ID NO:9648, SEQ ID NO:9649, SEQ ID NO:9650, SEQ ID NO:9651, SEQ ID NO:9652, SEQ ID NO:9653, SEQ ID NO:9654, SEQ ID NO:9655, SEQ ID NO:9656, SEQ ID NO:9657, SEQ ID NO:9658, SEQ ID NO:9659, SEQ ID NO:9660, SEQ ID NO:9661, SEQ ID NO:9662, SEQ ID NO:9663, SEQ ID NO:9664, SEQ ID NO:9665, SEQ ID NO:9666, SEQ ID NO:9667, SEQ ID NO:9668, SEQ ID NO:9669, SEQ ID NO:9670, SEQ ID NO:9671, SEQ ID NO:9672, SEQ ID NO:9673, SEQ ID NO:9674, SEQ ID NO:9675, SEQ ID NO:9676, SEQ ID NO:9677, SEQ ID NO:9678, SEQ ID NO:9679, SEQ ID NO:9680, SEQ ID NO:9681, SEQ ID NO:9682, SEQ ID NO:9683, SEQ ID NO:9684, SEQ ID NO:9685, SEQ ID NO:9686, SEQ ID NO:9687, SEQ ID NO:9688, SEQ ID NO:9689, SEQ ID NO:9690, SEQ ID NO:9691, SEQ ID NO:9692, SEQ ID NO:9693, SEQ ID NO:9694, SEQ ID NO:9695, SEQ ID NO:9696, SEQ ID NO:9697, SEQ ID NO:9698, SEQ ID NO:9699, SEQ ID NO:9700, SEQ ID NO:9701, SEQ ID NO:9702, SEQ ID NO:9703, SEQ ID NO:9704, SEQ ID NO:9705, SEQ ID NO:9706, SEQ ID NO:9707, SEQ ID NO:9708, SEQ ID NO:9709, SEQ ID NO:9710, SEQ ID NO:9711, SEQ ID NO:9712, SEQ ID NO:9713, SEQ ID NO:9714, SEQ ID NO:9715, SEQ ID NO:9716, SEQ ID NO:9717, SEQ ID NO:9718, SEQ ID NO:9719, SEQ ID NO:9720, SEQ ID NO:9721, SEQ ID NO:9722, SEQ ID NO:9723, SEQ ID NO:9724, SEQ ID NO:9725, SEQ ID NO:9726, SEQ ID NO:9727, SEQ ID NO:9728, SEQ ID NO:9729, SEQ ID NO:9730, SEQ ID NO:9731, SEQ ID NO:9732, SEQ ID NO:9733, SEQ ID NO:9734, SEQ ID NO:9735, SEQ ID NO:9736, SEQ ID NO:9737, SEQ ID NO:9738, SEQ ID NO:9739, SEQ ID NO:9740, SEQ ID NO:9741, SEQ ID NO:9742, SEQ ID NO:9743, SEQ ID NO:9744, SEQ ID NO:9745, SEQ ID NO:9746, SEQ ID NO:9747, SEQ ID NO:9748, SEQ ID NO:9749, SEQ ID NO:9750, SEQ ID NO:9751, SEQ ID NO:9752, SEQ ID NO:9753, SEQ ID NO:9754, SEQ ID NO:9755, SEQ ID NO:9756, SEQ ID NO:9757, SEQ ID NO:9758, SEQ ID NO:9759, SEQ ID NO:9760, SEQ ID NO:9761, SEQ ID NO:9762, SEQ ID NO:9763, SEQ ID NO:9764, SEQ ID NO:9765, SEQ ID NO:9766, SEQ ID NO:9767, SEQ ID NO:9768, SEQ ID NO:9769, SEQ ID NO:9770, SEQ ID NO:9771, SEQ ID NO:9772, SEQ ID NO:9773, SEQ ID NO:9774, SEQ ID NO:9775, SEQ ID NO:9776, SEQ ID NO:9777, SEQ ID NO:9778, SEQ ID NO:9779, SEQ ID NO:9780, SEQ ID NO:9781, SEQ ID NO:9782, SEQ ID NO:9783, SEQ ID NO:9784, SEQ ID NO:9785, SEQ ID NO:9786, SEQ ID NO:9787, SEQ ID NO:9788, SEQ ID NO:9789, SEQ ID NO:9790, SEQ ID NO:9791, SEQ ID NO:9792, SEQ ID NO:9793, SEQ ID NO:9794, SEQ ID NO:9795, SEQ ID NO:9796, SEQ ID NO:9797, SEQ ID NO:9798, SEQ ID NO:9799, SEQ ID NO:9800, SEQ ID NO:9801, SEQ ID NO:9802, SEQ ID NO:9803, SEQ ID NO:9804, SEQ ID NO:9805, SEQ ID NO:9806, SEQ ID NO:9807, SEQ ID NO:9808, SEQ ID NO:9809, SEQ ID NO:9810, SEQ ID NO:9811, SEQ ID NO:9812, SEQ ID NO:9813, SEQ ID NO:9814, SEQ ID NO:9815, SEQ ID NO:9816, SEQ ID NO:9817, SEQ ID NO:9818, SEQ ID NO:9819, SEQ ID NO:9820, SEQ ID NO:9821, SEQ ID NO:9822, SEQ ID NO:9823, SEQ ID NO:9824, SEQ ID NO:9825, SEQ ID NO:9826, SEQ ID NO:9827, SEQ ID NO:9828, SEQ ID NO:9829, SEQ ID NO:9830, SEQ ID NO:9831, SEQ ID NO:9832, SEQ ID NO:9833, SEQ ID NO:9834, SEQ ID NO:9835, SEQ ID NO:9836, SEQ ID NO:9837, SEQ ID NO:9838, SEQ ID NO:9839, SEQ ID NO:9840, and SEQ ID NO:9841.

[0023] The present invention is also directed to oligonucleotides having at least 90% sequence identity to any of the oligonucleotides disclosed herein.

[0024] The present invention is also directed to oligonucleotides including DNA, cDNA, PNA, genomic DNA, or synthetic oligonucleotides.

[0025] The present invention is also directed to kits including any of the sequences listed herein.

[0026] The present invention is further directed to a method of immune monitoring, comprising detecting the expression level of one or more genes in a patient wherein the one or more genes include a nucleotide sequence listed in the attached sequence listing. The nucleotide sequence may be a sequence listed above for transplant rejection. The method of immune monitoring may further comprise detecting the expression level of one or more additional genes in the patient wherein the one or more additional genes include a nucleotide sequence selected from those listed for transplant rejection, cardiac rejection, or renal rejection. The expression level may be measured by detecting the level of one or more proteins expressed by the one or more genes or one or more additional genes. The one or more proteins corresponding to the one or more genes or one or more additional genes may be selected from the protein sequences listed above for transplant rejection, cardiac transplant rejection, or renal transplant rejection.

[0027] The present invention is further directed to a method of diagnosing or monitoring cytomegolovirus infection in a patient, by detecting the expression level of one or more genes or surrogates derived therefrom in the patient to diagnose or monitor cytomegolovirus infection in the patient wherein the genes include a nucleotide sequence selected from: SEQ ID NO:4; SEQ ID NO:26; SEQ ID NO:60; SEQ ID NO:130; SEQ ID NO:176; SEQ ID NO:184; SEQ ID NO:261; SEQ ID NO:707; SEQ ID NO:792; SEQ ID NO:841; SEQ ID NO:1024; SEQ ID NO:1128; SEQ ID NO:1140; SEQ ID NO:1333; SEQ ID NO:1345; SEQ ID NO:1435; SEQ ID NO:1749; SEQ ID NO:1778; SEQ ID NO:1956; SEQ ID NO:2086; SEQ ID NO:2228; SEQ ID NO:2518; SEQ ID NO:2519; SEQ ID NO:2770; SEQ ID NO:2801; SEQ ID NO:3134; SEQ ID NO:3263; SEQ ID NO:3842; SEQ ID NO:4092; SEQ ID NO:4191; SEQ ID NO:4460; SEQ ID NO:4515; SEQ ID NO:5108; SEQ ID NO:5280; SEQ ID NO:5573; SEQ ID NO:5673; SEQ ID NO:5834; SEQ ID NO:6091; SEQ ID NO:6112; SEQ ID NO:6221; SEQ ID NO:6309; SEQ ID NO:6347; SEQ ID NO:6514; SEQ ID NO:6573; SEQ ID NO:7094; SEQ ID NO:7199; SEQ ID NO:7481; SEQ ID NO:7482; SEQ ID NO:7605; SEQ ID NO:8076; SEQ ID NO:8089; SEQ ID NO: 4132; SEQ ID NO:4604; SEQ ID NO:2630; SEQ ID NO:3305; SEQ ID NO:3717; SEQ ID NO:5471; SEQ ID NO:5559; SEQ ID NO:6308; SEQ ID NO:1983; SEQ ID NO:4761; SEQ ID NO:5509; SEQ ID NO:2004; SEQ ID NO:1685; SEQ ID NO:2428; SEQ ID NO:4113; SEQ ID NO:6059; SEQ ID NO:1754 and SEQ ID NO:375.

[0028] The present invention is further directed to a diagnostic agent comprising an oligonucleotide wherein the oligonucleotide has a nucleotide sequence selected from the Sequence Listing wherein the oligonucleotide detects expression of a gene that is differentially expressed in leukocytes in an individual over time.

[0029] The present invention is further directed to a system for detecting gene expression in leukocytes comprising an isolated DNA molecule wherein the isolated DNA molecule detects expression of a gene wherein the gene is selected from the group of genes corresponding to the oligonucleotides depicted in the Sequence Listing and the gene is differentially expressed in the leukocytes in an individual with at least one disease criterion for a disease selected from Table 1 as compared to the expression of the gene in leukocytes in an individual without the at least one disease criterion.

[0030] The present invention is further directed to a gene expression candidate library comprising at least two oligonucleotides wherein the oligonucleotides have a sequence selected from those oligonucleotide sequences listed in Table 2, Table 3, and the Sequence Listing. Table 3 encompasses Tables 3A, 3B and 3C. The oligonucleotides of the candidate library may comprise deoxyribonucleic acid (DNA), ribonucleic acid (RNA), protein nucleic acid (PNA), synthetic oligonucleotides, or genomic DNA.

[0031] In one embodiment, the candidate library is immobilized on an array. The array may comprises one or more of: a chip array, a plate array, a bead array, a pin array, a membrane array, a solid surface array, a liquid array, an oligonucleotide array, a polynucleotide array or a cDNA array, a microtiter plate, a pin array, a bead array, a membrane or a chip. Individual members of the libraries are may be separately immobilized.

[0032] The present invention is further directed to a diagnostic oligonucleotide set for a disease having at least two oligonucleotides wherein the oligonucleotides have a sequence selected from those oligonucleotide sequences listed in Table 2, Table 3, or the Sequence Listing which are differentially expressed in leukocytes genes in an individual with at least one disease criterion for at least one leukocyte-related disease as compared to the expression in leukocytes in an individual without the at least one disease criterion, wherein expression of the two or more genes of the gene expression library is correlated with at least one disease criterion.

[0033] The present invention is further directed to a diagnostic oligonucleotide set for a disease having at least one oligonucleotide wherein the oligonucleotide has a sequence selected from those sequences listed in Table 2, Table 3, or the sequence listing which is differentially expressed in leukocytes in an individual with at least one disease criterion for a disease selected from Table 1 as compared toleukocytes in an individual without at least one disease criterion, wherein expression of the at least one gene from the gene expression library is correlated with at least one disease criterion, wherein the differential expression of the at least one gene has not previously been described. In one format, two or more oligonucleotides are utilized.

[0034] In the diagnostic oligonucleotide sets of the invention the disease criterion may include data selected from patient historic, diagnostic, prognostic, risk prediction, therapeutic progress, and therapeutic outcome data. This includes lab results, radiology results, pathology results such as histology, cytology and the like, physical examination findings, and medication lists.

[0035] In the diagnostic oligonucleotide sets of the invention the leukocytes comprise peripheral blood leukocytes or leukocytes derived from a non-blood fluid. The non-blood fluid may be selected from colon, sinus, spinal fluid, saliva, lymph fluid, esophagus, small bowel, pancreatic duct, biliary tree, ureter, vagina, cervix uterus and pulmonary lavage fluid.

[0036] In the diagnostic oligonucleotide sets of the invention the leukocytes may include leukocytes derived from urine or a joint biopsy sample or biopsy of any other tissue or may be T-lymphocytes.

[0037] In the diagnostic oligonucleotide sets of the invention the disease may be selected from cardiac allograft rejection, kidney allograft rejection, liver allograft rejection, atherosclerosis, congestive heart failure, systemic lupus erythematosis (SLE), rheumatoid arthritis, osteoarthritis, and cytomegalovirus infection.

[0038] The diagnostic oligonucleotide sets of the invention may further include one or more cytomegalovirus (CMV) nucleotide sequences, wherein expression of the CMV nucleotide sequence is correlated with CMV infection.

[0039] The diagnostic nucleotide sets of the invention may further include one or more Epstein-Barr virus (EBV) nucleotide sequences, wherein expression of the one or more EBV nucleotide sequences is correlated with EBV infection.

[0040] In the present invention, expression may be differential expression, wherein the differential expression is one or more of a relative increase in expression, a relative decrease in expression, presence of expression or absence of expression, presence of disease or absence of disease. The differential expression may be RNA expression or protein expression. The differential expression may be between two or more samples from the same patient taken on separate occasions or between two or more separate patients or between two or more genes relative to each other.

[0041] The present invention is further directed to a diagnostic probe set for a disease where the probes correspond to at least one oligonucleotide wherein the oligonucleotides have a sequence such as those listed in Table 2, Table 3, or the Sequence Listing which is differentially expressed in leukocytes in an individual with at least one disease criterion for a disease selected from Table 1 as compared to leukocytes in an individual without the at least one disease criterion, wherein expression of the oligonucleotide is correlated with at least one disease criterion, and further wherein the differential expression of the at least one nucleotide sequence has not previously been described.

[0042] The present invention is further directed to a diagnostic probe set wherein the probes include one or more of probes useful for proteomics and probes for nucleic acids cDNA, or synthetic oligonucleotides.

[0043] The present invention is further directed to an isolated nucleic acid having a sequences such as those listed in Table 3B or Table 3C or the Sequence Listing.

[0044] The present invention is further directed to polypeptides wherein the polypeptides are encoded by the nucleic acid sequences in Tables 3B, 3C and the Sequence Listing.

[0045] The present invention is further directed to a polynucleotide expression vector containing the polynucleotide of Tables 3B–3C or the Sequence Listing in operative association with a regulatory element which controls expression of the polynucleotide in a host cell. The present invention is further directed to host cells transformed with the expression vectors of the invention. The host cell may be prokaryotic or eukaryotic.

[0046] The present invention is further directed to fusion proteins produced by the host cells of the invention. The present invention is further directed to antibodies directed to the fusion proteins of the invention. The antibodies may be monoclonal or polyclonal antibodies.

[0047] The present invention is further directed to kits comprising the diagnostic oligonucleotide sets of the invention. The kits may include instructions for use of the kit.

[0048] The present invention is further directed to a method of diagnosing a disease by obtaining a leukocyte sample from an individual, hybridizing nucleic acid derived from the leukocyte sample with a diagnostic oligonucleotide set, and comparing the expression of the diagnostic oligonucleotide set with a molecular signature indicative of the presence or absence of the disease.

[0049] The present invention is further directed to a method of detecting gene expression by a) isolating RNA and b) hybridizing the RNA to isolated DNA molecules wherein the isolated DNA molecules detect expression of a gene wherein the gene corresponds to one of the oligonucleotides depicted in the Sequence Listing.

[0050] The present invention is further directed to a method of detecting gene expression by a) isolating RNA; b) converting the RNA to nucleic acid derived from the RNA and c) hybridizing the nucleic acid derived from the RNA to isolated DNA molecules wherein the isolated DNA molecules detect expression of a gene wherein the gene corresponds to one of the oligonucleotides depicted in the Sequence Listing. In one format, the nucleic acid derived from the RNA is cDNA.

[0051] The present invention is further directed to a method of detecting gene expression by a) isolating RNA; b) converting the RNA to cRNA or aRNA and c) hybridizing the cRNA or aRNA to isolated DNA molecules wherein the isolated DNA molecules detect expression of a gene corresponding to one of the oligonucleotides depicted in the Sequence Listing.

[0052] The present invention is further directed to a method of monitoring progression of a disease by obtaining a leukocyte sample from an individual, hybridizing the nucleic acid derived from leukocyte sample with a diagnostic oligonucleotide set, and comparing the expression of the diagnostic oligonucleotide set with a molecular signature indicative of the presence or absence of disease progression.

[0053] The present invention is further directed to a method of monitoring the rate of progression of a disease by obtaining a leukocyte sample from an individual, hybridizing the nucleic acid derived from leukocyte sample with a diagnostic oligonucleotide set, and comparing the expression of the diagnostic oligonucleotide set with a molecular signature indicative of the presence or absence of disease progression.

[0054] The present invention is further directed to a method of predicting therapeutic outcome by obtaining a leukocyte sample from an individual, hybridizing the nucleic acid derived from leukocyte sample with a diagnostic oligonucleotide set, and comparing the expression of the diagnostic oligonucleotide set with a molecular signature indicative of the predicted therapeutic outcome.

[0055] The present invention is further directed to a method of determining prognosis by obtaining a leukocyte sample from an individual, hybridizing the nucleic acid derived from leukocyte sample with a diagnostic oligonucleotide set, and comparing the expression of the diagnostic oligonucleotide set with a molecular signature indicative of the prognosis.

[0056] The present invention is further directed to a method of predicting disease complications by obtaining a leukocyte sample from an individual, hybridizing nucleic acid derived from the leukocyte sample with a diagnostic oligonucleotide set, and comparing the expression of the diagnostic oligonucleotide set with a molecular signature indicative of the presence or absence of disease complications.

[0057] The present invention is further directed to a method of monitoring response to treatment, by obtaining a leukocyte sample from an individual, hybridizing the nucleic acid derived from leukocyte sample with a diagnostic oligonucleotide set, and comparing the expression of the diagnostic oligonucleotide set with a molecular signature indicative of the presence or absence of response to treatment.

[0058] In the methods of the invention the invention may further include characterizing the genotype of the individual, and comparing the genotype of the individual with a diagnostic genotype, wherein the diagnostic genotype is correlated with at least one disease criterion. The genotype may be analyzed by one or more methods selected from the group consisting of Southern analysis, RFLP analysis, PCR, single stranded conformation polymorphism and SNP analysis.

[0059] The present invention is further directed to a method of non-invasive imaging by providing an imaging probe for a nucleotide sequence that is differentially expressed in leukocytes from an individual with at least one disease criterion for at least one leukocyte-implicated disease where leukocytes localize at the site of disease, wherein the expression of the at least one nucleotide sequence is correlated with the at least one disease criterion by (a) contacting the probe with a population of leukocytes; (b) allowing leukocytes to localize to the site of disease or injury and (c) detecting an image.

[0060] The present invention is further directed to a control RNA for use in expression profile analysis, where the RNA extracted from the buffy coat samples is from at least four individuals.

[0061] The present invention is further directed to a method of collecting expression profiles, comprising comparing the expression profile of an individual with the expression profile of buffy coat control RNA, and analyzing the profile.

[0062] The present invention is further directed to a method of RNA preparation suitable for diagnostic expression profiling by obtaining a leukocyte sample from a subject, adding actinomycin-D to a final concentration of 1 ug/ml, adding cycloheximide to a final concentration of 10 ug/ml, and extracting RNA from the leukocyte sample. In the method of RNA preparation of the invention the actinomycin-D and cycloheximide may be present in a sample tube to which the leukocyte sample is added. The method may further include centrifuging the sample at 4° C. to separate mononuclear cells.

[0063] The present invention is further directed to a leukocyte oligonucleotide set including at least two oligonucleotides which are differentially expressed in leukocytes undergoing adhesion to an endothelium relative to expression in leukocytes not undergoing adhesion to an endothelium, wherein expression of the two oligonucleotides is correlated with the at least one indicator of adhesion state.

[0064] The present invention is further directed to a method of identifying at least one diagnostic probe set for assessing atherosclerosis by (a) providing a library of candidate oligonucleotides, which candidate oligonucleotides are differentially expressed in leukocytes which are undergoing adhesion to an endothelium relative to their expression in leukocytes that are not undergoing adhesion to an endothelium; (b) assessing expression of two or more oligonucleotides, which two or more oligonucleotides correspond to components of the library of candidate oligonucleotides, in a subject sample of leukocytes; (c) correlating expression of the two or more oligonucleotides with at least one criterion, which criterion includes one or more indicators of adhesion to an endothelium; and, (d) recording the molecular signature in a database.

[0065] The present invention is further directed to a method of identifying at least one diagnostic probe set for assessing atherosclerosis by (a) providing a library of candidate oligonucleotides, which candidate oligonucleotides are differentially expressed in leukocytes which are undergoing adhesion to an endothelium relative to their expression in leukocytes that are not undergoing adhesion to an endothelium; (b) assessing expression of two or more oligonucleotides, which two or more oligonucleotides correspond to components of the library of candidate nucleotide sequences, in a subject sample of epithelial cells; (c) correlating expression of the two or more nucleotide sequences with at least one criterion, which criterion comprises one or more indicator of adhesion to an endothelium; and (d) recording the molecular signature in a database.

[0066] The present invention is further directed to methods of leukocyte expression profiling including methods of analyzing longitudinal clinical and expression data. The rate of change and/or magnitude and direction of change of gene expression can be correlated with disease states and the rate of change of clinical conditions/data and/or the magnitude and direction of changes in clinical data. Correlations may be discovered by examining these expression or clinical changes that are not found in the absence of such changes. The present invention is further directed to methods of leukocyte profiling for analysis and/or detection of one or more viruses. The virus may be CMV, HIV, hepatitis or other viruses. Both viral and human leukocyte genes can be subjected to expression profiling for these purposes.

[0067] The table below gives a description of the sequence listing. There are 9966 entries. The Sequence Listing presents 50mer oligonucleotide sequences derived from human leukocyte, plant and viral genes. These are listed as SEQ IDs 1–8143. The 50mer sequences and their sources are also displayed in Table 8. Most of these 50mers were designed from sequences of genes in Tables 2, 3A, B and C, Tables 8, 11–12, 14, 15, 19, and the Sequence listing.

[0068] SEQ IDs 8144–8766 are the cDNA sequences derived from human leukocytes that were not homologous to UniGene sequences or sequences found in dbEST at the time they were searched. Some of these sequences match human genomic sequences and are listed in Tables 3B and C. The remaining clones are putative cDNA sequences that contained less than 50% masked nucleotides when submitted to RepeatMasker, were longer than 147 nucleotides, and did not have significant similarity to the UniGene Unique database, dbEST, the NR nucleotide database of Genbank or the assembled human genome of Genbank.

[0069] SEQ IDs 8767–8770, 8828–8830 and 8832 are sequences that appear in the text and examples (primer, masked sequences, exemplary sequences, etc.).

[0070] SEQ IDs 8771–8827 are CMV PCR primers described in Example 17.

[0071] SEQ IDs 8833–9109 are PCR primers used to measure expression of genes in SEQ IDs 9109–9173 which were correlated with rejection.

[0072] SEQ IDs 9109–9173 are genes whose expression can be correlated with rejection.

[0073] SEQ IDs 9174–9183 are PCR primers to measure gene expression.

[0074] SEQ ID 9184 described in the exemplary sequences and is cDNA encoding Granzyme B.

[0075] SEQ IDs 9185–9190 are Primers.

[0076] SEQ IDs 9191–9308 are sequences of cDNAs of top gene markers whose expression is correlated with rejection.

[0077] SEQ IDs 9309–9401 are proteins encoded by cDNA sequences 9191–9308.

[0078] SEQ IDs 9409–9674 are PCR primers for sequences 9191–9308.

[0079] SEQ IDs 9675–9807 are Taqman probes for sequences defined by primers 9409–9674.

[0080] SEQ IDs 9808–9831 and 9842–9916 are additional PCR primers for sequences 9191–9308.

[0081] SEQ IDs 9917–9955 and 9832–9841 are additional Taqman probes for sequences defined by primers 9808–9831 and 9842–9916.

[0082] SEQ IDs 9956–9958 are sequence examples for describing the design of PCR primers and probes in the examples.

[0083] SEQ IDs 9959–9966 are the resultant PCR primers designed for sequences 9956–9958.

[0109] Table 1: Table 1 lists diseases or conditions amenable to study by leukocyte profiling.

[0110] Table 2: Table 2 describes genes and other nucleotide sequences identified using data mining of publicly available publication databases and nucleotide sequence databases. Corresponding Unigene (build 133) cluster numbers are listed with each gene or other nucleotide sequence.

[0111] Table 3A: Table 3A describes differentially expressed nucleotide sequences useful for the prediction of clinical outcomes. This table contains 4517 identified cDNAs and cDNA regions of genes that are members of a leukocyte candidate library, for use in measuring the expression of nucleotide sequences that could subsequently be correlated with human clinical conditions. The regions of similarity were found by searching three different databases for pair wise similarity using blastn. The three databases were UniGene Unique build Mar. 30, 2001, the downloadable NCBI human EST database with date Apr. 8, 2001 which is a section of Genbank version 122; and the non-redundant section of Genbank ver 123. The sequences correspond to UniGene accession numbers from the Unigene file of Mar. 30, 2001. The clone sequences are not in the sequence listing.

[0112] Table 3B: Table 3B describes Identified Genomic Regions that code for novel mRNAs. The table contains 591 identified genomic regions that are highly similar to the cDNA clones. Those regions that are within ˜100 to 200 Kb of each other on the same contig are likely to represent exons of the same gene. The indicated clone is exemplary of the cDNA clones that match the indicated genomic region. The “number clones” column indicates how many clones were isolated from the libraries that are similar to the indicated region of the chromosome. The probability number is the likelihood that region of similarity would occur by chance on a random sequence. The Accession numbers are from the Mar. 15, 2001 build of the human genome. The file date for the downloaded data was Apr. 17, 2001. These sequences may prove useful for the prediction of clinical outcomes.

[0113] Table 3C: Table 3C describes 48 clones whose sequences align to two or more non-contiguous sequences on the same assembled human contig of genomic sequence. The Accession numbers are from the Mar. 15, 2001 build of the human genome. The file date for the downloaded data was Apr. 17, 2001. The alignments of the clone and the contig are indicated in the table. The start and stop offset of each matching region is indicated in the table. The sequence of the clones themselves is included in the sequence listing. The alignments of these clones strongly suggest that they are novel nucleotide sequences. Furthermore, no EST or mRNA aligning to the clone was found in the database. These sequences may prove useful for the prediction of clinical outcomes.

[0114] Table 4: Table 4 describes patient groups and diagnostic gene sets

[0115] Table 5: Table 5 describes the nucleotide sequence databases used in the sequence analysis described herein.

[0116] Table 6: Table 6 describes the algorithms and software packages used for exon and polypeptide prediction used in the sequence analysis described herein.

[0117] Table 7: Table 7 describes the databases and algorithms used for the protein sequence analysis described herein.

[0118] Table 8: Table 8 provides a listing of all oligonucleotides designed for the arrays and their associated genes. In this table, the sequence ID is given which corresponds to the sequence listing. The origin of the sequence for inclusion on the array is noted as coming from one of the cDNA libraries described in example 1, mining from databases as described in examples 2 and 20 or identification from the published literature. The unigene number, genebank accession and GI number are also given for each sequence when known. These data were obtained from the Unigene unique database, build 137. The name of the gene associated with the accession number is noted. The strand is noted as −1 or 1, meaning that the probe was designed from the complement of the sequence (−1) or directly from the sequence (1). Finally, the nucleotide sequence of each probe is also given.

[0119] Table 9: Database mining. The Library Browser at the NCBI UniGene web site was used to identify genes that are specifically expressed in leukocyte cell populations. The table lists the library name and type, the number of sequences in each library and the number used for the array.

[0120] Table 10: Viral gene for arrays. Viral genomes were used to design oligonucleotides for the microarrays. The accession numbers for the viral genomes used are given, along with the gene name and location of the region used for oligonucleotide design.

[0121] Table 11A. CMV gene expression markers. This table lists the oligonucleotides and associated genes identified as having value for the diagnosis and monitoring of CMV infection. The first column gives the SEQ ID that corresponds to the oligonuclotide in the sequence listing. The origin of the sequence for inclusion on the array is noted as coming from one of the cDNA libraries described in example 1, mining from databases as described in examples 2 and 20 or identification from the published literature. The unigene number, genebank accession and GI number are also given for each sequence when known. The SEQ ID for the sequence listing for the full-length genes corresponding to the accession numbers in the table are also given. These data were obtained from the Unigene unique database, build 149, and the Genbank Version 129. The full length sequences are presented in Table 14A. The name of the gene associated with the accession number is noted. The strand is noted as −1 or 1, meaning that the probe was designed from the complement of the sequence (−1) or directly from the sequence (1). Next, the nucleotide sequence of each probe is also given. For each gene, the false detection rate (FDR) from the significance analysis described in example 17 is given if applicable. WBC is the white blood cell count. WPT is the number of weeks past transplant.

[0122] Table 11B. Primers for PCR. For each of the CMV gene expression markers identified in Table 11A, 2 sets of PCR primer pairs are shown that were derived by the methods described in example 25. The first column gives the SEQ ID of the oligonucleotide probe used on the microarrays. The melting temperature (Tm) for each primer is shown, as is the corresponding SEQ ID for the primer in the sequence listing.

[0123] Table 12A. Cardiac rejection gene expression markers. This table lists the oligonucleotides and associated genes identified as having value for the diagnosis and monitoring of cardiac rejection. The first column gives the SEQ ID that corresponds to the oligonuclotide in the sequence listing. The origin of the sequence for inclusion on the array is noted as coming from one of the cDNA libraries described in example 1, mining from databases as described in examples 2 and 20 or identification from the published literature. The unigene number, genebank accession and GI number are also given for each sequence when known. The full length sequence for the accession number given is in the sequence listing. The SEQ ID for the sequence listing for the full-length genes corresponding to the accession numbers in the table are also given. These data were obtained from the Unigene unique database, build 149, and the Genbank Version 129. The full length sequences are presented in Table 14B. The name of the gene associated with the accession number is noted. The strand is noted as −1 or 1, meaning that the probe was designed from the complement of the sequence (−1) or directly from the sequence (1). Next, the nucleotide sequence of each probe is given. The remaining columns give data from the significance analysis and classification analysis that identified each gene as a rejection marker. Data is reported from 2 types of data: static data or referenced data, as described in the example. For each gene, the false detection rate (FDR) from the significance analysis described in example 24 is given if applicable. The FDR is given as a %. The final columns report the criteria for selection of each gene as a rejection marker from either the static or referenced data sets:

[0124] A1: Low FDR (<20) in two independent data sets with SAM

[0125] A2: Occurred greater than 50% of the time in the top 20 genes with low FDR (<20) by SAM AND occurred at all in Isolator or CART analyses

[0126] B1: Occurred greater than 50% of the time in the top 5 terms with Isolator

[0127] B2: Identified among the top 10 based on Isolator term occurrences

[0128] C1: A primary splitter in a CART analysis of ˜6000 genes

[0129] C2: Identified among the top 10 based on CART splitter occurrences

[0130] D: Part of a logistic regression model

[0131] E: Part of a K-nearest neighbor model

[0132] Table 12B. Primers for PCR. For each of the rejection gene expression markers identified in Table 12A, 2 sets of PCR primer pairs are shown that were derived by the methods described in example 25. The first column gives the SEQ ID of the oligonucleotide probe used on the microarrays. The melting temperature (Tm) for each primer is shown, as is the corresponding SEQ ID for each primer in the sequence listing.

[0133] Table 12C. Surrogates for rejection gene expression markers. For some of the rejection marker genes identified in Table 12A, genes are identified by the SEQ ID number as surrogates. The surrogates are identified as such by the CART algorithm or by hierarchical clustering (see text). Surrogates identified by hierarchical clustering are static expression values and those identified by CART are either static or referenced (see example 24).

[0134] Table 13. Dependent variables for discovery of gene expression markers of cardiac allograft rejection. A stable Grade 0 is a Grade 0 biopsy in a patient who does not experience rejection with the subsequent biopsy. HG or highest grade means that the higher of the biopsy grades from the centralized and local pathologists was used for a definition of the dependent variable.

[0135] Table 14A shows the full length CMV gene sequences referred to in Table 11A.

[0136] Table 14B shows the full length gene sequences for genes related to transplant rejection referred to in Table 12A.

[0137] Table 15: Allograft rejection gene expression markers. Identification of and data for 133 transplant rejection marker genes are given in the tables.

• [0138] A: Genes. For each of the 133 genes the following information is given: The Array Probe SEQ ID is given for the 50mer oligonucleotide associated with each gene. The Genbank accession number and the UniGene number (Build 156) for each gene is given as is the SEQ ID for the full-length gene (Full Length SEQ ID). When a protein is known or predicted it is identified and in the sequence listing (RefSeq Peptide Accession #, Protein SEQ ID). Finally, some genes are represented by more than one 50mer oligonucleotide. These alternative oligonuclotides are identified (Alternate Array Probe SEQ IDs).
• [0139] B: Supporting Data. Each of the 133 genes is identified by an Array Probe SEQ ID and Gene name. Data associated with each gene are noted with an “x” for criteria that the gene has met. Significant in SAM means that the gene was found to be a marker of acute rejection using leukocyte array expression profiling and the algorithm Significance analysis of microarrays (SAM). Models means the gene was found to be useful for classification modeling of acute rejection. The Array Score is a global measure of usefulness and significance of the gene in analysis of rejection using the leukocyte arrays. KTx indicates that the gene was identified by expression profiling of renal allograft tissue. DB mining, pathways identifies genes that were selected based on known involvement in a key pathway of the immune response or rejection. When a gene was identified to be co-expressed with an established rejection marker, the SEQ ID for the established gene is given in the column Cluster Name.
• [0140] C. PCR Primers. For each of the 133 genes, one or two sets of real time PCR reagents are listed along with the SEQ ID for the sequence in the listing. Forward and reverse PCR primers are given, along with a probe sequence for the amplified portion of the gene.

[0141] Table 16: Real-time PCR assay reporter and quencher dyes. Various combinations of reporter and quencher dyes are useful for real-time PCR assays. Reporter and quencher dyes work optimally in specific combinations defined by their spectra. For each reporter, appropriate choices for quencher dyes are given.

[0142] Table 17: Rejection marker PCR assay results

[0143] Results of real-time PCR assays are listed for the comparison of rejection samples to no rejection samples. The fold change is given for expression of each gene in rejection/no rejection samples. The p-value for the t-test comparing the rejection and no rejection classes is given.

[0144] Table 18: Summary results of array rejection significance analysis. Summary results are given for correlation analysis of leukocyte gene expression to acute rejection using significance analysis for microarrays (SAM). Five analyses are described. The ISHLT grades used to define the rejection and no rejection classes are given. In each case the highest grade from three pathology reading was taken for analysis. All samples are used for two analyses. The other analyses reduce redundancy of patients used in the analysis by using only one sample per patient (“Non-redundant”) or using only one sample per patient within a given class (“Non-redundant within class”). The number of samples used in the analysis is given and the lowest false detection rate (FDR) achieved is noted.

[0145] Table 19: Renal tissue rejection array significance analysis. Genes are listed that were identified as upregulated using microarrays on renal tissue with acute rejection versus controls. Significance analysis for microarrays (SAM) was used to determine the false detection rate for each gene (FDR). Genes with known expression in leukocytes are noted in the table.

[0146] Table 20: Rejection marker sequence analysis. For 81 of the allograft rejection markers listed in Table 15, an analysis of the gene sequence was done. The genes and proteins are identified by accession numbers. The cellular localization of each gene is described as either secreted, nuclear, mitochondrial, cytoplasmic or cellular membrane. The function of the gene is also described.

Definitions

[0147] Unless defined otherwise, all scientific and technical terms are understood to have the same meaning as commonly used in the art to which they pertain. For the purpose of the present invention, the following terms are defined below.

[0148] In the context of the invention, the term “gene expression system” refers to any system, device or means to detect gene expression and includes diagnostic agents, candidate libraries, oligonucleotide sets or probe sets.

[0149] The term “monitoring” is used herein to describe the use of gene sets to provide useful information about an individual or an individual's health or disease status. “Monitoring” can include, determination of prognosis, risk-stratification, selection of drug therapy, assessment of ongoing drug therapy, prediction of outcomes, determining response to therapy, diagnosis of a disease or disease complication, following progression of a disease or providing any information relating to a patients health status.

[0150] The term “diagnostic oligonucleotide set” generally refers to a set of two or more oligonucleotides that, when evaluated for differential expression of their products, collectively yields predictive data. Such predictive data typically relates to diagnosis, prognosis, monitoring of therapeutic outcomes, and the like. In general, the components of a diagnostic oligonucleotide set are distinguished from nucleotide sequences that are evaluated by analysis of the DNA to directly determine the genotype of an individual as it correlates with a specified trait or phenotype, such as a disease, in that it is the pattern of expression of the components of the diagnostic nucleotide set, rather than mutation or polymorphism of the DNA sequence that provides predictive value. It will be understood that a particular component (or member) of a diagnostic nucleotide set can, in some cases, also present one or more mutations, or polymorphisms that are amenable to direct genotyping by any of a variety of well known analysis methods, e.g., Southern blotting, RFLP, AFLP, SSCP, SNP, and the like.

[0151] A “disease specific target oligonucleotide sequence” is a gene or other oligonucleotide that encodes a polypeptide, most typically a protein, or a subunit of a multi-subunit protein, that is a therapeutic target for a disease, or group of diseases.

[0152] A “candidate library” or a “candidate oligonucleotide library” refers to a collection of oligonucleotide sequences (or gene sequences) that by one or more criteria have an increased probability of being associated with a particular disease or group of diseases. The criteria can be, for example, a differential expression pattern in a disease state or in activated or resting leukocytes in vitro as reported in the scientific or technical literature, tissue specific expression as reported in a sequence database, differential expression in a tissue or cell type of interest, or the like. Typically, a candidate library has at least 2 members or components; more typically, the library has in excess of about 10, or about 100, or about 1000, or even more, members or components.

[0153] The term “disease criterion” is used herein to designate an indicator of a disease, such as a diagnostic factor, a prognostic factor, a factor indicated by a medical or family history, a genetic factor, or a symptom, as well as an overt or confirmed diagnosis of a disease associated with several indicators such as those selected from the above list. A disease criterian includes data describing a patient's health status, including retrospective or prospective health data, e.g. in the form of the patient's medical history, laboratory test results, diagnostic test result, clinical events, medications, lists, response(s) to treatment and risk factors, etc.

[0154] The terms “molecular signature” or “expression profile” refers to the collection of expression values for a plurality (e.g., at least 2, but frequently about 10, about 100, about 1000, or more) of members of a candidate library. In many cases, the molecular signature represents the expression pattern for all of the nucleotide sequences in a library or array of candidate or diagnostic nucleotide sequences or genes. Alternatively, the molecular signature represents the expression pattern for one or more subsets of the candidate library. The term “oligonucleotide” refers to two or more nucleotides. Nucleotides may be DNA or RNA, naturally occurring or synthetic.

[0155] The term “healthy individual,” as used herein, is relative to a specified disease or disease criterion. That is, the individual does not exhibit the specified disease criterion or is not diagnosed with the specified disease. It will be understood, that the individual in question, can, of course, exhibit symptoms, or possess various indicator factors for another disease.

[0156] Similarly, an “individual diagnosed with a disease” refers to an individual diagnosed with a specified disease (or disease criterion). Such an individual may, or may not, also exhibit a disease criterion associated with, or be diagnosed with another (related or unrelated) disease.

[0157] An “array” is a spatially or logically organized collection, e.g., of oligonucleotide sequences or nucleotide sequence products such as RNA or proteins encoded by an oligonucleotide sequence. In some embodiments, an array includes antibodies or other binding reagents specific for products of a candidate library.

[0158] When referring to a pattern of expression, a “qualitative” difference in gene expression refers to a difference that is not assigned a relative value. That is, such a difference is designated by an “all or nothing” valuation. Such an all or nothing variation can be, for example, expression above or below a threshold of detection (an on/off pattern of expression). Alternatively, a qualitative difference can refer to expression of different types of expression products, e.g., different alleles (e.g., a mutant or polymorphic allele), variants (including sequence variants as well as post-translationally modified variants), etc.

[0159] In contrast, a “quantitative” difference, when referring to a pattern of gene expression, refers to a difference in expression that can be assigned a value on a graduated scale, (e.g., a 0–5 or 1–10 scale, a +–+++ scale, a grade 1–grade 5 scale, or the like; it will be understood that the numbers selected for illustration are entirely arbitrary and in no-way are meant to be interpreted to limit the invention).

Gene Expression Systems of the Invention

[0160] The invention is directed to a gene expression system having one or more DNA molecules wherein the one or more DNA molecules has a nucleotide sequence which detects expression of a gene corresponding to the oligonucleotides depicted in the Sequence Listing. In one format, the oligonucleotide detects expression of a gene that is differentially expressed in leukocytes. The gene expression system may be a candidate library, a diagnostic agent, a diagnostic oligonucleotide set or a diagnostic probe set. The DNA molecules may be genomic DNA, protein nucleic acid (PNA), cDNA or synthetic oligonucleotides. Following the procedures taught herein, one can identity sequences of interest for analyzing gene expression in leukocytes. Such sequences may be predictive of a disease state.

Diagnostic Oligonucleotides of the Invention

[0161] The invention relates to diagnostic nucleotide set(s) comprising members of the leukocyte candidate library listed in Table 2, Table 3 and in the Sequence Listing, for which a correlation exists between the health status of an individual, and the individual's expression of RNA or protein products corresponding to the nucleotide sequence. In some instances, only one oligonucleotide is necessary for such detection. Members of a diagnostic oligonucleotide set may be identified by any means capable of detecting expression of RNA or protein products, including but not limited to differential expression screening, PCR, RT-PCR, SAGE analysis, high-throughput sequencing, microarrays, liquid or other arrays, protein-based methods (e.g., western blotting, proteomics, and other methods described herein), and data mining methods, as further described herein.

[0162] In one embodiment, a diagnostic oligonucleotide set comprises at least two oligonucleotide sequences listed in Table 2 or Table 3 or the Sequence Listing which are differentially expressed in leukocytes in an individual with at least one disease criterion for at least one leukocyte-implicated disease relative to the expression in individual without the at least one disease criterion, wherein expression of the two or more nucleotide sequences is correlated with at least one disease criterion, as described below. In another embodiment, a diagnostic nucleotide set comprises at least one oligonucleotide having an oligonucleotide sequence listed in Table 2 or 3 or the Sequence Listing which is differentially expressed, and further wherein the differential expression/correlation has not previously been described. In some embodiments, the diagnostic nucleotide set is immobilized on an array.

[0163] In another embodiment, diagnostic nucleotides (or nucleotide sets) are related to the members of the leukocyte candidate library listed in Table 2, Table 3 and in the Sequence Listing, for which a correlation exists between the health status (or disease criterion) of an individual. The diagnostic nucleotides are partially or totally contained in (or derived from) full-length gene sequences (or predicted full-length gene sequences) for the members of the candidate library listed in Table 2, 3, and Tables 8, 11–12, 14, 15, and 19. This includes sequences from accession numbers and unigene numbers from Table 8. Table 8 shows the accession and unigene number (when known) for each oligonucleotide used on the 8134 gene leukocyte array described in examples 11–13. In some cases, oligonucleotide sequences are designed from EST or Chromosomal sequences from a public database. In these cases the full-length gene sequences may not be known. Full-length sequences in these cases can be predicted using gene prediction algorithms (examples 4–6). Alternatively the full-length can be determined by cloning and sequencing the full-length gene or genes that contain the sequence of interest using standard molecular biology approaches described here. The same is true for oligonucleotides designed from our sequencing of cDNA libraries (see examples 1–4) where the cDNA does not match any sequence in the public databases.

[0164] The diagnostic nucleotides may also be derived from other genes that are coexpressed with the correlated sequence or full-length gene. Genes may share expression patterns because they are regulated in the same molecular pathway. Because of the similarity of expression behavior genes are identified as surrogates in that they can substitute for a diagnostic gene in a diagnostic gene set. Example 10 demonstrates the discovery of surrogates from the data and the sequence listing identify and give the sequence for surrogates for lupus diagnostic genes.

[0165] As used herein the term “gene cluster” or “cluster” refers to a group of genes related by expression pattern. In other words, a cluster of genes is a group of genes with similar regulation across different conditions, such as graft non-rejection verus graft rejection. The expression profile for each gene in a cluster should be correlated with the expression profile of at least one other gene in that cluster. Correlation may be evaluated using a variety of statistical methods. As used herein the term “surrogate” refers to a gene with an expression profile such that it can substitute for a diagnostic gene in a diagnostic assay. Such genes are often members of the same gene cluster as the diagnostic gene. For each member of a diagnostic gene set, a set of potential surrogates can be identified through identification of genes with similar expression patterns as described below. Many statistical analyses produce a correlation coefficient to describe the relatedness between two gene expression patterns. Patterns may be considered correlated if the correlation coefficient is greater than or equal to 0.8. In preferred embodiments, the correlation coefficient should be greater than 0.85, 0.9 or 0.95. Other statistical methods produce a measure of mutual information to describe the relatedness between two gene expression patterns. Patterns may be considered correlated if the normalized mutual information value is greater than or equal to 0.7. In preferred embodiments, the normalized mutual information value should be greater than 0.8, 0.9 or 0.95. Patterns may also be considered similar if they cluster closely upon hierarchical clustering of gene expression data (Eisen et al. 1998). Similar patterns may be those genes that are among the 1, 2, 5, 10, 20, 50 or 100 nearest neighbors in a hierarchical clustering or have a similarity score (Eisen et al. 1998) of >0.5, 0.7, 0.8, 0.9, 0.95 or 0.99. Similar patterns may also be identified as those genes found to be surrogates in a classification tree by CART (Breiman et al. 1994). Often, but not always, members of a gene cluster have similar biological functions in addition to similar gene expression patterns.

[0166] Correlated genes, clusters and surrogates are identified for the diagnostic genes of the invention. These surrogates may be used as diagnostic genes in an assay instead of, or in addition to, the diagnostic genes for which they are surrogates.

[0167] The invention also provides diagnostic probe sets. It is understood that a probe includes any reagent capable of specifically identifying a nucleotide sequence of the diagnostic nucleotide set, including but not limited to amplified DNA, amplified RNA, cDNA, synthetic oligonucleotide, partial or full-length nucleic acid sequences. In addition, the probe may identify the protein product of a diagnostic nucleotide sequence, including, for example, antibodies and other affinity reagents.

[0168] It is also understood that each probe can correspond to one gene, or multiple probes can correspond to one gene, or both, or one probe can correspond to more than one gene. Homologs and variants of the disclosed nucleic acid molecules may be used in the present invention. Homologs and variants of these nucleic acid molecules will possess a relatively high degree of sequence identity when aligned using standard methods. The sequences encompassed by the invention have at least 40–50, 50–60, 70–80, 80–85, 85–90, 90–95 or 95–100% sequence identity to the sequences disclosed herein.

[0169] It is understood that for expression profiling, variations in the disclosed sequences will still permit detection of gene expression. The degree of sequence identity required to detect gene expression varies depending on the length of the oligomer. For a 60 mer, 6–8 random mutations or 6–8 random deletions in a 60 mer do not affect gene expression detection. Hughes, T R, et al. “Expression profiling using microarrays fabricated by an ink-jet oligonucleotide synthesizer. Nature Biotechnology, 19:343–347(2001). As the length of the DNA sequence is increased, the number of mutations or deletions permitted while still allowing gene expression detection is increased.

[0170] As will be appreciated by those skilled in the art, the sequences of the present invention may contain sequencing errors. That is, there may be incorrect nucleotides, frameshifts, unknown nucleotides, or other types of sequencing errors in any of the sequences; however, the correct sequences will fall within the homology and stringency definitions herein.

[0171] The minimum length of an oligonucleotide probe necessary for specific hybridization in the human genome can be estimated using two approaches. The first method uses a statistical argument that the probe will be unique in the human genome by chance.

[0172] Briefly, the number of independent perfect matches (Po) expected for an oligonucleotide of length L in a genome of complexity C can be calculated from the equation (Laird C D, Chromosoma 32:378 (1971):
Po=(1/4)^L*2C

[0173] In the case of mammalian genomes, 2C=˜3.6×10⁹, and an oligonucleotide of 14–15 nucleotides is expected to be represented only once in the genome. However, the distribution of nucleotides in the coding sequence of mammalian genomes is nonrandom (Lathe, R. J. Mol. Biol. 183:1 (1985) and longer oligonucleotides may be preferred in order to in increase the specificity of hybridization. In practical terms, this works out to probes that are 19–40 nucleotides long (Sambrook J et al., infra). The second method for estimating the length of a specific probe is to use a probe long enough to hybridize under the chosen conditions and use a computer to search for that sequence or close matches to the sequence in the human genome and choose a unique match. Probe sequences are chosen based on the desired hybridization properties as described in Chapter 11 of Sambrook et al, infra. The PRIMER3 program is useful for designing these probes (S. Rozen and H. Skaletsky 1996,1997; Primer3 code available at the web site located at genome.wi.mit.edu/genome_software/other/primer3.html). The sequences of these probes are then compared pair wise against a database of the human genome sequences using a program such as BLAST or MEGABLAST (Madden, T. L et al.(1996) Meth. Enzymol. 266:131–141). Since most of the human genome is now contained in the database, the number of matches will be determined. Probe sequences are chosen that are unique to the desired target sequence.

[0174] In some embodiments, a diagnostic probe set is immobilized on an array. The array is optionally comprises one or more of: a chip array, a plate array, a bead array, a pin array, a membrane array, a solid surface array, a liquid array, an oligonucleotide array, a polynucleotide array or a cDNA array, a microtiter plate, a pin array, a bead array, a membrane or a chip.

[0175] In some embodiments, the leukocyte-implicated disease is selected from the diseases listed in Table 1. In other embodiments, the disease is atherosclerosis or cardiac allograft rejection. In other embodiments, the disease is congestive heart failure, angina, myocardial infarction, systemic lupus erythematosis (SLE) and rheumatoid arthritis.

[0176] In some embodiments, diagnostic nucleotides of the invention are used as a diagnostic gene set in combination with genes that are know to be associated with a disease state (“known markers”). The use of the diagnostic nucleotides in combination with the known markers can provide information that is not obtainable through the known markers alone. The known markers include those identified by the prior art listing provided.

General Molecular Biology References

[0177] In the context of the invention, nucleic acids and/or proteins are manipulated according to well known molecular biology techniques. Detailed protocols for numerous such procedures are described in, e.g., in Ausubel et al. Current Protocols in Molecular Biology (supplemented through 2000) John Wiley & Sons, New York (“Ausubel”); Sambrook et al. Molecular Cloning—A Laboratory Manual (2nd Ed.), Vol. 1–3, Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y., 1989 (“Sambrook”), and Berger and Kimmel Guide to Molecular Cloning Techniques, Methods in Enzymology volume 152 Academic Press, Inc., San Diego, Calif. (“Berger”).

[0178] In addition to the above references, protocols for in vitro amplification techniques, such as the polymerase chain reaction (PCR), the ligase chain reaction (LCR), Q-replicase amplification, and other RNA polymerase mediated techniques (e.g., NASBA), useful e.g., for amplifying cDNA probes of the invention, are found in Mullis et al. (1987) U.S. Pat. No. 4,683,202; PCR Protocols A Guide to Methods and Applications (Innis et al. eds) Academic Press Inc. San Diego, Calif. (1990) (“Innis”); Arnheim and Levinson (1990) C&EN 36; The Journal Of NIH Research (1991) 3:81; Kwoh et al. (1989) Proc Natl Acad Sci USA 86, 1173; Guatelli et al. (1990) Proc Natl Acad Sci USA 87:1874; Lomell et al. (1989) J Clin Chem 35:1826; Landegren et al. (1988) Science 241:1077; Van Brunt (1990) Biotechnology 8:291; Wu and Wallace (1989) Gene 4: 560; Barringer et al. (1990) Gene 89:117, and Sooknanan and Malek (1995) Biotechnology 13:563.

[0179] Additional methods, useful for cloning nucleic acids in the context of the present invention, include Wallace et al. U.S. Pat. No. 5,426,039. Improved methods of amplifying large nucleic acids by PCR are summarized in Cheng et al. (1994) Nature 369:684 and the references therein.

[0180] Certain polynucleotides of the invention, e.g., oligonucleotides can be synthesized utilizing various solid-phase strategies involving mononucleotide- and/or trinucleotide-based phosphoramidite coupling chemistry. For example, nucleic acid sequences can be synthesized by the sequential addition of activated monomers and/or trimers to an elongating polynucleotide chain. See e.g., Caruthers, M. H. et al. (1992) Meth Enzymol 211:3.

[0181] In lieu of synthesizing the desired sequences, essentially any nucleic acid can be custom ordered from any of a variety of commercial sources, such as The Midland Certified Reagent Company ([email protected]), The Great American Gene Company ExpressGen, Inc., Operon Technologies, Inc. and many others.

[0182] Similarly, commercial sources for nucleic acid and protein microarrays are available, and include, e.g., Agilent Technologies, Palo Alto, Calif. Affymetrix, Santa Clara, Calif.; and others.

Identification of Diagnostic Nucleotide Sets

Candidate Library

[0183] Libraries of candidates that are differentially expressed in leukocytes are substrates for the identification and evaluation of diagnostic oligonucleotide sets and disease specific target nucleotide sequences.

[0184] The term leukocyte is used generically to refer to any nucleated blood cell that is not a nucleated erythrocyte. More specifically, leukocytes can be subdivided into two broad classes. The first class includes granulocytes, including, most prevalently, neutrophils, as well as eosinophils and basophils at low frequency. The second class, the non-granular or mononuclear leukocytes, includes monocytes and lymphocytes (e.g., T cells and B cells). There is an extensive literature in the art implicating leukocytes, e.g., neutrophils, monocytes and lymphocytes in a wide variety of disease processes, including inflammatory and rheumatic diseases, neurodegenerative diseases (such as Alzheimer's dementia), cardiovascular disease, endocrine diseases, transplant rejection, malignancy and infectious diseases, and other diseases listed in Table 1. Mononuclear cells are involved in the chronic immune response, while granulocytes, which make up approximately 60% of the leukocytes, have a non-specific and stereotyped response to acute inflammatory stimuli and often have a life span of only 24 hours.

[0185] In addition to their widespread involvement and/or implication in numerous disease related processes, leukocytes are particularly attractive substrates for clinical and experimental evaluation for a variety of reasons. Most importantly, they are readily accessible at low cost from essentially every potential subject. Collection is minimally invasive and associated with little pain, disability or recovery time. Collection can be performed by minimally trained personnel (e.g., phlebotomists, medical technicians, etc.) in a variety of clinical and non-clinical settings without significant technological expenditure. Additionally, leukocytes are renewable, and thus available at multiple time points for a single subject.

Assembly of Candidate Libraries

[0186] At least two conceptually distinct approaches to the assembly of candidate libraries exist. Either, or both, or other, approaches can be favorably employed. The method of assembling, or identifying, candidate libraries is secondary to the criteria utilized for selecting appropriate library members. Most importantly, library members are assembled based on differential expression of RNA or protein products in leukocyte populations.

[0187] More specifically, candidate nucleotide sequences are induced or suppressed, or expressed at increased or decreased levels in leukocytes from a subject with one or more disease or disease state (a disease criterion) relative to leukocytes from a subject lacking the specified disease criterion. Alternatively, or in addition, library members can be assembled from among nucleotide sequences that are differentially expressed in activated or resting leukocytes relative to other cell types.

[0188] Firstly, publication and sequence databases can be “mined” using a variety of search strategies, including, e.g., a variety of genomics and proteomics approaches. For example, currently available scientific and medical publication databases such as Medline, Current Contents, OMIM (online Mendelian inheritance in man) various Biological and Chemical Abstracts, Journal indexes, and the like can be searched using term or key-word searches, or by author, title, or other relevant search parameters. Many such databases are publicly available, and one of skill is well versed in strategies and procedures for identifying publications and their contents, e.g., genes, other nucleotide sequences, descriptions, indications, expression pattern, etc. Numerous databases are available through the internet for free or by subscription, see, e.g., NCBI PubMed website, Science magazine website, Infotrieve website, and ISI website. 3 Additional or alternative publication or citation databases are also available that provide identical or similar types of information, any of which are favorable employed in the context of the invention. These databases can be searched for publications describing differential gene expression in leukocytes between patient with and without diseases or conditions listed in Table 1. We identified the nucleotide sequences listed in Table 2 and some of the sequences listed in Table 8 (Example 20), using data mining methods.

[0189] Alternatively, a variety of publicly available and proprietary sequence databases (including GenBank, dbEST, UniGene, and TIGR and SAGE databases) including sequences corresponding to expressed nucleotide sequences, such as expressed sequence tags (ESTs) are available. For example, Genbank™ ncbi.nlm.nih.gov/Genbank/) among others can be readily accessed and searched via the internet. These and other sequence and clone database resources are currently available; however, any number of additional or alternative databases comprising nucleotide sequence sequences, EST sequences, clone repositories, PCR primer sequences, and the like corresponding to individual nucleotide sequence sequences are also suitable for the purposes of the invention.

[0190] Sequences from nucleotide sequences can be identified that are only found in libraries derived from leukocytes or sub-populations of leukocytes, for example see Table 2. Alternatively, the representation, or relative frequency, of a nucleotide sequence may be determined in a leukocyte-derived nucleic acid library and compared to the representation of the sequence in non-leukocyte derived libraries. The representation of a nucleotide sequence correlates with the relative expression level of the nucleotide sequence in leukocytes and non-leukocytes. An oligonucleotide sequence which has increased or decreased representation in a leukocyte-derived nucleic acid library relative to a non-leukocyte-derived libraries is a candidate for a leukocyte-specific gene.

[0191] Nucleotide sequences identified as having specificity to activated or resting leukocytes or to leukocytes from patients or patient samples with a variety of disease types can be isolated for use in a candidate library for leukocyte expression profiling through a variety of mechanisms. These include, but are not limited to, the amplification of the nucleotide sequence from RNA or DNA using nucleotide sequence specific primers for PCR or RT-PCR, isolation of the nucleotide sequence using conventional cloning methods, the purchase of an IMAGE consortium cDNA clone (EST) with complimentary sequence or from the same expressed nucleotide sequence, design of oligonucleotides, preparation of synthetic nucleic acid sequence, or any other nucleic-acid based method. In addition, the protein product of the nucleotide sequence can be isolated or prepared, and represented in a candidate library, using standard methods in the art, as described further below.

[0192] While the above discussion related primarily to “genomics” approaches, it is appreciated that numerous, analogous “proteomics” approaches are suitable to the present invention. For example, a differentially expressed protein product can, for example, be detected using western analysis, two-dimensional gel analysis, chromatographic separation, mass spectrometric detection, protein-fusion reporter constructs, calorimetric assays, binding to a protein array, or by characterization of polysomal mRNA. The protein is further characterized and the nucleotide sequence encoding the protein is identified using standard techniques, e.g. by screening a cDNA library using a probe based on protein sequence information.

[0193] The second approach involves the construction of a differential expression library by any of a variety of means. Any one or more of differential screening, differential display or subtractive hybridization procedures, or other techniques that preferentially identify, isolate or amplify differentially expressed nucleotide sequences can be employed to produce a library of differentially expressed candidate nucleotide sequences, a subset of such a library, a partial library, or the like. Such methods are well known in the art. For example, peripheral blood leukocytes, (i.e., a mixed population including lymphocytes, monocytes and neutrophils), from multiple donor samples are pooled to prevent bias due to a single-donor's unique genotype. The pooled leukocytes are cultured in standard medium and stimulated with individual cytokines or growth factors e.g., with IL-2, IL-1, MCP1, TNFα, and/or IL8 according to well known procedures (see, e.g., Tough et al. (1999); Winston et al. (1999); Hansson et al. (1989)). Typically, leukocytes are recovered from Buffy coat preparations produced by centrifugation of whole blood. Alternatively, mononuclear cells (monocytes and lymphocytes) can be obtained by density gradient centrifugation of whole blood, or specific cell types (such as a T lymphocyte) can be isolated using affinity reagents to cell specific surface markers. Leukocytes may also be stimulated by incubation with ionomycin, and phorbol myristate acetate (PMA). This stimulation protocol is intended to non-specifically mimic “activation” of numerous pathways due to variety of disease conditions rather than to simulate any single disease condition or paradigm.

[0194] Using well known subtractive hybridization procedures (as described in, e.g., U.S. Pat. Nos. 5,958,738; 5,589,339; 5,827,658; 5,712,127; 5,643,761) each of which are hereby incorporated by reference, a library is produced that is enriched for RNA species (messages) that are differentially expressed between test and control leukocyte populations. In some embodiments, the test population of leukocytes are simply stimulated as described above to emulate non-specific activation events, while in other embodiments the test population can be selected from subjects (or patients) with a specified disease or class of diseases. Typically, the control leukocyte population lacks the defining test condition, e.g., stimulation, disease state, diagnosis, genotype, etc. Alternatively, the total RNA from control and test leukocyte populations are prepared by established techniques, treated with DNAseI, and selected for messenger RNA with an intact 3′ end (i.e., polyA(+) messenger RNA) e.g., using commercially available kits according to the manufacturer's instructions e.g. Clontech. Double stranded cDNA is synthesized utilizing reverse transcriptase. Double stranded cDNA is then cut with a first restriction enzyme (e.g., NlaIII, that cuts at the recognition site: CATG, and cuts the cDNA sequence at approximately 256 bp intervals) that cuts the cDNA molecules into conveniently sized fragments.

[0195] The cDNAs prepared from the test population of leukocytes are divided into (typically 2) “tester” pools, while cDNAs prepared from the control population of leukocytes are designated the “driver” pool. Typically, pooled populations of cells from multiple individual donors are utilized and in the case of stimulated versus unstimulated cells, the corresponding tester and driver pools for any single subtraction reaction are derived from the same donor pool.

[0196] A unique double-stranded adapter is ligated to each of the tester cDNA populations using unphosphorylated primers so that only the sense strand is covalently linked to the adapter. An initial hybridization is performed consisting of each of the tester pools of cDNA (each with its corresponding adapter) and an excess of the driver cDNA. Typically, an excess of about 10–100 fold driver relative to tester is employed, although significantly lower or higher ratios can be empirically determined to provide more favorable results. The initial hybridization results in an initial normalization of the cDNAs such that high and low abundance messages become more equally represented following hybridization due to a failure of driver/tester hybrids to amplify.

[0197] A second hybridization involves pooling un-hybridized sequences from initial hybridizations together with the addition of supplemental driver cDNA. In this step, the expressed sequences enriched in the two tester pools following the initial hybridization can hybridize. Hybrids resulting from the hybridization between members of each of the two tester pools are then recovered by amplification in a polymerase chain reaction (PCR) using primers specific for the unique adapters. Again, sequences originating in a tester pool that form hybrids with components of the driver pool are not amplified. Hybrids resulting between members of the same tester pool are eliminated by the formation of “panhandles” between their common 5′ and 3′ ends. For additional details, see, e.g., Lukyanov et al. (1997) Biochem Biophys Res Commun 230:285–8.

[0198] Typically, the tester and driver pools are designated in the alternative, such that the hybridization is performed in both directions to ensure recovery of messenger RNAs that are differentially expressed in either a positive or negative manner (i.e., that are turned on or turned off, up-regulated or down-regulated). Accordingly, it will be understood that the designation of test and control populations is to some extent arbitrary, and that a test population can just as easily be compared to leukocytes derived from a patient with the same of another disease of interest.

[0199] If so desired, the efficacy of the process can be assessed by such techniques as semi-quantitative PCR of known (i.e., control) nucleotide sequences, of varying abundance such as β-actin. The resulting PCR products representing partial cDNAs of differentially expressed nucleotide sequences are then cloned (i.e., ligated) into an appropriate vector (e.g., a commercially available TA cloning vector, such as pGEM from Promega) and, optionally, transformed into competent bacteria for selection and screening.

[0200] Either of the above approaches, or both in combination, or indeed, any procedure, which permits the assembly of a collection of nucleotide sequences that are expressed in leukocytes, is favorably employed to produce the libraries of candidates useful for the identification of diagnostic nucleotide sets and disease specific target nucleotides of the invention. Additionally, any method that permits the assembly of a collection of nucleotides that are expressed in leukocytes and preferentially associated with one or more disease or condition, whether or not the nucleotide sequences are differentially expressed, is favorably employed in the context of the invention. Typically, libraries of about 2,000–10,000 members are produced (although libraries in excess of 10,000 are not uncommon). Following additional evaluation procedures, as described below, the proportion of unique clones in the candidate library can approximate 100%.

[0201] A candidate oligonucleotide sequence may be represented in a candidate library by a full-length or partial nucleic acid sequence, deoxyribonucleic acid (DNA) sequence, cDNA sequence, RNA sequence, synthetic oligonucleotides, etc. The nucleic acid sequence can be at least 19 nucleotides in length, at least 25 nucleotides, at least 40 nucleotides, at least 100 nucleotides, or larger. Alternatively, the protein product of a candidate nucleotide sequence may be represented in a candidate library using standard methods, as further described below.

Characterization of Candidate Oligonucleotide Sequences

[0202] The sequence of individual members (e.g., clones, partial sequence listing in a database such as an EST, etc.) of the candidate oligonucleotide libraries is then determined by conventional sequencing methods well known in the art, e.g., by the dideoxy-chain termination method of Sanger et al. (1977) Proc Natl Acad Sci USA 74:5463–7; by chemical procedures, e.g., Maxam and Gilbert (1977) Proc Natl Acad Sci USA 74:560–4; or by polymerase chain reaction cycle sequencing methods, e.g., Olsen and Eckstein (1989) Nuc Acid Res 17:9613–20, DNA chip based sequencing techniques or variations, including automated variations (e.g., as described in Hunkapiller et al. (1991) Science 254:59–67; Pease et al. (1994) Proc Natl Acad Sci USA 91:5022–6), thereof. Numerous kits for performing the above procedures are commercially available and well known to those of skill in the art. Character strings corresponding to the resulting nucleotide sequences are then recorded (i.e., stored) in a database. Most commonly the character strings are recorded on a computer readable medium for processing by a computational device.

[0203] Generally, to facilitate subsequent analysis, a custom algorithm is employed to query existing databases in an ongoing fashion, to determine the identity, expression pattern and potential function of the particular members of a candidate library. The sequence is first processed, by removing low quality sequence. Next the vector sequences are identified and removed and sequence repeats are identified and masked. The remaining sequence is then used in a Blast algorithm against multiple publicly available, and/or proprietary databases, e.g., NCBI nucleotide, EST and protein databases, Unigene, and Human Genome Sequence. Sequences are also compared to all previously sequenced members of the candidate libraries to detect redundancy.

[0204] In some cases, sequences are of high quality, but do not match any sequence in the NCBI nr, human EST or Unigene databases. In this case the sequence is queried against the human genomic sequence. If a single chromosomal site is matched with a high degree of confidence, that region of genomic DNA is identified and subjected to further analysis with a gene prediction program such as GRAIL. This analysis may lead to the identification of a new gene in the genomic sequence. This sequence can then be translated to identify the protein sequence that is encoded and that sequence can be further analyzed using tools such as Pfam, Blast P, or other protein structure prediction programs, as illustrated in Table 7. Typically, the above analysis is directed towards the identification of putative coding regions, e.g., previously unidentified open reading frames, confirming the presence of known coding sequences, and determining structural motifs or sequence similarities of the predicted protein (i.e., the conceptual translation product) in relation to known sequences. In addition, it has become increasingly possible to assemble “virtual cDNAs” containing large portions of coding region, simply through the assembly of available expressed sequence tags (ESTs). In turn, these extended nucleic acid and amino acid sequences allow the rapid expansion of substrate sequences for homology searches and structural and functional motif characterization. The results of these analysis permits the categorization of sequences according to structural characteristics, e.g., as structural proteins, proteins involved in signal transduction, cell surface or secreted proteins etc.

[0205] It is understood that full-length nucleotide sequences may also be identified using conventional methods, for example, library screening, RT-PCR, chromosome walking, etc., as described in Sambrook and Ausubel, infra.

Candidate Nucleotide Library of the Invention

[0206] We identified members of a candidate nucleotide library that are differentially expressed in activated leukocytes and resting leukocytes. Accordingly, the invention provides the candidate leukocyte nucleotide library comprising the nucleotide sequences listed in Table 2, Table 3 and in the sequence listing. In another embodiment, the invention provides a candidate library comprising at least one nucleotide sequence listed in Table 2, Table 3, Tables 8, 11–12, 14, 15, and 19 and the sequence listing. In another embodiment, the invention provides a candidate library comprising at least two nucleotide sequences listed in Table 2, Table 3, Tables 8, 11–12, 14, 15, and 19 and the sequence listing. In another embodiment, the at least two nucleotide sequence are at least 19 nucleotides in length, at least 35 nucleotides, at least 40 nucleotides or at least 100 nucleotides. In some embodiments, the nucleotide sequences comprises deoxyribonucleic acid (DNA) sequence, ribonucleic acid (RNA) sequence, synthetic oligonucleotide sequence, or genomic DNA sequence. It is understood that the nucleotide sequences may each correspond to one gene, or that several nucleotide sequences may correspond to one gene, or both.

[0207] The invention also provides probes to the candidate nucleotide library. In one embodiment of the invention, the probes comprise at least two nucleotide sequences listed in Table 2, Table 3, or the sequence listing which are differentially expressed in leukocytes in an individual with a least one disease criterion for at least one leukocyte-related disease and in leukocytes in an individual without the at least one disease criterion, wherein expression of the two or more nucleotide sequences is correlated with at least one disease criterion. It is understood that a probe may detect either the RNA expression or protein product expression of the candidate nucleotide library.

[0208] Alternatively, or in addition, a probe can detect a genotype associated with a candidate nucleotide sequence, as further described below. In another embodiment, the probes for the candidate nucleotide library are immobilized on an array.

[0209] The candidate nucleotide library of the invention is useful in identifying diagnostic nucleotide sets of the invention, as described below. The candidate nucleotide sequences may be further characterized, and may be identified as a disease target nucleotide sequence and/or a novel nucleotide sequence, as described below. The candidate nucleotide sequences may also be suitable for use as imaging reagents, as described below.

Detection of Non-Leukocyte Expressed Genes

[0210] When measuring gene expression levels in a blood sample, RNAs may be measured that are not derived from leukocytes. Examples are viral genes, free RNAs that have been released from damaged non-leukocyte cell types or RNA from circulating non-leukocyte cell types. For example, in the process of acute allograft rejection, tissue damage may result in release of allograft cells or RNAs derived from allograft cells into the circulation. In the case of cardiac allografts, such transcripts may be specific to muscle (myoglobin) or to cardiac muscle (Troponin I, Toponin T, CK-MB). Presence of cardiac specific mRNAs in peripheral blood may indicate ongoing or recent cardiac cellular damage (resulting from acute rejection). Therefore, such genes may be excellent diagnostic markers for allograft rejection.

Generation of Expression Patterns

RNA, DNA or Protein Sample Procurement

[0211] Following identification or assembly of a library of differentially expressed candidate nucleotide sequences, leukocyte expression profiles corresponding to multiple members of the candidate library are obtained. Leukocyte samples from one or more subjects are obtained by standard methods. Most typically, these methods involve trans-cutaneous venous sampling of peripheral blood. While sampling of circulating leukocytes from whole blood from the peripheral vasculature is generally the simplest, least invasive, and lowest cost alternative, it will be appreciated that numerous alternative sampling procedures exist, and are favorably employed in some circumstances. No pertinent distinction exists, in fact, between leukocytes sampled from the peripheral vasculature, and those obtained, e.g., from a central line, from a central artery, or indeed from a cardiac catheter, or during a surgical procedure which accesses the central vasculature. In addition, other body fluids and tissues that are, at least in part, composed of leukocytes are also desirable leukocyte samples. For example, fluid samples obtained from the lung during bronchoscopy may be rich in leukocytes, and amenable to expression profiling in the context of the invention, e.g., for the diagnosis, prognosis, or monitoring of lung transplant rejection, inflammatory lung diseases or infectious lung disease. Fluid samples from other tissues, e.g., obtained by endoscopy of the colon, sinuses, esophagus, stomach, small bowel, pancreatic duct, biliary tree, bladder, ureter, vagina, cervix or uterus, etc., are also suitable. Samples may also be obtained other sources containing leukocytes, e.g., from urine, bile, cerebrospinal fluid, feces, gastric or intestinal secretions, semen, or solid organ or joint biopsies.

[0212] Most frequently, mixed populations of leukocytes, such as are found in whole blood are utilized in the methods of the present invention. A crude separation, e.g., of mixed leukocytes from red blood cells, and/or concentration, e.g., over a sucrose, percoll or ficoll gradient, or by other methods known in the art, can be employed to facilitate the recovery of RNA or protein expression products at sufficient concentrations, and to reduce non-specific background. In some instances, it can be desirable to purify sub-populations of leukocytes, and methods for doing so, such as density or affinity gradients, flow cytometry, fluorescence Activated Cell Sorting (FACS), immuno-magnetic separation, “panning,” and the like, are described in the available literature and below.

Obtaining DNA, RNA and Protein Samples for Expression Profiling

[0213] Expression patterns can be evaluated at the level of DNA, or RNA or protein products. For example, a variety of techniques are available for the isolation of RNA from whole blood. Any technique that allows isolation of mRNA from cells (in the presence or absence of rRNA and tRNA) can be utilized. In brief, one method that allows reliable isolation of total RNA suitable for subsequent gene expression analysis, is described as follows. Peripheral blood (either venous or arterial) is drawn from a subject, into one or more sterile, endotoxin free, tubes containing an anticoagulant (e.g., EDTA, citrate, heparin, etc.). Typically, the sample is divided into at least two portions. One portion, e.g., of 5–8 ml of whole blood is frozen and stored for future analysis, e.g., of DNA or protein. A second portion, e.g., of approximately 8 ml whole blood is processed for isolation of total RNA by any of a variety of techniques as described in, e.g, Sambook, Ausubel, below, as well as U.S. Pat. Nos. 5,728,822 and 4,843,155.

[0214] Typically, a subject sample of mononuclear leukocytes obtained from about 8 ml of whole blood, a quantity readily available from an adult human subject under most circumstances, yields 5–20 μg of total RNA. This amount is ample, e.g., for labeling and hybridization to at least two probe arrays. Labeled probes for analysis of expression patterns of nucleotides of the candidate libraries are prepared from the subject's sample of RNA using standard methods. In many cases, cDNA is synthesized from total RNA using a polyT primer and labeled, e.g., radioactive or fluorescent, nucleotides. The resulting labeled cDNA is then hybridized to probes corresponding to members of the candidate nucleotide library, and expression data is obtained for each nucleotide sequence in the library. RNA isolated from subject samples (e.g., peripheral blood leukocytes, or leukocytes obtained from other biological fluids and samples) is next used for analysis of expression patterns of nucleotides of the candidate libraries.

[0215] In some cases, however, the amount of RNA that is extracted from the leukocyte sample is limiting, and amplification of the RNA is desirable. Amplification may be accomplished by increasing the efficiency of probe labeling, or by amplifying the RNA sample prior to labeling. It is appreciated that care must be taken to select an amplification procedure that does not introduce any bias (with respect to gene expression levels) during the amplification process.

[0216] Several methods are available that increase the signal from limiting amounts of RNA, e.g. use of the Clontech (Glass Fluorescent Labeling Kit) or Stratagene (Fairplay Microarray Labeling Kit), or the Micromax kit (New England Nuclear, Inc.). Alternatively, cDNA is synthesized from RNA using a T7-polyT primer, in the absence of label, and DNA dendrimers from Genisphere (3DNA Submicro) are hybridized to the poly T sequence on the primer, or to a different “capture sequence” which is complementary to a fluorescently labeled sequence. Each 3DNA molecule has 250 fluorescent molecules and therefore can strongly label each cDNA.

[0217] Alternatively, the RNA sample is amplified prior to labeling. For example, linear amplification may be performed, as described in U.S. Pat. No. 6,132,997. A T7-polyT primer is used to generate the cDNA copy of the RNA. A second DNA strand is then made to complete the substrate for amplification. The T7 promoter incorporated into the primer is used by a T7 polymerase to produce numerous antisense copies of the original RNA. Fluorescent dye labeled nucleotides are directly incorporated into the RNA. Alternatively, amino allyl labeled nucleotides are incorporated into the RNA, and then fluorescent dyes are chemically coupled to the amino allyl groups, as described in Hughes. Other exemplary methods for amplification are described below.

[0218] It is appreciated that the RNA isolated must contain RNA derived from leukocytes, but may also contain RNA from other cell types to a variable degree. Additionally, the isolated RNA may come from subsets of leukocytes, e.g. monocytes and/or T-lymphocytes, as described above. Such consideration of cell type used for the derivation of RNA depend on the method of expression profiling used. Subsets of leukocytes can be obtained by fluorescence activated cell sorting (FACS), microfluidics cell separation systems or a variety of other methods. Cell sorting may be necessary for the discovery of diagnostic gene sets, for the implementation of gene sets as products or both. Cell sorting can be achieved with a variety of technologies (See Galbraith et al. 1999, Cantor et al. 1975, see also the technology of Guava Technologies, Hayward, Calif.).

[0219] DNA samples may be obtained for analysis of the presence of DNA mutations, single nucleotide polymorphisms (SNPs), or other polymorphisms. DNA is isolated using standard techniques, e.g. Maniatus, supra.

[0220] Expression of products of candidate nucleotides may also be assessed using proteomics. Protein(s) are detected in samples of patient serum or from leukocyte cellular protein. Serum is prepared by centrifugation of whole blood, using standard methods. Proteins present in the serum may have been produced from any of a variety of leukocytes and non-leukocyte cells, and include secreted proteins from leukocytes. Alternatively, leukocytes or a desired sub-population of leukocytes are prepared as described above. Cellular protein is prepared from leukocyte samples using methods well known in the art, e.g., Trizol (Invitrogen Life Technologies, cat # 15596108; Chomczynski, P. and Sacchi, N. (1987) Anal. Biochem. 162, 156; Simms, D., Cizdziel, P. E., and Chomczynski, P. (1993) Focus® 15, 99; Chomczynski, P., Bowers-Finn, R., and Sabatini, L. (1987) J. of NIH Res. 6, 83; Chomczynski, P. (1993) Bio/Techniques 15, 532; Bracete, A. M., Fox, D. K., and Simms, D. (1998) Focus 20, 82; Sewall, A. and McRae, S. (1998) Focus 20, 36; Anal Biochem 1984 April;138(1):141–3, A method for the quantitative recovery of protein in dilute solution in the presence of detergents and lipids; Wessel D, Flugge U I. (1984) Anal Biochem. 1984 April;138(1):141–143.

[0221] The assay itself may be a cell sorting assay in which cells are sorted and/or counted based on cell surface expression of a protein marker. (See Cantor et al. 1975, Galbraith et al. 1999)

Obtaining Expression Patterns

[0222] Expression patterns, or profiles, of a plurality of nucleotides corresponding to members of the candidate library are then evaluated in one or more samples of leukocytes.

[0223] Typically, the leukocytes are derived from patient peripheral blood samples, although, as indicated above, many other sample sources are also suitable. These expression patterns constitute a set of relative or absolute expression values for a some number of RNAs or protein products corresponding to the plurality of nucleotide sequences evaluated, which is referred to herein as the subject's “expression profile” for those nucleotide sequences.

[0224] While expression patterns for as few as one independent member of the candidate library can be obtained, it is generally preferable to obtain expression patterns corresponding to a larger number of nucleotide sequences, e.g., about 2, about 5, about 10, about 20, about 50, about 100, about 200, about 500, or about 1000, or more. The expression pattern for each differentially expressed component member of the library provides a finite specificity and sensitivity with respect to predictive value, e.g., for diagnosis, prognosis, monitoring, and the like.

Clinical Studies, Data and Patient Groups

[0225] For the purpose of discussion, the term subject, or subject sample of leukocytes, refers to an individual regardless of health and/or disease status. A subject can be a patient, a study participant, a control subject, a screening subject, or any other class of individual from whom a leukocyte sample is obtained and assessed in the context of the invention. Accordingly, a subject can be diagnosed with a disease, can present with one or more symptom of a disease, or a predisposing factor, such as a family (genetic) or medical history (medical) factor, for a disease, or the like. Alternatively, a subject can be healthy with respect to any of the aforementioned factors or criteria. It will be appreciated that the term “healthy” as used herein, is relative to a specified disease, or disease factor, or disease criterion, as the term “healthy” cannot be defined to correspond to any absolute evaluation or status. Thus, an individual defined as healthy with reference to any specified disease or disease criterion, can in fact be diagnosed with any other one or more disease, or exhibit any other one or more disease criterion.

[0226] Furthermore, while the discussion of the invention focuses, and is exemplified using human sequences and samples, the invention is equally applicable, through construction or selection of appropriate candidate libraries, to non-human animals, such as laboratory animals, e.g., mice, rats, guinea pigs, rabbits; domesticated livestock, e.g., cows, horses, goats, sheep, chicken, etc.; and companion animals, e.g., dogs, cats, etc.

Methods for Obtaining Expression Data

[0227] Numerous methods for obtaining expression data are known, and any one or more of these techniques, singly or in combination, are suitable for determining expression profiles in the context of the present invention. For example, expression patterns can be evaluated by northern analysis, PCR, RT-PCR, Taq Man analysis, FRET detection, monitoring one or more molecular beacon, hybridization to an oligonucleotide array, hybridization to a cDNA array, hybridization to a polynucleotide array, hybridization to a liquid microarray, hybridization to a microelectric array, molecular beacons, cDNA sequencing, clone hybridization, cDNA fragment fingerprinting, serial analysis of gene expression (SAGE), subtractive hybridization, differential display and/or differential screening (see, e.g., Lockhart and Winzeler (2000) Nature 405:827–836, and references cited therein).

[0228] For example, specific PCR primers are designed to a member(s) of a candidate nucleotide library. cDNA is prepared from subject sample RNA by reverse transcription from a poly-dT oligonucleotide primer, and subjected to PCR. Double stranded cDNA may be prepared using primers suitable for reverse transcription of the PCR product, followed by amplification of the cDNA using in vitro transcription. The product of in vitro transcription is a sense-RNA corresponding to the original member(s) of the candidate library. PCR product may be also be evaluated in a number of ways known in the art, including real-time assessment using detection of labeled primers, e.g. TaqMan or molecular beacon probes. Technology platforms suitable for analysis of PCR products include the ABI 7700, 5700, or 7000 Sequence Detection Systems (Applied Biosystems, Foster City, Calif.), the MJ Research Opticon (MJ Research, Waltham, Mass.), the Roche Light Cycler (Roche Diagnositics, Indianapolis, Ind.), the Stratagene MX4000 (Stratagene, La Jolla, Calif.), and the Bio-Rad iCycler (Bio-Rad Laboratories, Hercules, Calif.).

[0229] Alternatively, molecular beacons are used to detect presence of a nucleic acid sequence in an unamplified RNA or cDNA sample, or following amplification of the sequence using any method, e.g. IVT (In Vitro transcription) or NASBA (nucleic acid sequence based amplification). Molecular beacons are designed with sequences complementary to member(s) of a candidate nucleotide library, and are linked to fluorescent labels. Each probe has a different fluorescent label with non-overlapping emission wavelengths. For example, expression of ten genes may be assessed using ten different sequence-specific molecular beacons.

[0230] Alternatively, or in addition, molecular beacons are used to assess expression of multiple nucleotide sequences at once. Molecular beacons with sequence complimentary to the members of a diagnostic nucleotide set are designed and linked to fluorescent labels.

[0231] Each fluorescent label used must have a non-overlapping emission wavelength. For example, 10 nucleotide sequences can be assessed by hybridizing 10 sequence specific molecular beacons (each labeled with a different fluorescent molecule) to an amplified or un-amplified RNA or cDNA sample. Such an assay bypasses the need for sample labeling procedures.

[0232] Alternatively, or in addition bead arrays can be used to assess expression of multiple sequences at once. See, e.g, LabMAP 100, Luminex Corp, Austin, Tex.).

[0233] Alternatively, or in addition electric arrays are used to assess expression of multiple sequences, as exemplified by the e-Sensor technology of Motorola (Chicago, Ill.) or Nanochip technology of Nanogen (San Diego, Calif.)

[0234] Of course, the particular method elected will be dependent on such factors as quantity of RNA recovered, practitioner preference, available reagents and equipment, detectors, and the like. Typically, however, the elected method(s) will be appropriate for processing the number of samples and probes of interest. Methods for high-throughput expression analysis are discussed below.

[0235] Alternatively, expression at the level of protein products of gene expression is performed. For example, protein expression, in a sample of leukocytes, can be evaluated by one or more method selected from among: western analysis, two-dimensional gel analysis, chromatographic separation, mass spectrometric detection, protein-fusion reporter constructs, colorimetric assays, binding to a protein array and characterization of polysomal mRNA. One particularly favorable approach involves binding of labeled protein expression products to an array of antibodies specific for members of the candidate library. Methods for producing and evaluating antibodies are widespread in the art, see, e.g., Coligan, supra; and Harlow and Lane (1989) Antibodies: A Laboratory Manual, Cold Spring Harbor Press, NY (“Harlow and Lane”). Additional details regarding a variety of immunological and immunoassay procedures adaptable to the present invention by selection of antibody reagents specific for the products of candidate nucleotide sequences can be found in, e.g., Stites and Terr (eds.)(1991) Basic and Clinical Immunology, 7^th ed., and Paul, supra. Another approach uses systems for performing desorption spectrometry. Commercially available systems, e.g., from Ciphergen Biosystems, Inc. (Fremont, Calif.) are particularly well suited to quantitative analysis of protein expression. Indeed, Protein Chip® arrays (see, e.g., the web site ciphergen.com) used in desorption spectrometry approaches provide arrays for detection of protein expression. Alternatively, affinity reagents, e.g., antibodies, small molecules, etc.) are developed that recognize epitopes of the protein product. Affinity assays are used in protein array assays, e.g. to detect the presence or absence of particular proteins.

[0236] Alternatively, affinity reagents are used to detect expression using the methods described above. In the case of a protein that is expressed on the cell surface of leukocytes, labeled affinity reagents are bound to populations of leukocytes, and leukocytes expressing the protein are identified and counted using fluorescent activated cell sorting (FACS).

[0237] It is appreciated that the methods of expression evaluation discussed herein, although discussed in the context of discovery of diagnostic nucleotide sets, are equally applicable for expression evaluation when using diagnostic nucleotide sets for, e.g. diagnosis of diseases, as further discussed below.

High Throughput Expression Assays

[0238] A number of suitable high throughput formats exist for evaluating gene expression. Typically, the term high throughput refers to a format that performs at least about 100 assays, or at least about 500 assays, or at least about 1000 assays, or at least about 5000 assays, or at least about 10,000 assays, or more per day. When enumerating assays, either the number of samples or the number of candidate nucleotide sequences evaluated can be considered. For example, a northern analysis of, e.g., about 100 samples performed in a gridded array, e.g., a dot blot, using a single probe corresponding to a candidate nucleotide sequence can be considered a high throughput assay. More typically, however, such an assay is performed as a series of duplicate blots, each evaluated with a distinct probe corresponding to a different member of the candidate library. Alternatively, methods that simultaneously evaluate expression of about 100 or more candidate nucleotide sequences in one or more samples, or in multiple samples, are considered high throughput.

[0239] Numerous technological platforms for performing high throughput expression analysis are known. Generally, such methods involve a logical or physical array of either the subject samples, or the candidate library, or both. Common array formats include both liquid and solid phase arrays. For example, assays employing liquid phase arrays, e.g., for hybridization of nucleic acids, binding of antibodies or other receptors to ligand, etc., can be performed in multiwell, or microtiter, plates. Microtiter plates with 96, 384 or 1536 wells are widely available, and even higher numbers of wells, e.g, 3456 and 9600 can be used. In general, the choice of microtiter plates is determined by the methods and equipment, e.g., robotic handling and loading systems, used for sample preparation and analysis. Exemplary systems include, e.g., the ORCA™ system from Beckman-Coulter, Inc. (Fullerton, Calif.) and the Zymate systems from Zymark Corporation (Hopkinton, Mass.).

[0240] Alternatively, a variety of solid phase arrays can favorably be employed in to determine expression patterns in the context of the invention. Exemplary formats include membrane or filter arrays (e.g, nitrocellulose, nylon), pin arrays, and bead arrays (e.g., in a liquid “slurry”). Typically, probes corresponding to nucleic acid or protein reagents that specifically interact with (e.g., hybridize to or bind to) an expression product corresponding to a member of the candidate library are immobilized, for example by direct or indirect cross-linking, to the solid support. Essentially any solid support capable of withstanding the reagents and conditions necessary for performing the particular expression assay can be utilized. For example, functionalized glass, silicon, silicon dioxide, modified silicon, any of a variety of polymers, such as (poly)tetrafluoroethylene, (poly)vinylidenedifluoride, polystyrene, polycarbonate, or combinations thereof can all serve as the substrate for a solid phase array.

[0241] In a preferred embodiment, the array is a “chip” composed, e.g., of one of the above specified materials. Polynucleotide probes, e.g., RNA or DNA, such as cDNA, synthetic oligonucleotides, and the like, or binding proteins such as antibodies, that specifically interact with expression products of individual components of the candidate library are affixed to the chip in a logically ordered manner, i.e., in an array. In addition, any molecule with a specific affinity for either the sense or anti-sense sequence of the marker nucleotide sequence (depending on the design of the sample labeling), can be fixed to the array surface without loss of specific affinity for the marker and can be obtained and produced for array production, for example, proteins that specifically recognize the specific nucleic acid sequence of the marker, ribozymes, peptide nucleic acids (PNA), or other chemicals or molecules with specific affinity.

[0242] Detailed discussion of methods for linking nucleic acids and proteins to a chip substrate, are found in, e.g., U.S. Pat. No. 5,143,854 “LARGE SCALE PHOTOLITHOGRAPHIC SOLID PHASE SYNTHESIS OF POLYPEPTIDES AND RECEPTOR BINDING SCREENING THEREOF” to Pirrung et al., issued, Sep. 1, 1992; U.S. Pat. No. 5,837,832 “ARRAYS OF NUCLEIC ACID PROBES ON BIOLOGICAL CHIPS” to Chee et al., issued Nov. 17, 1998; U.S. Pat. No. 6,087,112 “ARRAYS WITH MODIFIED OLIGONUCLEOTIDE AND POLYNUCLEOTIDE COMPOSITIONS” to Dale, issued Jul. 11, 2000; U.S. Pat. No. 5,215,882 “METHOD OF IMMOBILIZING NUCLEIC ACID ON A SOLID SUBSTRATE FOR USE IN NUCLEIC ACID HYBRIDIZATION ASSAYS” to Bahl et al., issued Jun. 1, 1993; U.S. Pat. No. 5,707,807 “MOLECULAR INDEXING FOR EXPRESSED GENE ANALYSIS” to Kato, issued Jan. 13, 1998; U.S. Pat. No. 5,807,522 “METHODS FOR FABRICATING MICROARRAYS OF BIOLOGICAL SAMPLES” to Brown et al., issued Sep. 15, 1998; U.S. Pat. No. 5,958,342 “JET DROPLET DEVICE” to Gamble et al., issued Sep. 28, 1999; U.S. Pat. No. 5,994,076 “METHODS OF ASSAYING DIFFERENTIAL EXPRESSION” to Chenchik et al., issued Nov. 30, 1999; U.S. Pat. No. 6,004,755 “QUANTITATIVE MICROARRAY HYBRIDIZATION ASSAYS” to Wang, issued Dec. 21, 1999; U.S. Pat. No. 6,048,695 “CHEMICALLY MODIFIED NUCLEIC ACIDS AND METHOD FOR COUPLING NUCLEIC ACIDS TO SOLID SUPPORT” to Bradley et al., issued Apr. 11, 2000; U.S. Pat. No. 6,060,240 “METHODS FOR MEASURING RELATIVE AMOUNTS OF NUCLEIC ACIDS IN A COMPLEX MIXTURE AND RETRIEVAL OF SPECIFIC SEQUENCES THEREFROM” to Kamb et al., issued May 9, 2000; U.S. Pat. No. 6,090,556 “METHOD FOR QUANTITATIVELY DETERMINING THE EXPRESSION OF A GENE” to Kato, issued Jul. 18, 2000; and U.S. Pat. No. 6,040,138 “EXPRESSION MONITORING BY HYBRIDIZATION TO HIGH DENSITY OLIGONUCLEOTIDE ARRAYS” to Lockhart et al., issued Mar. 21, 2000 each of which are hereby incorporated by reference in their entirety.

[0243] For example, cDNA inserts corresponding to candidate nucleotide sequences, in a standard TA cloning vector are amplified by a polymerase chain reaction for approximately 30–40 cycles. The amplified PCR products are then arrayed onto a glass support by any of a variety of well known techniques, e.g., the VSLIPS™ technology described in U.S. Pat. No. 5,143,854. RNA, or cDNA corresponding to RNA, isolated from a subject sample of leukocytes is labeled, e.g., with a fluorescent tag, and a solution containing the RNA (or cDNA) is incubated under conditions favorable for hybridization, with the “probe” chip. Following incubation, and washing to eliminate non-specific hybridization, the labeled nucleic acid bound to the chip is detected qualitatively or quantitatively, and the resulting expression profile for the corresponding candidate nucleotide sequences is recorded. It is appreciated that the probe used for diagnostic purposes may be identical to the probe used during diagnostic nucleotide sequence discovery and validation. Alternatively, the probe sequence may be different than the sequence used in diagnostic nucleotide sequence discovery and validation. Multiple cDNAs from a nucleotide sequence that are non-overlapping or partially overlapping may also be used.

[0244] In another approach, oligonucleotides corresponding to members of a candidate nucleotide library are synthesized and spotted onto an array. Alternatively, oligonucleotides are synthesized onto the array using methods known in the art, e.g. Hughes, et al. supra. The oligonucleotide is designed to be complementary to any portion of the candidate nucleotide sequence. In addition, in the context of expression analysis for, e.g. diagnostic use of diagnostic nucleotide sets, an oligonucleotide can be designed to exhibit particular hybridization characteristics, or to exhibit a particular specificity and/or sensitivity, as further described below.

[0245] Hybridization signal may be amplified using methods known in the art, and as described herein, for example use of the Clontech kit (Glass Fluorescent Labeling Kit), Stratagene kit (Fairplay Microarray Labeling Kit), the Micromax kit (New England Nuclear, Inc.), the Genisphere kit (3DNA Submicro), linear amplification, e.g. as described in U.S. Pat. No. 6,132,997 or described in Hughes, T R, et al., Nature Biotechnology, 19:343–347 (2001) and/or Westin et al. Nat Biotech. 18:199–204.

[0246] Alternatively, fluorescently labeled cDNA are hybridized directly to the microarray using methods known in the art. For example, labeled cDNA are generated by reverse transcription using Cy3- and Cy5-conjugated deoxynucleotides, and the reaction products purified using standard methods. It is appreciated that the methods for signal amplification of expression data useful for identifying diagnostic nucleotide sets are also useful for amplification of expression data for diagnostic purposes.

[0247] Microarray expression may be detected by scanning the microarray with a variety of laser or CCD-based scanners, and extracting features with numerous software packages, for example, Imagene (Biodiscovery), Feature Extraction (Agilent), Scanalyze (Eisen, M. 1999. SCANALYZE User Manual; Stanford Univ., Stanford, Calif. Ver 2.32.), GenePix (Axon Instruments).

[0248] In another approach, hybridization to microelectric arrays is performed, e.g. as described in Umek et al (2001) J Mol Diagn. 3:74–84. An affinity probe, e.g. DNA, is deposited on a metal surface. The metal surface underlying each probe is connected to a metal wire and electrical signal detection system. Unlabelled RNA or cDNA is hybridized to the array, or alternatively, RNA or cDNA sample is amplified before hybridization, e.g. by PCR. Specific hybridization of sample RNA or cDNA results in generation of an electrical signal, which is transmitted to a detector. See Westin (2000) Nat Biotech. 18:199–204 (describing anchored multiplex amplification of a microelectronic chip array); Edman (1997) NAR 25:4907–14; Vignali (2000) J Immunol Methods 243:243–55. In another approach, a microfluidics chip is used for RNA sample preparation and analysis. This approach increases efficiency because sample preparation and analysis are streamlined. Briefly, microfluidics may be used to sort specific leukocyte sub-populations prior to RNA preparation and analysis. Microfluidics chips are also useful for, e.g., RNA preparation, and reactions involving RNA (reverse transcription, RT-PCR). Briefly, a small volume of whole, anti-coagulated blood is loaded onto a microfluidics chip, for example chips available from Caliper (Mountain View, Calif.) or Nanogen (San Diego, Calif.) A microfluidics chip may contain channels and reservoirs in which cells are moved and reactions are performed. Mechanical, electrical, magnetic, gravitational, centrifugal or other forces are used to move the cells and to expose them to reagents. For example, cells of whole blood are moved into a chamber containing hypotonic saline, which results in selective lysis of red blood cells after a 20-minute incubation. Next, the remaining cells (leukocytes) are moved into a wash chamber and finally, moved into a chamber containing a lysis buffer such as guanidine isothiocyanate. The leukocyte cell lysate is further processed for RNA isolation in the chip, or is then removed for further processing, for example, RNA extraction by standard methods.

[0249] Alternatively, the microfluidics chip is a circular disk containing ficoll or another density reagent. The blood sample is injected into the center of the disc, the disc is rotated at a speed that generates a centrifugal force appropriate for density gradient separation of mononuclear cells, and the separated mononuclear cells are then harvested for further analysis or processing.

[0250] It is understood that the methods of expression evaluation, above, although discussed in the context of discovery of diagnostic nucleotide sets, are also applicable for expression evaluation when using diagnostic nucleotide sets for, e.g. diagnosis of diseases, as further discussed below.

Evaluation of Expression Patterns

[0251] Expression patterns can be evaluated by qualitative and/or quantitative measures. Certain of the above described techniques for evaluating gene expression (as RNA or protein products) yield data that are predominantly qualitative in nature. That is, the methods detect differences in expression that classify expression into distinct modes without providing significant information regarding quantitative aspects of expression. For example, a technique can be described as a qualitative technique if it detects the presence or absence of expression of a candidate nucleotide sequence, i.e., an on/off pattern of expression. Alternatively, a qualitative technique measures the presence (and/or absence) of different alleles, or variants, of a gene product.

[0252] In contrast, some methods provide data that characterizes expression in a quantitative manner. That is, the methods relate expression on a numerical scale, e.g., a scale of 0–5, a scale of 1–10, a scale of +–+++, from grade 1 to grade 5, a grade from a to z, or the like. It will be understood that the numerical, and symbolic examples provided are arbitrary, and that any graduated scale (or any symbolic representation of a graduated scale) can be employed in the context of the present invention to describe quantitative differences in nucleotide sequence expression. Typically, such methods yield information corresponding to a relative increase or decrease in expression.

[0253] Any method that yields either quantitative or qualitative expression data is suitable for evaluating expression of candidate nucleotide sequence in a subject sample of leukocytes. In some cases, e.g., when multiple methods are employed to determine expression patterns for a plurality of candidate nucleotide sequences, the recovered data, e.g., the expression profile, for the nucleotide sequences is a combination of quantitative and qualitative data.

[0254] In some applications, expression of the plurality of candidate nucleotide sequences is evaluated sequentially. This is typically the case for methods that can be characterized as low- to moderate-throughput. In contrast, as the throughput of the elected assay increases, expression for the plurality of candidate nucleotide sequences in a sample or multiple samples of leukocytes, is assayed simultaneously. Again, the methods (and throughput) are largely determined by the individual practitioner, although, typically, it is preferable to employ methods that permit rapid, e.g. automated or partially automated, preparation and detection, on a scale that is time-efficient and cost-effective.

[0255] It is understood that the preceding discussion, while directed at the assessment of expression of the members of candidate libraries, is also applies to the assessment of the expression of members of diagnostic nucleotide sets, as further discussed below.

Genotyping

[0256] In addition to, or in conjunction with the correlation of expression profiles and clinical data, it is often desirable to correlate expression patterns with the subject's genotype at one or more genetic loci. The selected loci can be, for example, chromosomal loci corresponding to one or more member of the candidate library, polymorphic alleles for marker loci, or alternative disease related loci (not contributing to the candidate library) known to be, or putatively associated with, a disease (or disease criterion). Indeed, it will be appreciated, that where a (polymorphic) allele at a locus is linked to a disease (or to a predisposition to a disease), the presence of the allele can itself be a disease criterion. Numerous well known methods exist for evaluating the genotype of an individual, including southern analysis, restriction fragment length polymorphism (RFLP) analysis, polymerase chain reaction (PCR), amplification length polymorphism (AFLP) analysis, single stranded conformation polymorphism (SSCP) analysis, single nucleotide polymorphism (SNP) analysis (e.g., via PCR, Taqman or molecular beacons), among many other useful methods. Many such procedures are readily adaptable to high throughput and/or automated (or semi-automated) sample preparation and analysis methods. Most, can be performed on nucleic acid samples recovered via simple procedures from the same sample of leukocytes as yielded the material for expression profiling. Exemplary techniques are described in, e.g., Sambrook, and Ausubel, supra.

Identification of the Diagnostic Nucleotide Sets of the Invention

[0257] Identification of diagnostic nucleotide sets and disease specific target nucleotide sequence proceeds by correlating the leukocyte expression profiles with data regarding the subject's health status to produce a data set designated a “molecular signature.”

[0258] Examples of data regarding a patient's health status, also termed “disease criteria(ion)”, is described below and in the Section titled “selected diseases,” below. Methods useful for correlation analysis are further described elsewhere in the specification.

[0259] Generally, relevant data regarding the subject's health status includes retrospective or prospective health data, e.g., in the form of the subject's medical history, as provided by the subject, physician or third party, such as, medical diagnoses, laboratory test results, diagnostic test results, clinical events, or medication lists, as further described below. Such data may include information regarding a patient's response to treatment and/or a particular medication and data regarding the presence of previously characterized “risk factors.” For example, cigarette smoking and obesity are previously identified risk factors for heart disease. Further examples of health status information, including diseases and disease criteria, is described in the section titled Selected diseases, below. Typically, the data describes prior events and evaluations (i.e., retrospective data). However, it is envisioned that data collected subsequent to the sampling (i.e., prospective data) can also be correlated with the expression profile. The tissue sampled, e.g., peripheral blood, bronchial lavage, etc., can be obtained at one or more multiple time points and subject data is considered retrospective or prospective with respect to the time of sample procurement.

[0260] Data collected at multiple time points, called “longitudinal data”, is often useful, and thus, the invention encompasses the analysis of patient data collected from the same patient at different time points. Analysis of paired samples, such as samples from a patient at different time, allows identification of differences that are specifically related to the disease state since the genetic variability specific to the patient is controlled for by the comparison. Additionally, other variables that exist between patients may be controlled for in this way, for example, the presence or absence of inflammatory diseases (e.g., rheumatoid arthritis) the use of medications that may effect leukocyte gene expression, the presence or absence of co-morbid conditions, etc. Methods for analysis of paired samples are further described below. Moreover, the analysis of a pattern of expression profiles (generated by collecting multiple expression profiles) provides information relating to changes in expression level over time, and may permit the determination of a rate of change, a trajectory, or an expression curve. Two longitudinal samples may provide information on the change in expression of a gene over time, while three longitudinal samples may be necessary to determine the “trajectory” of expression of a gene. Such information may be relevant to the diagnosis of a disease. For example, the expression of a gene may vary from individual to individual, but a clinical event, for example, a heart attack, may cause the level of expression to double in each patient. In this example, clinically interesting information is gleaned from the change in expression level, as opposed to the absolute level of expression in each individual.

[0261] When a single patient sample is obtained, it may still be desirable to compare the expression profile of that sample to some reference expression profile. In this case, one can determine the change of expression between the patient's sample and a reference expression profile that is appropriate for that patient and the medical condition in question. For example, a reference expression profile can be determined for all patients without the disease criterion in question who have similar characteristics, such as age, sex, race, diagnoses etc.

[0262] Generally, small sample sizes of 20–100 samples are used to identify a diagnostic nucleotide set. Larger sample sizes are generally necessary to validate the diagnostic nucleotide set for use in large and varied patient populations, as further described below. For example, extension of gene expression correlations to varied ethnic groups, demographic groups, nations, peoples or races may require expression correlation experiments on the population of interest.

Expression Reference Standards

[0263] Expression profiles derived from a patient (i.e., subjects diagnosed with, or exhibiting symptoms of, or exhibiting a disease criterion, or under a doctor's care for a disease) sample are compared to a control or standard expression RNA to facilitate comparison of expression profiles (e.g. of a set of candidate nucleotide sequences) from a group of patients relative to each other (i.e., from one patient in the group to other patients in the group, or to patients in another group).

[0264] The reference RNA used should have desirable features of low cost and simplicity of production on a large scale. Additionally, the reference RNA should contain measurable amounts of as many of the genes of the candidate library as possible.

[0265] For example, in one approach to identifying diagnostic nucleotide sets, expression profiles derived from patient samples are compared to a expression reference “standard.” Standard expression reference can be, for example, RNA derived from resting cultured leukocytes or commercially available reference RNA, such as Universal reference RNA from Stratagene. See Nature, V406, Aug. 17, 2000, p. 747–752. Use of an expression reference standard is particularly useful when the expression of large numbers of nucleotide sequences is assayed, e.g. in an array, and in certain other applications, e.g. qualitative PCR, RT-PCR, etc., where it is desirable to compare a sample profile to a standard profile, and/or when large numbers of expression profiles, e.g. a patient population, are to be compared. Generally, an expression reference standard should be available in large quantities, should be a good substrate for amplification and labeling reactions, and should be capable of detecting a large percentage of candidate nucleic acids using suitable expression profiling technology.

[0266] Alternatively, or in addition, the expression profile derived from a patient sample is compared with the expression of an internal reference control gene, for example, β-actin or CD4. The relative expression of the profiled genes and the internal reference control gene (from the same individual) is obtained. An internal reference control may also be used with a reference RNA. For example, an expression profile for “gene 1” and the gene encoding CD4 can be determined in a patient sample and in a reference RNA. The expression of each gene can be expressed as the “relative” ratio of expression the gene in the patient sample compared with expression of the gene in the reference RNA. The expression ratio (sample/reference) for gene 1 may be divided by the expression ration for CD4 (sample/reference) and thus the relative expression of gene 1 to CD4 is obtained. The invention also provides a buffy coat control RNA useful for expression profiling, and a method of using control RNA produced from a population of buffy coat cells, the white blood cell layer derived from the centrifugation of whole blood. Buffy coat contains all white blood cells, including granulocytes, mononuclear cells and platelets. The invention also provides a method of preparing control RNA from buffy coat cells for use in expression profile analysis of leukocytes. Buffy coat fractions are obtained, e.g. from a blood bank or directly from individuals, preferably from a large number of individuals such that bias from individual samples is avoided and so that the RNA sample represents an average expression of a healthy population. Buffy coat fractions from about 50 or about 100, or more individuals are preferred. 10 ml buffy coat from each individual is used. Buffy coat samples are treated with an erthythrocyte lysis buffer, so that erthythrocytes are selectively removed. The leukocytes of the buffy coat layer are collected by centrifugation. Alternatively, the buffy cell sample can be further enriched for a particular leukocyte sub-populations, e.g. mononuclear cells, T-lymphocytes, etc.

[0267] To enrich for mononuclear cells, the buffy cell pellet, above, is diluted in PBS (phosphate buffered saline) and loaded onto a non-polystyrene tube containing a polysucrose and sodium diatrizoate solution adjusted to a density of 1.077+/−0.001 g/ml. To enrich for T-lymphocytes, 45 ml of whole blood is treated with RosetteSep (Stem Cell Technologies), and incubated at room temperature for 20 minutes. The mixture is diluted with an equal volume of PBS plus 2% FBS and mixed by inversion. 30 ml of diluted mixture is layered on top of 15 ml DML medium (Stem Cell Technologies). The tube is centrifuged at 1200×g, and the enriched cell layer at the plasma: medium interface is removed, washed with PBS+2% FBS, and cells collected by centrifugation at 1200×g. The cell pellet is treated with 5 ml of erythrocyte lysis buffer (EL buffer, Qiagen) for 10 minutes on ice, and enriched T-lymphocytes are collected by centrifugation.

[0268] In addition or alternatively, the buffy cells (whole buffy coat or sub-population, e.g. mononuclear fraction) can be cultured in vitro and subjected to stimulation with cytokines or activating chemicals such as phorbol esters or ionomycin. Such stimuli may increase expression of nucleotide sequences that are expressed in activated immune cells and might be of interest for leukocyte expression profiling experiments.

[0269] Following sub-population selection and/or further treatment, e.g. stimulation as described above, RNA is prepared using standard methods. For example, cells are pelleted and lysed with a phenol/guanidinium thiocyanate and RNA is prepared. RNA can also be isolated using a silica gel-based purification column or the column method can be used on RNA isolated by the phenol/guanidinium thiocyanate method. RNA from individual buffy coat samples can be pooled during this process, so that the resulting reference RNA represents the RNA of many individuals and individual bias is minimized or eliminated. In addition, a new batch of buffy coat reference RNA can be directly compared to the last batch to ensure similar expression pattern from one batch to another, using methods of collecting and comparing expression profiles described above/below. One or more expression reference controls are used in an experiment. For example, RNA derived from one or more of the following sources can be used as controls for an experiment: stimulated or unstimulated whole buffy coat, stimulated or unstimulated peripheral mononuclear cells, or stimulated or unstimulated T-lymphocytes.

[0270] Alternatively, the expression reference standard can be derived from any subject or class of subjects including healthy subjects or subjects diagnosed with the same or a different disease or disease criterion. Expression profiles from subjects in two distinct classes are compared to determine which subset of nucleotide sequences in the candidate library best distinguish between the two subject classes, as further discussed below. It will be appreciated that in the present context, the term “distinct classes” is relevant to at least one distinguishable criterion relevant to a disease of interest, a “disease criterion.” The classes can, of course, demonstrate significant overlap (or identity) with respect to other disease criteria, or with respect to disease diagnoses, prognoses, or the like. The mode of discovery involves, e.g., comparing the molecular signature of different subject classes to each other (such as patient to control, patients with a first diagnosis to patients with a second diagnosis, etc.) or by comparing the molecular signatures of a single individual taken at different time points. The invention can be applied to a broad range of diseases, disease criteria, conditions and other clinical and/or epidemiological questions, as further discussed above/below.

[0271] It is appreciated that while the present discussion pertains to the use of expression reference controls while identifying diagnostic nucleotide sets, expression reference controls are also useful during use of diagnostic nucleotide sets, e.g. use of a diagnostic nucleotide set for diagnosis of a disease, as further described below.

Analysis of Expression Profiles

[0272] In order to facilitate ready access, e.g., for comparison, review, recovery, and/or modification, the molecular signatures/expression profiles are typically recorded in a database. Most typically, the database is a relational database accessible by a computational device, although other formats, e.g., manually accessible indexed files of expression profiles as photographs, analogue or digital imaging readouts, spreadsheets, etc. can be used. Further details regarding preferred embodiments are provided below. Regardless of whether the expression patterns initially recorded are analog or digital in nature and/or whether they represent quantitative or qualitative differences in expression, the expression patterns, expression profiles (collective expression patterns), and molecular signatures (correlated expression patterns) are stored digitally and accessed via a database. Typically, the database is compiled and maintained at a central facility, with access being available locally and/or remotely.

[0273] As additional samples are obtained, and their expression profiles determined and correlated with relevant subject data, the ensuing molecular signatures are likewise recorded in the database. However, rather than each subsequent addition being added in an essentially passive manner in which the data from one sample has little relation to data from a second (prior or subsequent) sample, the algorithms optionally additionally query additional samples against the existing database to further refine the association between a molecular signature and disease criterion. Furthermore, the data set comprising the one (or more) molecular signatures is optionally queried against an expanding set of additional or other disease criteria. The use of the database in integrated systems and web embodiments is further described below.

Analysis of Expression Profile Data from Arrays

[0274] Expression data is analyzed using methods well known in the art, including the software packages Imagene (Biodiscovery, Marina del Rey, Calif.), Feature Extraction Software (Agilent, Palo Alto, Calif.), and Scanalyze (Stanford University). In the discussion that follows, a “feature” refers to an individual spot of DNA on an array. Each gene may be represented by more than one feature. For example, hybridized microarrays are scanned and analyzed on an Axon Instruments scanner using GenePix 3.0 software (Axon Instruments, Union City, Calif.). The data extracted by GenePix is used for all downstream quality control and expression evaluation. The data is derived as follows. The data for all features flagged as “not found” by the software is removed from the dataset for individual hybridizations. The “not found” flag by GenePix indicates that the software was unable to discriminate the feature from the background. Each feature is examined to determine the value of its signal. The median pixel intensity of the background (B_n) is subtracted from the median pixel intensity of the feature (F_n) to produce the background-subtracted signal (hereinafter, “BGSS”). The BGSS is divided by the standard deviation of the background pixels to provide the signal-to-noise ratio (hereinafter, “S/N”).

[0275] Features with a S/N of three or greater in both the Cy3 channel (corresponding to the sample RNA) and Cy5 channel (corresponding to the reference RNA) are used for further analysis (hereinafter denoted “useable features”). Alternatively, different S/Ns are used for selecting expression data for an analysis. For example, only expression data with signal to noise ratios >3 might be used in an analysis. Alternatively, features with S/N values <3 may be flagged as such and included in the analysis. Such flagged data sets include more values and may allow one to discover expression markers that would be missed otherwise. However, such data sets may have a higher variability than filtered data, which may decrease significance of findings or performance of correlation statistics. For each usable feature (i), the expression level (e) is expressed as the logarithm of the ratio (R) of the Background Subtracted Signal (hereinafter “BGSS”) for the Cy3 (sample RNA) channel divided by the BGSS for the Cy5 channel (reference RNA). This “log ratio” value is used for comparison to other experiments.

[0276] $\begin{array}{cc}{R}_i=\frac{{\mathrm{BGSS}}_{\mathrm{sample}}}{{\mathrm{BGSS}}_{\mathrm{reference}}}& \left(0.1\right)\end{array}$
e_i=log r_i  (0.2)

[0277] Variation in signal across hybridizations may be caused by a number of factors affecting hybridization, DNA spotting, wash conditions, and labeling efficiency.

[0278] A single reference RNA may be used with all of the experimental RNAs, permitting multiple comparisons in addition to individual comparisons. By comparing sample RNAs to the same reference, the gene expression levels from each sample are compared across arrays, permitting the use of a consistent denominator for our experimental ratios.

Scaling

[0279] The data may be scaled (normalized) to control for labeling and hybridization variability within the experiment, using methods known in the art. Scaling is desirable because it facilitates the comparison of data between different experiments, patients, etc. Generally the BGSS are scaled to a factor such as the median, the mean, the trimmed mean, and percentile. Additional methods of scaling include: to scale between 0 and 1, to subtract the mean, or to subtract the median.

[0280] Scaling is also performed by comparison to expression patterns obtained using a common reference RNA, as described in greater detail above. As with other scaling methods, the reference RNA facilitates multiple comparisons of the expression data, e.g., between patients, between samples, etc. Use of a reference RNA provides a consistent denominator for experimental ratios.

[0281] In addition to the use of a reference RNA, individual expression levels may be adjusted to correct for differences in labeling efficiency between different hybridization experiments, allowing direct comparison between experiments with different overall signal intensities, for example. A scaling factor (a) may be used to adjust individual expression levels as follows. The median of the scaling factor (a), for example, BGSS, is determined for the set of all features with a S/N greater than three. Next, the BGSS_i (the BGSS for each feature “i”) is divided by the median for all features (a), generating a scaled ratio. The scaled ration is used to determine the expression value for the feature (e_i), or the log ratio.

[0282] $\begin{array}{cc}{S}_i=\frac{{\mathrm{BGSS}}_i}{a}& \left(0.3\right)\end{array}$

[0283] $\begin{array}{cc}{e}_i=\log \left(\frac{\mathrm{Cy}3S_i}{\mathrm{Cy}5S_i}\right)& \left(0.4\right)\end{array}$

[0284] In addition, or alternatively, control features are used to normalize the data for labeling and hybridization variability within the experiment. Control feature may be cDNA for genes from the plant, Arabidopsis thaliana, that are included when spotting the mini-array. Equal amounts of RNA complementary to control cDNAs are added to each of the samples before they were labeled. Using the signal from these control genes, a normalization constant (L) is determined according to the following formula:

[0285] $L_j=\frac{\frac{\sum _{i=1}^N{\mathrm{BGSS}}_{j,i}}{N}}{\frac{\sum _{j=1}^K\frac{\sum _{i=1}^N{\mathrm{BGSS}}_{j,i}}{N}}{K}}$
where BGSS_i is the signal for a specific feature, N is the number of A. thaliana control features, K is the number of hybridizations, and L_j is the normalization constant for each individual hybridization.

[0286] Using the formula above, the mean for all control features of a particular hybridization and dye (e.g., Cy3) is calculated. The control feature means for all Cy3 hybridizations are averaged, and the control feature mean in one hybridization divided by the average of all hybridizations to generate a normalization constant for that particular Cy3 hybridization (L_j), which is used as a in equation (0.3). The same normalization steps may be performed for Cy3 and Cy5 values.

[0287] An alternative scaling method can also be used. The log of the ratio of Green/Red is determined for all features. The median log ratio value for all features is determined. The feature values are then scaled using the following formula:
Log_Scaled_Feature_Ratio=Log_Feature_Ratio−Median_Log_Ratio.

[0288] Many additional methods for normalization exist and can be applied to the data. In one method, the average ratio of Cy3 BGSS/Cy5 BGSS is determined for all features on an array. This ratio is then scaled to some arbitrary number, such as 1 or some other number. The ratio for each probe is then multiplied by the scaling factor required to bring the average ratio to the chosen level. This is performed for each array in an analysis. Alternatively, the ratios are normalized to the average ratio across all arrays in an analysis.

[0289] If multiple features are used per gene sequence or oligonucleotide, these repeats can be used to derive an average expression value for each gene. If some of the replicate features are of poor quality and don't meet requirements for analysis, the remaining features can be used to represent the gene or gene sequence.

Correlation Analysis

[0290] Correlation analysis is performed to determine which array probes have expression behavior that best distinguishes or serves as markers for relevant groups of samples representing a particular clinical condition. Correlation analysis, or comparison among samples representing different disease criteria (e.g., clinical conditions), is performed using standard statistical methods. Numerous algorithms are useful for correlation analysis of expression data, and the selection of algorithms depends in part on the data analysis to be performed. For example, algorithms can be used to identify the single most informative gene with expression behavior that reliably classifies samples, or to identify all the genes useful to classify samples. Alternatively, algorithms can be applied that determine which set of 2 or more genes have collective expression behavior that accurately classifies samples. The use of multiple expression markers for diagnostics may overcome the variability in expression of a gene between individuals, or overcome the variability intrinsic to the assay. Multiple expression markers may include redundant markers (surrogates), in that two or more genes or probes may provide the same information with respect to diagnosis. This may occur, for example, when two or more genes or gene probes are coordinately expressed. For diagnostic application, it may be appropriate to utilize a gene and one or more of its surrogates in the assay. This redundancy may overcome failures (technical or biological) of a single marker to distinguish samples. Alternatively, one or more surrogates may have properties that make them more suitable for assay development, such as a higher baseline level of expression, better cell specificity, a higher fold change between sample groups or more specific sequence for the design of PCR primers or complimentary probes. It will be appreciated that while the discussion above pertains to the analysis of RNA expression profiles the discussion is equally applicable to the analysis of profiles of proteins or other molecular markers.

[0291] Prior to analysis, expression profile data may be formatted or prepared for analysis using methods known in the art. For example, often the log ratio of scaled expression data for every array probe is calculated using the following formula:

log (Cy 3 BGSS/Cy5 BGSS), where Cy 3 signal corresponds to the expression of the gene in the clinical sample, and Cy5 signal corresponds to expression of the gene in the reference RNA.

[0292] Data may be further filtered depending on the specific analysis to be done as noted below. For example, filtering may be aimed at selecting only samples with expression above a certain level, or probes with variability above a certain level between sample sets.

[0293] The following non-limiting discussion consider several statistical methods known in the art. Briefly, the t-test and ANOVA are used to identify single genes with expression differences between or among populations, respectively. Multivariate methods are used to identify a set of two or more genes for which expression discriminates between two disease states more specifically than expression of any single gene.

t-Test

[0294] The simplest measure of a difference between two groups is the Student's t test. See, e.g., Welsh et al. (2001) Proc Natl Acad Sci USA 98:1176–81 (demonstrating the use of an unpaired Student's t-test for the discovery of differential gene expression in ovarian cancer samples and control tissue samples). The t-test assumes equal variance and normally distributed data. This test identifies the probability that there is a difference in expression of a single gene between two groups of samples. The number of samples within each group that is required to achieve statistical significance is dependent upon the variation among the samples within each group. The standard formula for a t-test is:

[0295] $\begin{array}{cc}t\left(e_i\right)=\frac{{\stackrel{\_}{e}}_{i,c}-{\stackrel{\_}{e}}_{i,t}}{\sqrt{\left(s_{i,c}^2/n_c\right)+\left(s_{i,t}^2/n_t\right)}},& \left(0.5\right)\end{array}$
where ē_i is the difference between the mean expression level of gene i in groups c and t, s_i,c is the variance of gene x in group c and s_i,t is the variance of gene x in group t. n_c and n_t are the numbers of samples in groups c and t.

[0296] The combination of the t statistic and the degrees of freedom [min(n_t, n_c)−1] provides a p value, the probability of rejecting the null hypothesis. A p-value of ≦0.01, signifying a 99 percent probability the mean expression levels are different between the two groups (a 1% chance that the mean expression levels are in fact not different and that the observed difference occurred by statistical chance), is often considered acceptable.

[0297] When performing tests on a large scale, for example, on a large dataset of about 8000 genes, a correction factor must be included to adjust for the number of individual tests being performed. The most common and simplest correction is the Bonferroni correction for multiple tests, which divides the p-value by the number of tests run. Using this test on an 8000 member dataset indicates that a p value of ≦0.00000125 is required to identify genes that are likely to be truly different between the two test conditions.

Significance Analysis for Microarrays (SAM)

[0298] Significance analysis for microarrays (SAM) (Tusher 2001) is a method through which genes with a correlation between their expression values and the response vector are statistically discovered and assigned a statistical significance. The ratio of false significant to significant genes is the False Discovery Rate (FDR). This means that for each threshold there are a set of genes which are called significant, and the FDR gives a confidence level for this claim. If a gene is called differentially expressed between 2 classes by SAM, with a FDR of 5%, there is a 95% chance that the gene is actually differentially expressed between the classes. SAM takes into account the variability and large number of variables of microarrays. SAM will identify genes that are most globally differentially expressed between the classes. Thus, important genes for identifying and classifying outlier samples or patients may not be identified by SAM.

Wilcoxon's Signed Ranks Test

[0299] This method is non-parametric and is utilized for paired comparisons. See e.g., Sokal and Rohlf (1987) Introduction to Biostatistics 2^nd edition, WH Freeman, New York. At least 6 pairs are necessary to apply this statistic. This test is useful for analysis of paired expression data (for example, a set of patients who have cardiac transplant biopsy on 2 occasions and have a grade 0 on one occasion and a grade 3A on another).

ANOVA

[0300] Differences in gene expression across multiple related groups may be assessed using an Analysis of Variance (ANOVA), a method well known in the art (Michelson and Schofield, 1996).

Multivariate Analysis

[0301] Many algorithms suitable for multivariate analysis are known in the art. Generally, a set of two or more genes for which expression discriminates between two disease states more specifically than expression of any single gene is identified by searching through the possible combinations of genes using a criterion for discrimination, for example the expression of gene X must increase from normal 300 percent, while the expression of genes Y and Z must decrease from normal by 75 percent. Ordinarily, the search starts with a single gene, then adds the next best fit at each step of the search. Alternatively, the search starts with all of the genes and genes that do not aid in the discrimination are eliminated step-wise.

Paired Samples

[0302] Paired samples, or samples collected at different time-points from the same patient, are often useful, as described above. For example, use of paired samples permits the reduction of variation due to genetic variation among individuals. In addition, the use of paired samples has a statistical significance, in that data derived from paired samples can be calculated in a different manner that recognizes the reduced variability. For example, the formula for a t-test for paired samples is:

[0303] $\begin{array}{cc}t\left(e_x\right)=\frac{{\stackrel{\_}{D}}_{{\stackrel{\_}{e}}_x}}{\sqrt{\frac{\sum D^2-{\left(\sum D\right)}^2/b}{b-1}}},& \left(0.5\right)\end{array}$
where D is the difference between each set of paired samples and b is the number of sample pairs. D is the mean of the differences between the members of the pairs. In this test, only the differences between the paired samples are considered, then grouped together (as opposed to taking all possible differences between groups, as would be the case with an ordinary t-test). Additional statistical tests useful with paired data, e.g., ANOVA and Wilcoxon's signed rank test, are discussed above.
Diagnostic Classification

[0304] Once a discriminating set of genes is identified, the diagnostic classifier (a mathematical function that assigns samples to diagnostic categories based on expression data) is applied to unknown sample expression levels.

[0305] Methods that can be used for this analysis include the following non-limiting list: CLEAVER is an algorithm used for classification of useful expression profile data. See Raychaudhuri et al. (2001) Trends Biotechnol 19:189–193. CLEAVER uses positive training samples (e.g., expression profiles from samples known to be derived from a particular patient or sample diagnostic category, disease or disease criteria), negative training samples (e.g., expression profiles from samples known not to be derived from a particular patient or sample diagnostic category, disease or disease criteria) and test samples (e.g., expression profiles obtained from a patient), and determines whether the test sample correlates with the particular disease or disease criteria, or does not correlate with a particular disease or disease criteria. CLEAVER also generates a list of the 20 most predictive genes for classification.

[0306] Artificial neural networks (hereinafter, “ANN”) can be used to recognize patterns in complex data sets and can discover expression criteria that classify samples into more than 2 groups. The use of artificial neural networks for discovery of gene expression diagnostics for cancers using expression data generated by oligonucleotide expression microarrays is demonstrated by Khan et al. (2001) Nature Med. 7:673–9. Khan found that 96 genes provided 0% error rate in classification of the tumors. The most important of these genes for classification was then determined by measuring the sensitivity of the classification to a change in expression of each gene. Hierarchical clustering using the 96 genes results in correct grouping of the cancers into diagnostic categories.

[0307] Golub uses cDNA microarrays and a distinction calculation to identify genes with expression behavior that distinguishes myeloid and lymphoid leukemias. See Golub et al. (1999) Science 286:531–7. Self organizing maps were used for new class discovery. Cross validation was done with a “leave one out” analysis. 50 genes were identified as useful markers. This was reduced to as few as 10 genes with equivalent diagnostic accuracy.

[0308] Hierarchical and non-hierarchical clustering methods are also useful for identifying groups of genes that correlate with a subset of clinical samples such as with transplant rejection grade. Alizadeh used hierarchical clustering as the primary tool to distinguish different types of diffuse B-cell lymphomas based on gene expression profile data. See Alizadeh et al. (2000) Nature 403:503–11. Alizadeh used hierarchical clustering as the primary tool to distinguish different types of diffuse B-cell lymphomas based on gene expression profile data. A cDNA array carrying 17856 probes was used for these experiments, 96 samples were assessed on 128 arrays, and a set of 380 genes was identified as being useful for sample classification.

[0309] Perou demonstrates the use of hierarchical clustering for the molecular classification of breast tumor samples based on expression profile data. See Perou et al. (2000) Nature 406:747–52. In this work, a cDNA array carrying 8102 gene probes was used. 1753 of these genes were found to have high variation between breast tumors and were used for the analysis.

[0310] Hastie describes the use of gene shaving for discovery of expression markers. Hastie et al. (2000) Genome Biol. 1(2):RESEARCH 0003.1–0003.21. The gene shaving algorithm identifies sets of genes with similar or coherent expression patterns, but large variation across conditions (RNA samples, sample classes, patient classes). In this manner, genes with a tight expression pattern within a transplant rejection grade, but also with high variability across rejection grades are grouped together. The algorithm takes advantage of both characteristics in one grouping step. For example, gene shaving can identify useful marker genes with co-regulated expression. Sets of useful marker genes can be reduced to a smaller set, with each gene providing some non-redundant value in classification. This algorithm was used on the data set described in Alizadeh et al., supra, and the set of 380 informative gene markers was reduced to 234.

[0311] Supervised harvesting of expression trees (Hastie 2001) identifies genes or clusters that best distinguish one class from all the others on the data set. The method is used to identify the genes/clusters that can best separate one class versus all the others for datasets that include two or more classes or all classes from each other. This algorithm can be used for discovery or testing of a diagnostic gene set.

[0312] CART is a decision tree classification algorithm (Breiman 1984). From gene expression and or other data, CART can develop a decision tree for the classification of samples. Each node on the decision tree involves a query about the expression level of one or more genes or variables. Samples that are above the threshold go down one branch of the decision tree and samples that are not go down the other branch. See examples 10 and 16 for further description of its use in classification analysis and examples of its usefulness in discovering and implementing a diagnostic gene set. CART identifies surrogates for each splitter (genes that are the next best substitute for a useful gene in classification. Multiple Additive Regression Trees (Friedman, J H 1999, MART) is similar to CART in that it is a classification algorithm that builds decision trees to distinguish groups.

[0313] MART builds numerous trees for any classification problem and the resulting model involves a combination of the multiple trees. MART can select variables as it build models and thus can be used on large data sets, such as those derived from an 8000 gene microarray. Because MART uses a combination of many trees and does not take too much information from any one tree, it resists over training. MART identifies a set of genes and an algorithm for their use as a classifier.

[0314] A Nearest Shrunken Centroids Classifier can be applied to microarray or other data sets by the methods described by Tibshirani et al. 2002. This algorithms also identified gene sets for classification and determines their 10 fold cross validation error rates for each class of samples. The algorithm determines the error rates for models of any size, from one gene to all genes in the set. The error rates for either or both sample classes can are minimized when a particular number of genes are used. When this gene number is determined, the algorithm associated with the selected genes can be identified and employed as a classifier on prospective sample.

[0315] Once a set of genes and expression criteria for those genes have been established for classification, cross validation is done. There are many approaches, including a 10 fold cross validation analysis in which 10% of the training samples are left out of the analysis and the classification algorithm is built with the remaining 90%. The 10% are then used as a test set for the algorithm. The process is repeated 10 times with 10% of the samples being left out as a test set each time. Through this analysis, one can derive a cross validation error which helps estimate the robustness of the algorithm for use on prospective (test) samples.

[0316] Clinical data are gathered for every patient sample used for expression analysis. Clinical variables can be quantitative or non-quantitative. A clinical variable that is quantitative can be used as a variable for significance or classification analysis. Non-quantitative clinical variables, such as the sex of the patient, can also be used in a significance analysis or classification analysis with some statistical tool. It is appreciated that the most useful diagnostic gene set for a condition may be optimal when considered along with one or more predictive clinical variables. Clinical data can also be used as supervising vectors for a correlation analysis. That is to say that the clinical data associated with each sample can be used to divide the samples into meaningful diagnostic categories for analysis. For example, samples can be divided into 2 or more groups based on the presence or absence of some diagnostic criterion (a). In addition, clinical data can be utilized to select patients for a correlation analysis or to exclude them based on some undesirable characteristic, such as an ongoing infection, a medicine or some other issue. Clinical data can also be used to assess the pre-test probability of an outcome. For example, patients who are female are much more likely to be diagnosed as having systemic lupus erythematosis than patients who are male.

[0317] Once a set of genes are identified that classify samples with acceptable accuracy. These genes are validated as a set using new samples that were not used to discover the gene set. These samples can be taken from frozen archives from the discovery clinical study or can be taken from new patients prospectively. Validation using a “test set” of samples can be done using expression profiling of the gene set with microarrays or using real-time PCR for each gene on the test set samples. Alternatively, a different expression profiling technology can be used.

Immune Monitoring

[0318] Leukocyte gene expression can be used to monitor the immune system. Gene expression patterns may be associated with activation or the resting state of cells of the immune system that are responsible for or responsive to a disease state. For example, in the process of transplant rejection, cells of the immune system are activated by the presence of the foreign tissue. Genes and gene sets that monitor and diagnose this process are providing a measure of the level and type of activation of the immune system. Genes and gene sets that are useful in monitoring the immune system may be useful for diagnosis and monitoring of all diseases that involve the immune system. Some examples are transplant rejection, rheumatoid arthritis, lupus, inflammatory bowel diseases, multiple sclerosis, HIV/AIDS, and viral, bacterial and fungal infection. All disorders and diseases disclosed herein are contemplated. Genes and gene sets that monitor immune activation are useful for monitoring response to immunosuppressive drug therapy, which is used to decrease immune activation. Genes are found to correlate with immune activation by correlation of expression patterns to the known presence of immune activation or quiescence in a sample as determined by some other test.

Selected Diseases

[0319] In principle, diagnostic nucleotide sets of the invention may be developed and applied to essentially any disease, or disease criterion, as long as at least one subset of nucleotide sequences is differentially expressed in samples derived from one or more individuals with a disease criteria or disease and one or more individuals without the disease criteria or disease, wherein the individual may be the same individual sampled at different points in time, or the individuals may be different individuals (or populations of individuals). For example, the subset of nucleotide sequences may be differentially expressed in the sampled tissues of subjects with the disease or disease criterion (e.g., a patient with a disease or disease criteria) as compared to subjects without the disease or disease criterion (e.g., patients without a disease (control patients)). Alternatively, or in addition, the subset of nucleotide sequence(s) may be differentially expressed in different samples taken from the same patient, e.g at different points in time, at different disease stages, before and after a treatment, in the presence or absence of a risk factor, etc.

[0320] Expression profiles corresponding to sets of nucleotide sequences that correlate not with a diagnosis, but rather with a particular aspect of a disease can also be used to identify the diagnostic nucleotide sets and disease specific target nucleotide sequences of the invention. For example, such an aspect, or disease criterion, can relate to a subject's medical or family history, e.g., childhood illness, cause of death of a parent or other relative, prior surgery or other intervention, medications, symptoms (including onset and/or duration of symptoms), etc. Alternatively, the disease criterion can relate to a diagnosis, e.g., hypertension, diabetes, atherosclerosis, or prognosis (e.g., prediction of future diagnoses, events or complications), e.g., acute myocardial infarction, restenosis following angioplasty, reperfusion injury, allograft rejection, rheumatoid arthritis or systemic lupus erythematosis disease activity or the like. In other cases, the disease criterion corresponds to a therapeutic outcome, e.g., transplant rejection, bypass surgery or response to a medication, restenosis after stent implantation, collateral vessel growth due to therapeutic angiogenesis therapy, decreased angina due to revascularization, resolution of symptoms associated with a myriad of therapies, and the like. Alternatively, the disease criteria corresponds with previously identified or classic risk factors and may correspond to prognosis or future disease diagnosis. As indicated above, a disease criterion can also correspond to genotype for one or more loci. Disease criteria (including patient data) may be collected (and compared) from the same patient at different points in time, from different patients, between patients with a disease (criterion) and patients representing a control population, etc. Longitudinal data, i.e., data collected at different time points from an individual (or group of individuals) may be used for comparisons of samples obtained from an individual (group of individuals) at different points in time, to permit identification of differences specifically related to the disease state, and to obtain information relating to the change in expression over time, including a rate of change or trajectory of expression over time. The usefulness of longitudinal data is further discussed in the section titled “Identification of diagnostic nucleotide sets of the invention”.

[0321] It is further understood that diagnostic nucleotide sets may be developed for use in diagnosing conditions for which there is no present means of diagnosis. For example, in rheumatoid arthritis, joint destruction is often well under way before a patient experience symptoms of the condition. A diagnostic nucleotide set may be developed that diagnoses rheumatic joint destruction at an earlier stage than would be possible using present means of diagnosis, which rely in part on the presentation of symptoms by a patient. Diagnostic nucleotide sets may also be developed to replace or augment current diagnostic procedures. For example, the use of a diagnostic nucleotide set to diagnose cardiac allograft rejection may replace the current diagnostic test, a graft biopsy.

[0322] It is understood that the following discussion of diseases is exemplary and non-limiting, and further that the general criteria discussed above, e.g. use of family medical history, are generally applicable to the specific diseases discussed below.

[0323] In addition to leukocytes, as described throughout, the general method is applicable to nucleotide sequences that are differentially expressed in any subject tissue or cell type, by the collection and assessment of samples of that tissue or cell type. However, in many cases, collection of such samples presents significant technical or medical problems given the current state of the art.

Organ Transplant Rejection and Success

[0324] A frequent complication of organ transplantation is recognition of the transplanted organ as foreign by the immune system resulting in rejection. Diagnostic nucleotide sets can be identified and validated for monitoring organ transplant success, rejection and treatment. Medications currently exist that suppress the immune system, and thereby decrease the rate of and severity of rejection. However, these drugs also suppress the physiologic immune responses, leaving the patient susceptible to a wide variety of opportunistic infections and cancers. At present there is no easy, reliable way to diagnose transplant rejection. Organ biopsy is the preferred method, but this is expensive, painful and associated with significant risk and has inadequate sensitivity for focal rejection. Diagnostic nucleotide sets of the present invention can be developed and validated for use as diagnostic tests for transplant rejection and success. It is appreciated that the methods of identifying diagnostic nucleotide sets are applicable to any organ transplant population. For example, diagnostic nucleotide sets are developed for cardiac allograft rejection and success.

[0325] In some cases, disease criteria correspond to acute stage rejection diagnosis based on organ biopsy and graded using the International Society for Heart and Lung Transplantation (“ISHLT”) criteria. This grading system classifies endomyocardial biopsies on the histological level as Grade 0, 1A, 1B, 2, 3A, 3B, or 4. Grade 0 biopsies have no evidence of rejection, while each successive grade has increased severity of leukocyte infiltration and/or damage to the graft myocardial cells. It is appreciated that there is variability in the Grading systems between medical centers and pathologists and between repeated readings of the same pathologist at different times. When using the biopsy grade as a disease criterion for leukocyte gene expression correlation analysis, it may be desirable to have a single pathologist read all biopsy slides or have multiple pathologists read all slides to determine the variability in this disease criterion. It is also appreciated that cardiac biopsy, in part due to variability, is not 100% sensitive or 100% specific for diagnosing acute rejection. When using the cardiac biopsy grade as a disease criterion for the discovery of diagnostic gene sets, it may be desirable to divide patient samples into diagnostic categories based on the grades. Examples of such classes are those patients with: Grade 0 vs. Grades 1A–4, Grade 0 vs. Grades 1B–4, Grade 0 vs. Grades 2–4, Grade 0–1 vs. Grade 2–4, Grade 0–1 vs. Grade 3A–4, or Grade 0 vs. Grade 3A–4.

[0326] Other disease criteria correspond to the cardiac biopsy results and other criteria, such as the results of cardiac function testing by echocardiography, hemodynamics assessment by cardiac catheterization, CMV infection, weeks post transplant, medication regimen, demographics and/or results of other diagnostic tests.

[0327] Other disease criteria correspond to information from the patient's medical history and information regarding the organ donor. Alternatively, disease criteria include the presence or absence of cytomegalovirus (CMV) infection, Epstein-Barr virus (EBV) infection, allograft dysfunction measured by physiological tests of cardiac function (e.g., hemodynamic measurements from catheterization or echocardiograph data), and symptoms of other infections. Alternatively, disease criteria correspond to therapeutic outcome, e.g. graft failure, re-transplantation, death, hospitalization, need for intravenous immunosuppression, transplant vasculopathy, response to immunosuppressive medications, etc. Disease criteria may further correspond to a rejection episode of at least moderate histologic grade, which results in treatment of the patient with additional corticosteroids, anti-T cell antibodies, or total lymphoid irradiation; a rejection with histologic grade 2 or higher; a rejection with histologic grade <2; the absence of histologic rejection and normal or unchanged allograft function (based on hemodynamic measurements from catheterization or on echocardiographic data); the presence of severe allograft dysfunction or worsening allograft dysfunction during the study period (based on hemodynamic measurements from catheterization or on echocardiographic data); documented CMV infection by culture, histology, or PCR, and at least one clinical sign or symptom of infection; specific graft biopsy rejection grades; rejection of mild to moderate histologic severity prompting augmentation of the patient's chronic immunosuppressive regimen; rejection of mild to moderate severity with allograft dysfunction prompting plasmapheresis or a diagnosis of “humoral” rejection; infections other than CMV, especially infection with Epstein Barr virus (EBV); lymphoproliferative disorder (also called post-transplant lymphoma); transplant vasculopathy diagnosed by increased intimal thickness on intravascular ultrasound (IVUS), angiography, or acute myocardial infarction; graft failure or retransplantation; and all cause mortality. Further specific examples of clinical data useful as disease criteria are provided in Example 9. In another example, diagnostic nucleotide sets are developed and validated for use in diagnosis and monitoring of kidney allograft recipients. Disease criteria correspond to, e.g., results of biopsy analysis for kidney allograft rejection, serum creatine level, creatinine clearance, radiological imaging results for the kidney and urinalysis results. Another disease criterion corresponds to the need for hemodialysis, retransplantation, death or other renal replacement therapy. Diagnostic nucleotide sets are developed and validated for use in diagnosis and treatment of bone marrow transplant and liver transplantation pateints, respectively. Disease criteria for bone marrow transplant correspond to the diagnosis and monitoring of graft rejection and/or graft versus host disease, the recurrence of cancer, complications due to immunosuppression, hematologic abnormalities, infection, hospitalization and/or death. Disease criteria for liver transplant rejection include levels of serum markers for liver damage and liver function such as AST (aspartate aminotransferase), ALT (alanine aminotransferase), Alkaline phosphatase, GGT, (gamma-glutamyl transpeptidase) Bilirubin, Albumin and Prothrombin time. Further disease criteria correspond to hepatic encephalopathy, medication usage, ascites, graft failure, retransplantation, hospitalization, complications of immunosuppression, results of diagnostic tests, results of radiological testing, death and histological rejection on graft biopsy. In addition, urine can be utilized for at the target tissue for profiling in renal transplant, while biliary and intestinal secretions and feces may be used favorably for hepatic or intestinal organ allograft rejection. Diagnostic nucleotide sets can also be discovered and developed for the diagnosis and monitoring of chronic renal allograft rejection.

[0328] In the case of renal allografts, gene expression markers may be identified that are secreted proteins. These proteins may be detected in the urine of allograft recipients using standard immunoassays. Proteins are more likely to be present in the urine if they are of low molecular weight. Lower molecular weight proteins are more likely to pass through the glomerular membrane and into the urine.

[0329] In another example, diagnostic nucleotide sets are developed and validated for use in diagnosis and treatment of xenograft recipients. This can include the transplantation of any organ from a non-human animal to a human or between non-human animals. Considerations for discovery and application of diagnostics and therapeutics and for disease criterion are substantially similar to those for allograft transplantation between humans.

[0330] In another example, diagnostic nucleotide sets are developed and validated for use in diagnosis and treatment of artificial organ recipients. This includes, but is not limited to mechanical circulatory support, artificial hearts, left ventricular assist devices, renal replacement therapies, organ prostheses and the like. Disease criteria are thrombosis (blood clots), infection, death, hospitalization, and worsening measures of organ function (e.g., hemodynamics, creatinine, liver function testing, renal function testing, functional capacity).

[0331] In another example, diagnostic nucleotide sets are developed and validated for use in matching donor organs to appropriate recipients. Diagnostic gene set can be discovered that correlate with successful matching of donor organ to recipient. Disease criteria include graft failure, acute and chronic rejection, death, hospitalization, immunosuppressive drug use, and complications of immunosuppression. Gene sets may be assayed from the donor or recipient's peripheral blood, organ tissue or some other tissue.

[0332] In another example, diagnostic nucleotide sets are developed and validated for use in diagnosis and induction of patient immune tolerance (decrease rejection of an allograft by the host immune system). Disease criteria include rejection, assays of immune activation, need for immunosuppression and all disease criteria noted above for transplantation of each organ.

Viral Diseases

[0333] Diagnostic leukocyte nucleotide sets may be developed and validated for use in diagnosing viral disease. In another aspect, viral nucleotide sequences may be added to a leukocyte nucleotide set for use in diagnosis of viral diseases. Alternatively, viral nucleotide sets and leukocyte nucleotides sets may be used sequentially.

Epstein-Barr Virus (EBV)

[0334] EBV causes a variety of diseases such as mononucleosis, B-cell lymphoma, and pharyngeal carcinoma. It infects mononuclear cells and circulating atypical lymphocytes are a common manifestation of infection. Peripheral leukocyte gene expression is altered by infection. Transplant recipients and patients who are immunosuppressed are at increased risk for EBV-associated lymphoma.

[0335] Diagnostic nucleotide sets may be developed and validated for use in diagnosis and monitoring of EBV. In one aspect, the diagnostic nucleotide set is a leukocyte nucleotide set. Alternatively, EBV nucleotide sequences are added to a leukocyte nucleotide set, for use in diagnosing EBV. Disease criteria correspond with diagnosis of EBV, and, in patients who are EBV-sero-positive, presence (or prospective occurrence) of EBV-related illnesses such as mononucleosis, and EBV-associated lymphoma. Diagnostic nucleotide sets are useful for diagnosis of EBV, and prediction of occurrence of EBV-related illnesses.

Cytomegalovirus (CMV)

[0336] Cytomegalovirus cause inflammation and disease in almost any tissue, particularly the colon, lung, bone marrow and retina, and is a very important cause of disease in immunosuppressed patients, e.g. transplant, cancer, AIDS. Many patients are infected with or have been exposed to CMV, but not all patients develop clinical disease from the virus. Also, CMV negative recipients of allografts that come from CMV positive donors are at high risk for CMV infection. As immunosuppressive drugs are developed and used, it is increasingly important to identify patients with current or impending clinical CMV disease, because the potential benefit of immunosuppressive therapy must be balanced with the increased rate of clinical CMV infection and disease that may result from the use of immunosuppression therapy. CMV may also play a role in the occurrence of atherosclerosis or restenosis after angioplasty.

[0337] Diagnostic nucleotide sets are developed for use in diagnosis and monitoring of CMV infection or re-activation of CMV infection. In one aspect, the diagnostic nucleotide set is a leukocyte nucleotide set. In another aspect, CMV nucleotide sequences are added to a leukocyte nucleotide set, for use in diagnosing CMV. Disease criteria correspond to diagnosis of CMV (e.g., sero-positive state) and presence of clinically active CMV. Disease criteria may also correspond to prospective data, e.g. the likelihood that CMV will become clinically active or impending clinical CMV infection. Antiviral medications are available and diagnostic nucleotide sets can be used to select patients for early treatment, chronic suppression or prophylaxis of CMV activity.

Hepatitis B and C

[0338] These chronic viral infections affect about 1.25 and 2.7 million patients in the US, respectively. Many patients are infected, but suffer no clinical manifestations. Some patients with infection go on to suffer from chronic liver failure, cirrhosis and hepatic carcinoma.

[0339] Diagnostic nucleotide sets are developed for use in diagnosis and monitoring of HBV or HCV infection. In one aspect, the diagnostic nucleotide set is a leukocyte nucleotide set. In another aspect, viral nucleotide sequences are added to a leukocyte nucleotide set, for use in diagnosing the virus and monitoring progression of liver disease. Disease criteria correspond to diagnosis of the virus (e.g., sero-positive state or other disease symptoms). Alternatively, disease criteria correspond to liver damage, e.g., elevated alkaline phosphatase, ALT, AST or evidence of ongoing hepatic damage on liver biopsy.

[0340] Alternatively, disease criteria correspond to serum liver tests (AST, ALT, Alkaline Phosphatase, GGT, PT, bilirubin), liver biopsy, liver ultrasound, viral load by serum PCR, cirrhosis, hepatic cancer, need for hospitalization or listing for liver transplant. Diagnostic nucleotide sets are used to diagnose HBV and HCV, and to predict likelihood of disease progression. Antiviral therapeutic usage, such as Interferon gamma and Ribavirin, can also be disease criteria.

HIV

[0341] HIV infects T cells and certainly causes alterations in leukocyte expression. Diagnostic nucleotide sets are developed for diagnosis and monitoring of HIV. In one aspect, the diagnostic nucleotide set is a leukocyte nucleotide set. In another aspect, viral nucleotide sequences are added to a leukocyte nucleotide set, for use in diagnosing the virus. Disease criteria correspond to diagnosis of the virus (e.g., sero-positive state). In addition, disease criteria correspond to viral load, CD4 T cell counts, opportunistic infection, response to antiretroviral therapy, progression to AIDS, rate of progression and the occurrence of other HIV related outcomes (e.g., malignancy, CNS disturbance). Response to antiretrovirals may also be disease criteria.

Pharmacogenomics

[0342] Pharmocogenomics is the study of the individual propensity to respond to a particular drug therapy (combination of therapies). In this context, response can mean whether a particular drug will work on a particular patient, e.g. some patients respond to one drug but not to another drug. Response can also refer to the likelihood of successful treatment or the assessment of progress in treatment. Titration of drug therapy to a particular patient is also included in this description, e.g. different patients can respond to different doses of a given medication. This aspect may be important when drugs with side-effects or interactions with other drug therapies are contemplated.

[0343] Diagnostic nucleotide sets are developed and validated for use in assessing whether a patient will respond to a particular therapy and/or monitoring response of a patient to drug therapy (therapies). Disease criteria correspond to presence or absence of clinical symptoms or clinical endpoints, presence of side-effects or interaction with other drug(s). The diagnostic nucleotide set may further comprise nucleotide sequences that are targets of drug treatment or markers of active disease.

Validation and Accuracy of Diagnostic Nucleotide Sets

[0344] Prior to widespread application of the diagnostic probe sets of the invention the predictive value of the probe set is validated. When the diagnostic probe set is discovered by microarray based expression analysis, the differential expression of the member genes may be validated by a less variable and more quantitative and accurate technology such as real time PCR. In this type of experiment the amplification product is measured during the PCR reaction. This enables the researcher to observe the amplification before any reagent becomes rate limiting for amplification. In kinetic PCR the measurement is of C_T (threshold cycle) or C_P (crossing point). This measurement (C_T=C_P) is the point at which an amplification curve crosses a threshold fluorescence value. The threshold is set to a point within the area where all of the reactions were in their linear phase of amplification. When measuring C_T, a lower C_T value is indicative of a higher amount of starting material since an earlier cycle number means the threshold was crossed more quickly.

[0345] Several fluorescence methodologies are available to measure amplification product in real-time PCR. Taqman (Applied BioSystems, Foster City, Calif.) uses fluorescence resonance energy transfer (FRET) to inhibit signal from a probe until the probe is degraded by the sequence specific binding and Taq 3′ exonuclease activity. Molecular Beacons (Stratagene, La Jolla, Calif.) also use FRET technology, whereby the fluorescence is measured when a hairpin structure is relaxed by the specific probe binding to the amplified DNA. The third commonly used chemistry is Sybr Green, a DNA-binding dye (Molecular Probes, Eugene, Oreg.). The more amplified product that is produced, the higher the signal. The Sybr Green method is sensitive to non-specific amplification products, increasing the importance of primer design and selection. Other detection chemistries can also been used, such as ethidium bromide or other DNA-binding dyes and many modifications of the fluorescent dye/quencher dye Taqman chemistry, for example scorpions.

[0346] Real-time PCR validation can be done as described in Example 15.

[0347] Typically, the oligonucleotide sequence of each probe is confirmed, e.g. by DNA sequencing using an oligonucleotide-specific primer. Partial sequence obtained is generally sufficient to confirm the identity of the oligonucleotide probe. Alternatively, a complementary polynucleotide is fluorescently labeled and hybridized to the array, or to a different array containing a resynthesized version of the oligonucleotide probe, and detection of the correct probe is confirmed.

[0348] Typically, validation is performed by statistically evaluating the accuracy of the correspondence between the molecular signature for a diagnostic probe set and a selected indicator. For example, the expression differential for a nucleotide sequence between two subject classes can be expressed as a simple ratio of relative expression. The expression of the nucleotide sequence in subjects with selected indicator can be compared to the expression of that nucleotide sequence in subjects without the indicator, as described in the following equations.

ΣE_xai/N=E_xA the average expression of nucleotide sequence x in the members of group A;

ΣE_xbi/M=E_xB the average expression of nucleotide sequence x in the members of group B;

E_xA/ExB=ΔE_xAB the average differential expression of nucleotide sequence x between groups A and B:

where Σ indicates a sum; Ex is the expression of nucleotide sequence x relative to a standard; ai are the individual members of group A, group A has N members; bi are the individual members of group B, group B has M members.

[0349] The expression of at least two nucleotide sequences, e.g., nucleotide sequence X and nucleotide sequence Y are measured relative to a standard in at least one subject of group A (e.g., with a disease) and group B (e.g., without the disease). Ideally, for purposes of validation the indicator is independent from (i.e., not assigned based upon) the expression pattern. Alternatively, a minimum threshold of gene expression for nucleotide sequences X and Y, relative to the standard, are designated for assignment to group A. For nucleotide sequence x, this threshold is designated ΔEx, and for nucleotide sequence y, the threshold is designated ΔEy.

[0350] The following formulas are used in the calculations below:
Sensitivity=(true positives/true positives+false negatives)
Specificity=(true negatives/true negatives+false positives)

[0351] If, for example, expression of nucleotide sequence x above a threshold: x>ΔEx, is observed for 80/100 subjects in group A and for 10/100 subjects in group B, the sensitivity of nucleotide sequence x for the assignment to group A, at the given expression threshold ΔEx, is 80%, and the specificity is 90%.

[0352] If the expression of nucleotide sequence y is >ΔEy in 80/100 subjects in group A, and in 10/100 subjects in group B, then, similarly the sensitivity of nucleotide sequence y for the assignment to group A at the given threshold ΔEy is 80% and the specificity is 90%. If in addition, 60 of the 80 subjects in group A that meet the expression threshold for nucleotide sequence y also meet the expression threshold ΔEx and that 5 of the 10 subjects in group B that meet the expression threshold for nucleotide sequence y also meet the expression threshold ΔEx, the sensitivity of the test (x>ΔEx and y>ΔEy) for assignment of subjects to group A is 60% and the specificity is 95%.

[0353] Alternatively, if the criteria for assignment to group A are change to: Expression of x>ΔEx or expression of y>ΔEy, the sensitivity approaches 100% and the specificity is 85%.

[0354] Clearly, the predictive accuracy of any diagnostic probe set is dependent on the minimum expression threshold selected. The expression of nucleotide sequence X (relative to a standard) is measured in subjects of groups A (with disease) and B (without disease).

[0355] The minimum threshold of nucleotide sequence expression for x, required for assignment to group A is designated ΔEx 1.

[0356] If 90/100 patients in group A have expression of nucleotide sequence x>ΔEx 1 and 20/100 patients in group B have expression of nucleotide sequence x>ΔEx 1, then the sensitivity of the expression of nucleotide sequence x (using ΔEx 1 as a minimum expression threshold) for assignment of patients to group A will be 90% and the specificity will be 80%.

[0357] Altering the minimum expression threshold results in an alteration in the specificity and sensitivity of the nucleotide sequences in question. For example, if the minimum expression threshold of nucleotide sequence x for assignment of subjects to group A is lowered to ΔEx 2, such that 100/100 subjects in group A and 40/100 subjects in group B meet the threshold, then the sensitivity of the test for assignment of subjects to group A will be 100% and the specificity will be 60%.

[0358] Thus, for 2 nucleotide sequences X and Y: the expression of nucleotide sequence x and nucleotide sequence y (relative to a standard) are measured in subjects belonging to groups A (with disease) and B (without disease). Minimum thresholds of nucleotide sequence expression for nucleotide sequences X and Y (relative to common standards) are designated for assignment to group A. For nucleotide sequence x, this threshold is designated ΔEx1 and for nucleotide sequence y, this threshold is designated ΔEy1.

[0359] If in group A, 90/100 patients meet the minimum requirements of expression ΔEx1 and ΔEy1, and in group B, 10/100 subjects meet the minimum requirements of expression ΔEx1 and ΔEy1, then the sensitivity of the test for assignment of subjects to group A is 90% and the specificity is 90%.

[0360] Increasing the minimum expression thresholds for X and Y to ΔEx2 and ΔEy2, such that in group A, 70/100 subjects meet the minimum requirements of expression ΔEx2 and ΔEy2, and in group B, 3/100 subjects meet the minimum requirements of expression ΔEx2 and ΔEy2. Now the sensitivity of the test for assignment of subjects to group A is 70% and the specificity is 97%.

[0361] If the criteria for assignment to group A is that the subject in question meets either threshold, ΔEx2 or ΔEy2, and it is found that 100/100 subjects in group A meet the criteria and 20/100 subjects in group B meet the criteria, then the sensitivity of the test for assignment to group A is 100% and the specificity is 80%.

[0362] Individual components of a diagnostic probe set each have a defined sensitivity and specificity for distinguishing between subject groups. Such individual nucleotide sequences can be employed in concert as a diagnostic probe set to increase the sensitivity and specificity of the evaluation. The database of molecular signatures is queried by algorithms to identify the set of nucleotide sequences (i.e., corresponding to members of the probe set) with the highest average differential expression between subject groups. Typically, as the number of nucleotide sequences in the diagnostic probe set increases, so does the predictive value, that is, the sensitivity and specificity of the probe set. When the probe sets are defined they may be used for diagnosis and patient monitoring as discussed below. The diagnostic sensitivity and specificity of the probe sets for the defined use can be determined for a given probe set with specified expression levels as demonstrated above. By altering the expression threshold required for the use of each nucleotide sequence as a diagnostic, the sensitivity and specificity of the probe set can be altered by the practitioner. For example, by lowering the magnitude of the expression differential threshold for each nucleotide sequence in the set, the sensitivity of the test will increase, but the specificity will decrease. As is apparent from the foregoing discussion, sensitivity and specificity are inversely related and the predictive accuracy of the probe set is continuous and dependent on the expression threshold set for each nucleotide sequence. Although sensitivity and specificity tend to have an inverse relationship when expression thresholds are altered, both parameters can be increased as nucleotide sequences with predictive value are added to the diagnostic nucleotide set. In addition a single or a few markers may not be reliable expression markers across a population of patients. This is because of the variability in expression and measurement of expression that exists between measurements, individuals and individuals over time. Inclusion of a large number of candidate nucleotide sequences or large numbers of nucleotide sequences in a diagnostic nucleotide set allows for this variability as not all nucleotide sequences need to meet a threshold for diagnosis. Generally, more markers are better than a single marker. If many markers are used to make a diagnosis, the likelihood that all expression markers will not meet some thresholds based upon random variability is low and thus the test will give fewer false negatives.

[0363] It is appreciated that the desired diagnostic sensitivity and specificity of the diagnostic nucleotide set may vary depending on the intended use of the set. For example, in certain uses, high specificity and high sensitivity are desired. For example, a diagnostic nucleotide set for predicting which patient population may experience side effects may require high sensitivity so as to avoid treating such patients. In other settings, high sensitivity is desired, while reduced specificity may be tolerated. For example, in the case of a beneficial treatment with few side effects, it may be important to identify as many patients as possible (high sensitivity) who will respond to the drug, and treatment of some patients who will not respond is tolerated. In other settings, high specificity is desired and reduced sensitivity may be tolerated. For example, when identifying patients for an early-phase clinical trial, it is important to identify patients who may respond to the particular treatment. Lower sensitivity is tolerated in this setting as it merely results in reduced patients who enroll in the study or requires that more patients are screened for enrollment.

Methods of Using Diagnostic Nucleotide Sets.

[0364] The invention also provide methods of using the diagnostic nucleotide sets to: diagnose disease; assess severity of disease; predict future occurrence of disease; predict future complications of disease; determine disease prognosis; evaluate the patient's risk, or “stratify” a group of patients; assess response to current drug therapy; assess response to current non-pharmacological therapy; determine the most appropriate medication or treatment for the patient; predict whether a patient is likely to respond to a particular drug; and determine most appropriate additional diagnostic testing for the patient, among other clinically and epidemiologically relevant applications.

[0365] The nucleotide sets of the invention can be utilized for a variety of purposes by physicians, healthcare workers, hospitals, laboratories, patients, companies and other institutions. As indicated previously, essentially any disease, condition, or status for which at least one nucleotide sequence is differentially expressed in leukocyte populations (or sub-populations) can be evaluated, e.g., diagnosed, monitored, etc. using the diagnostic nucleotide sets and methods of the invention. In addition to assessing health status at an individual level, the diagnostic nucleotide sets of the present invention are suitable for evaluating subjects at a “population level,” e.g., for epidemiological studies, or for population screening for a condition or disease.

Collection and Preparation of Sample

[0366] RNA, protein and/or DNA is prepared using methods well-known in the art, as further described herein. It is appreciated that subject samples collected for use in the methods of the invention are generally collected in a clinical setting, where delays may be introduced before RNA samples are prepared from the subject samples of whole blood, e.g. the blood sample may not be promptly delivered to the clinical lab for further processing. Further delay may be introduced in the clinical lab setting where multiple samples are generally being processed at any given time. For this reason, methods which feature lengthy incubations of intact leukocytes at room temperature are not preferred, because the expression profile of the leukocytes may change during this extended time period. For example, RNA can be isolated from whole blood using a phenol/guanidine isothiocyanate reagent or another direct whole-blood lysis method, as described in, e.g., U.S. Pat. Nos. 5,346,994 and 4,843,155. This method may be less preferred under certain circumstances because the large majority of the RNA recovered from whole blood RNA extraction comes from erythrocytes since these cells outnumber leukocytes 1000:1. Care must be taken to ensure that the presence of erythrocyte RNA and protein does not introduce bias in the RNA expression profile data or lead to inadequate sensitivity or specificity of probes.

[0367] Alternatively, intact leukocytes may be collected from whole blood using a lysis buffer that selectively lyses erythrocytes, but not leukocytes, as described, e.g., in (U.S. Pat. Nos. 5,973,137, and 6,020,186). Intact leukocytes are then collected by centrifugation, and leukocyte RNA is isolated using standard protocols, as described herein. However, this method does not allow isolation of sub-populations of leukocytes, e.g. mononuclear cells, which may be desired. In addition, the expression profile may change during the lengthy incubation in lysis buffer, especially in a busy clinical lab where large numbers of samples are being prepared at any given time.

[0368] Alternatively, specific leukocyte cell types can be separated using density gradient reagents (Boyum, A, 1968.). For example, mononuclear cells may be separated from whole blood using density gradient centrifugation, as described, e.g., in U.S. Pat. Nos. 4,190,535, 4,350,593, 4,751,001, 4,818,418, and 5,053,134. Blood is drawn directly into a tube containing an anticoagulant and a density reagent (such as Ficoll or Percoll).

[0369] Centrifugation of this tube results in separation of blood into an erythrocyte and granulocyte layer, a mononuclear cell suspension, and a plasma layer. The mononuclear cell layer is easily removed and the cells can be collected by centrifugation, lysed, and frozen. Frozen samples are stable until RNA can be isolated. Density centrifugation, however, must be conducted at room temperature, and if processing is unduly lengthy, such as in a busy clinical lab, the expression profile may change.

[0370] Alternatively, cells can be separated using fluorescence activated cell sorting (FACS) or some other technique, which divides cells into subsets based on gene or protein expression. This may be desirable to enrich the sample for cells of interest, but it may also introduce cell manipulations and time delays, which result in alteration of gene expression profiles (Cantor et al. 1975; Galbraith et al. 1999).

[0371] The quality and quantity of each clinical RNA sample is desirably checked before amplification and labeling for array hybridization, using methods known in the art. For example, one microliter of each sample may be analyzed on a Bioanalyzer (Agilent 2100 Palo Alto, Calif. USA) using an RNA 6000 nano LabChip (Caliper, Mountain View, Calif. USA). Degraded RNA is identified by the reduction of the 28S to 18S ribosomal RNA ratio and/or the presence of large quantities of RNA in the 25–100 nucleotide range.

[0372] It is appreciated that the RNA sample for use with a diagnostic nucleotide set may be produced from the same or a different cell population, sub-population and/or cell type as used to identify the diagnostic nucleotide set. For example, a diagnostic nucleotide set identified using RNA extracted from mononuclear cells may be suitable for analysis of RNA extracted from whole blood or mononuclear cells, depending on the particular characteristics of the members of the diagnostic nucleotide set. Generally, diagnostic nucleotide sets must be tested and validated when used with RNA derived from a different cell population, sub-population or cell type than that used when obtaining the diagnostic gene set. Factors such as the cell-specific gene expression of diagnostic nucleotide set members, redundancy of the information provided by members of the diagnostic nucleotide set, expression level of the member of the diagnostic nucleotide set, and cell-specific alteration of expression of a member of the diagnostic nucleotide set will contribute to the usefulness of using a different RNA source than that used when identifying the members of the diagnostic nucleotide set. It is appreciated that it may be desirable to assay RNA derived from whole blood, obviating the need to isolate particular cell types from the blood.

Rapid Method of RNA Extraction Suitable for Production in a Clinical Setting of High Quality RNA for Expression Profiling

[0373] In a clinical setting, obtaining high quality RNA preparations suitable for expression profiling, from a desired population of leukocytes poses certain technical challenges, including: the lack of capacity for rapid, high-throughput sample processing in the clinical setting, and the possibility that delay in processing (in a busy lab or in the clinical setting) may adversely affect RNA quality, e.g. by a permitting the expression profile of certain nucleotide sequences to shift. Also, use of toxic and expensive reagents, such as phenol, may be disfavored in the clinical setting due to the added expense associated with shipping and handling such reagents.

[0374] A useful method for RNA isolation for leukocyte expression profiling would allow the isolation of monocyte and lymphocyte RNA in a timely manner, while preserving the expression profiles of the cells, and allowing inexpensive production of reproducible high-quality RNA samples. Accordingly, the invention provides a method of adding inhibitor(s) of RNA transcription and/or inhibitor(s) of protein synthesis, such that the expression profile is “frozen” and RNA degradation is reduced. A desired leukocyte population or sub-population is then isolated, and the sample may be frozen or lysed before further processing to extract the RNA. Blood is drawn from subject population and exposed to ActinomycinD (to a final concentration of 10 ug/ml) to inhibit transcription, and cycloheximide (to a final concentration of 10 ug/ml) to inhibit protein synthesis. The inhibitor(s) can be injected into the blood collection tube in liquid form as soon as the blood is drawn, or the tube can be manufactured to contain either lyophilized inhibitors or inhibitors that are in solution with the anticoagulant. At this point, the blood sample can be stored at room temperature until the desired leukocyte population or sub-population is isolated, as described elsewhere. RNA is isolated using standard methods, e.g., as described above, or a cell pellet or extract can be frozen until further processing of RNA is convenient.

[0375] The invention also provides a method of using a low-temperature density gradient for separation of a desired leukocyte sample. In another embodiment, the invention provides the combination of use of a low-temperature density gradient and the use of transcriptional and/or protein synthesis inhibitor(s). A desired leukocyte population is separated using a density gradient solution for cell separation that maintains the required density and viscosity for cell separation at 0–4° C. Blood is drawn into a tube containing this solution and may be refrigerated before and during processing as the low temperatures slow cellular processes and minimize expression profile changes.

[0376] Leukocytes are separated, and RNA is isolated using standard methods. Alternately, a cell pellet or extract is frozen until further processing of RNA is convenient. Care must be taken to avoid rewarming the sample during further processing steps.

[0377] Alternatively, the invention provides a method of using low-temperature density gradient separation, combined with the use of actinomycin A and cyclohexamide, as described above.

Assessing Expression for Diagnostics

[0378] Expression profiles for the set of diagnostic nucleotide sequences in a subject sample can be evaluated by any technique that determines the expression of each component nucleotide sequence. Methods suitable for expression analysis are known in the art, and numerous examples are discussed in the Sections titled “Methods of obtaining expression data” and “high throughput expression Assays”, above.

[0379] In many cases, evaluation of expression profiles is most efficiently, and cost effectively, performed by analyzing RNA expression. Alternatively, the proteins encoded by each component of the diagnostic nucleotide set are detected for diagnostic purposes by any technique capable of determining protein expression, e.g., as described above.

[0380] Expression profiles can be assessed in subject leukocyte sample using the same or different techniques as those used to identify and validate the diagnostic nucleotide set. For example, a diagnostic nucleotide set identified as a subset of sequences on a cDNA microarray can be utilized for diagnostic (or prognostic, or monitoring, etc.) purposes on the same array from which they were identified. Alternatively, the diagnostic nucleotide sets for a given disease or condition can be organized onto a dedicated sub-array for the indicated purpose. It is important to note that if diagnostic nucleotide sets are discovered using one technology, e.g. RNA expression profiling, but applied as a diagnostic using another technology, e.g. protein expression profiling, the nucleotide sets must generally be validated for diagnostic purposes with the new technology. In addition, it is appreciated that diagnostic nucleotide sets that are developed for one use, e.g. to diagnose a particular disease, may later be found to be useful for a different application, e.g. to predict the likelihood that the particular disease will occur. Generally, the diagnostic nucleotide set will need to be validated for use in the second circumstance. As discussed herein, the sequence of diagnostic nucleotide set members may be amplified from RNA or cDNA using methods known in the art providing specific amplification of the nucleotide sequences.

Identification of Novel Nucleotide Sequences that are Differentially Expressed in Leukocytes

[0381] Novel nucleotide sequences that are differentially expressed in leukocytes are also part of the invention. Previously unidentified open reading frames may be identified in a library of differentially expressed candidate nucleotide sequences, as described above, and the DNA and predicted protein sequence may be identified and characterized as noted above. We identified unnamed (not previously described as corresponding to a gene, or an expressed gene) nucleotide sequences in the our candidate nucleotide library, depicted in Table 3A, 3B, Tables 8, 11–12, 14, 15, and 19, and the sequence listing. Accordingly, further embodiments of the invention are the isolated nucleic acids described in Tables 3A and 3B, and in the sequence listing. The novel differentially expressed nucleotide sequences of the invention are useful in the diagnostic nucleotide set of the invention described above, and are further useful as members of a diagnostic nucleotide set immobilized on an array. The novel partial nucleotide sequences may be further characterized using sequence tools and publicly or privately accessible sequence databases, as is well known in the art: Novel differentially expressed nucleotide sequences may be identified as disease target nucleotide sequences, described below. Novel nucleotide sequences may also be used as imaging reagent, as further described below.

[0382] As used herein, “nucleotide sequence” refers to (a) a nucleotide sequence containing at least one of the DNA sequences disclosed herein (as shown in the Figures, Table 3A, 3B, Tables 8, 11–12, 14, 15, and 19, and the sequence listing); (b) any DNA sequence that encodes the amino acid sequence encoded by the DNA sequences disclosed herein; (c) any DNA sequence that hybridizes to the complement of the coding sequences disclosed herein, contained within the coding region of the nucleotide sequence to which the DNA sequences disclosed herein (as shown in Table 3A, 3B, Tables 8, 11–12, 14, 15, and 19, and the sequence listing) belong, under highly stringent conditions, e.g., hybridization to filter-bound DNA in 0.5 M NaHPO₄, 7% sodium dodecyl sulfate (SDS), 1 mM EDTA at 65° C., and washing in 0.1×SSC/0.1% SDS at 68° C. (Ausubel F. M. et al., eds., 1989, Current Protocols in Molecular Biology, Vol. I, Green Publishing Associates, Inc., and John Wiley & sons, Inc., New York, at p. 2.10.3), (d) any DNA sequence that hybridizes to the complement of the coding sequences disclosed herein, (as shown in Table 3A, 3B, Tables 8, 11–12, 14, 15, and 19, and the sequence listing) contained within the coding region of the nucleotide sequence to which DNA sequences disclosed herein (as shown in TABLES 3A, 3B, Tables 8, 11–12, 14, 15, and 19,) belong, under less stringent conditions, such as moderately stringent conditions, e.g., washing in 0.2×SSC/0.1% SDS at 42° C. (Ausubel et al., 1989, supra), yet which still encodes a functionally equivalent gene product; and/or (e) any DNA sequence that is at least 90% identical, at least 80% identical or at least 70% identical to the coding sequences disclosed herein (as shown in TABLES 3A, 3B Tables 8, 11–12, 14, 15, and 19, and the sequence listing), wherein % identity is determined using standard algorithms known in the art. For example, the BLAST approach, as detailed in Karlin, S. and S. F. Altschul (1993; Proc. Nat. Acad. Sci. 90:5873–7) and incorporated herein by reference, searches for matches between a query sequence and a database sequence, to evaluate the statistical significance of any matches found, and to report only those matches which satisfy the user-selected threshold of significance. The threshold may be set, for example, at 10⁻²⁵ for nucleotides and 10⁻¹⁴ for proteins.

[0383] The invention also includes nucleic acid molecules, preferably DNA molecules, that hybridize to, and are therefore the complements of, the DNA sequences (a) through (c), in the preceding paragraph. Such hybridization conditions may be highly stringent or less highly stringent, as described above. In instances wherein the nucleic acid molecules are deoxyoligonucleotides (“oligos”), highly stringent conditions may refer, e.g., to washing in 6×SSC/0.05% sodium pyrophosphate at 37° C. (for 14-base oligos), 48° C. (for 17-base oligos), 55° C. (for 20-base oligos), and 60° C. (for 23-base oligos). These nucleic acid molecules may act as target nucleotide sequence antisense molecules, useful, for example, in target nucleotide sequence regulation and/or as antisense primers in amplification reactions of target nucleotide sequence nucleic acid sequences. Further, such sequences may be used as part of ribozyme and/or triple helix sequences, also useful for target nucleotide sequence regulation. Still further, such molecules may be used as components of diagnostic methods whereby the presence of a disease-causing allele, may be detected.

[0384] The invention also encompasses nucleic acid molecules contained in full-length gene sequences that are related to or derived from sequences in Tables 2, 3, Tables 8, 11–12, 14, 15, and 19, and the sequence listing. One sequence may map to more than one full-length gene.

[0385] The invention also encompasses (a) DNA vectors that contain any of the foregoing coding sequences and/or their complements (i.e., antisense); (b) DNA expression vectors that contain any of the foregoing coding sequences operatively associated with a regulatory element that directs the expression of the coding sequences; and (c) genetically engineered host cells that contain any of the foregoing coding sequences operatively associated with a regulatory element that directs the expression of the coding sequences in the host cell. As used herein, regulatory elements include but are not limited to inducible and non-inducible promoters, enhancers, operators and other elements known to those skilled in the art that drive and regulate expression. The invention includes fragments of any of the DNA sequences disclosed herein. Fragments of the DNA sequences may be at least 5, at least 10, at least 15, at least 19 nucleotides, at least 25 nucleotides, at least 50 nucleotides, at least 100 nucleotides, at least 200, at least 500, or larger.

[0386] In addition to the nucleotide sequences described above, homologues of such sequences, as may, for example be present in other species, may be identified and may be readily isolated, without undue experimentation, by molecular biological techniques well known in the art, as well as use of gene analysis tools described above, and e.g., in Example 4. Further, there may exist nucleotide sequences at other genetic loci within the genome that encode proteins which have extensive homology to one or more domains of such gene products. These nucleotide sequences may also be identified via similar techniques.

[0387] For example, the isolated differentially expressed nucleotide sequence may be labeled and used to screen a cDNA library constructed from mRNA obtained from the organism of interest. Hybridization conditions will be of a lower stringency when the cDNA library was derived from an organism different from the type of organism from which the labeled sequence was derived. Alternatively, the labeled fragment may be used to screen a genomic library derived from the organism of interest, again, using appropriately stringent conditions. Such low stringency conditions will be well known to those of skill in the art, and will vary predictably depending on the specific organisms from which the library and the labeled sequences are derived. For guidance regarding such conditions see, for example, Sambrook et al., 1989, Molecular Cloning, A Laboratory Manual, Cold Springs Harbor Press, N.Y.; and Ausubel et al., 1989, Current Protocols in Molecular Biology, Green Publishing Associates and Wiley Interscience, N.Y.

Identification of Novel Proteins that are Differentially Expressed in Leukocytes

[0388] Novel nucleotide products include those proteins encoded by the novel nucleotide sequences described, above. Specifically, novel gene products may include polypeptides encoded by the novel nucleotide sequences contained in the coding regions of the nucleotide sequences to which DNA sequences disclosed herein (in TABLES 3A, 3B, Tables 8, 11–12, 14, 15, and 19, and the sequence listing).

[0389] In addition, novel protein products of novel nucleotide sequences may include proteins that represent functionally equivalent gene products. Such an equivalent novel gene product may contain deletions, additions or substitutions of amino acid residues within the amino acid sequence encoded by the novel nucleotide sequences described, above, but which result in a silent change, thus producing a functionally equivalent novel nucleotide sequence product. Amino acid substitutions may be made on the basis of similarity in polarity, charge, solubility, hydrophobicity, hydrophilicity, and/or the amphipathic nature of the residues involved.

[0390] For example, nonpolar (hydrophobic) amino acids include alanine, leucine, isoleucine, valine, proline, phenylalanine, tryptophan, and methionine; polar neutral amino acids include glycine, serine, threonine, cysteine, tyrosine, asparagine, and glutamine; positively charged (basic) amino acids include arginine, lysine, and histidine; and negatively charged (acidic) amino acids include aspartic acid and glutamic acid.

[0391] “Functionally equivalent”, as utilized herein, refers to a protein capable of exhibiting a substantially similar in vivo activity as the endogenous novel gene products encoded by the novel nucleotide described, above.

[0392] The novel gene products (protein products of the novel nucleotide sequences) may be produced by recombinant DNA technology using techniques well known in the art.

[0393] Thus, methods for preparing the novel gene polypeptides and peptides of the invention by expressing nucleic acid encoding novel nucleotide sequences are described herein.

[0394] Methods which are well known to those skilled in the art can be used to construct expression vectors containing novel nucleotide sequence protein coding sequences and appropriate transcriptional/translational control signals. These methods include, for example, in vitro recombinant DNA techniques, synthetic techniques and in vivo recombination/genetic recombination. See, for example, the techniques described in Sambrook et al., 1989, supra, and Ausubel et al., 1989, supra. Alternatively, RNA capable of encoding novel nucleotide sequence protein sequences may be chemically synthesized using, for example, synthesizers. See, for example, the techniques described in “Oligonucleotide Synthesis”, 1984, Gait, M. J. ed., IRL Press, Oxford, which is incorporated by reference herein in its entirety

[0395] A variety of host-expression vector systems may be utilized to express the novel nucleotide sequence coding sequences of the invention. Such host-expression systems represent vehicles by which the coding sequences of interest may be produced and subsequently purified, but also represent cells which may, when transformed or transfected with the appropriate nucleotide coding sequences, exhibit the novel protein encoded by the novel nucleotide sequence of the invention in situ. These include but are not limited to microorganisms such as bacteria (e.g., E. coli, B. subtilis) transformed with recombinant bacteriophage DNA, plasmid DNA or cosmid DNA expression vectors containing novel nucleotide sequence protein coding sequences; yeast (e.g. Saccharomyces, Pichia) transformed with recombinant yeast expression vectors containing the novel nucleotide sequence protein coding sequences; insect cell systems infected with recombinant virus expression vectors (e.g., baculovirus) containing the novel nucleotide sequence protein coding sequences; plant cell systems infected with recombinant virus expression vectors (e.g., cauliflower mosaic virus, CaMV; tobacco mosaic virus, TMV) or transformed with recombinant plasmid expression vectors (e.g., Ti plasmid) containing novel nucleotide sequence protein coding sequences; or mammalian cell systems (e.g. COS, CHO, BHK, 293, 3T3) harboring recombinant expression constructs containing promoters derived from the genome of mammalian cells (e.g., metallothionein promoter) or from mammalian viruses (e.g., the adenovirus late promoter; the vaccinia virus 7.5 K promoter).

[0396] In bacterial systems, a number of expression vectors may be advantageously selected depending upon the use intended for the novel nucleotide sequence protein being expressed. For example, when a large quantity of such a protein is to be produced, for the generation of antibodies or to screen peptide libraries, for example, vectors which direct the expression of high levels of fusion protein products that are readily purified may be desirable. Such vectors include, but are not limited, to the E. coli expression vector pUR278 (Ruther et al., 1983, EMBO J. 2:1791), in which the novel nucleotide sequence protein coding sequence may be ligated individually into the vector in frame with the lac Z coding region so that a fusion protein is produced; pIN vectors (Inouye & Inouye, 1985, Nucleic Acids Res. 13:3101–3109; Van Heeke & Schuster, 1989, J. Biol. Chem. 264:5503–5509); and the likes of pGEX vectors may also be used to express foreign polypeptides as fusion proteins with glutathione S-transferase (GST). In general, such fusion proteins are soluble and can easily be purified from lysed cells by adsorption to glutathione-agarose beads followed by elution in the presence of free glutathione. The pGEX vectors are designed to include thrombin or factor Xa protease cleavage sites so that the cloned target nucleotide sequence protein can be released from the GST moiety. Other systems useful in the invention include use of the FLAG epitope or the 6-HIS systems.

[0397] In an insect system, Autographa californica nuclear polyhedrosis virus (AcNPV) is used as a vector to express foreign nucleotide sequences. The virus grows in Spodoptera frugiperda cells. The novel nucleotide sequence coding sequence may be cloned individually into non-essential regions (for example the polyhedrin gene) of the virus and placed under control of an AcNPV promoter (for example the polyhedrin promoter).

[0398] Successful insertion of novel nucleotide sequence coding sequence will result in inactivation of the polyhedrin gene and production of non-occluded recombinant virus (i.e., virus lacking the proteinaceous coat coded for by the polyhedrin gene). These recombinant viruses are then used to infect Spodoptera frugiperda cells in which the inserted nucleotide sequence is expressed. (E.g., see Smith et al., 1983, J. Virol. 46: 584; Smith, U.S. Pat. No. 4,215,051).

[0399] In mammalian host cells, a number of viral-based expression systems may be utilized. In cases where an adenovirus is used as an expression vector, the novel nucleotide sequence coding sequence of interest may be ligated to an adenovirus transcription/translation control complex, e.g., the late promoter and tripartite leader sequence. This chimeric nucleotide sequence may then be inserted in the adenovirus genome by in vitro or in vivo recombination. Insertion in a non-essential region of the viral genome (e.g., region E1 or E3) will result in a recombinant virus that is viable and capable of expressing novel nucleotide sequence encoded protein in infected hosts. (E.g., See Logan & Shenk, 1984, Proc. Natl. Acad. Sci. USA 81:3655–3659). Specific initiation signals may also be required for efficient translation of inserted novel nucleotide sequence coding sequences. These signals include the ATG initiation codon and adjacent sequences. In cases where an entire novel nucleotide sequence, including its own initiation codon and adjacent sequences, is inserted into the appropriate expression vector, no additional translational control signals may be needed. However, in cases where only a portion of the novel nucleotide sequence coding sequence is inserted, exogenous translational control signals, including, perhaps, the ATG initiation codon, must be provided. Furthermore, the initiation codon must be in phase with the reading frame of the desired coding sequence to ensure translation of the entire insert. These exogenous translational control signals and initiation codons can be of a variety of origins, both natural and synthetic. The efficiency of expression may be enhanced by the inclusion of appropriate transcription enhancer elements, transcription terminators, etc. (see Bittner et al., 1987, Methods in Enzymol. 153:516–544).

[0400] In addition, a host cell strain may be chosen which modulates the expression of the inserted sequences, or modifies and processes the product of the nucleotide sequence in the specific fashion desired. Such modifications (e.g., glycosylation) and processing (e.g., cleavage) of protein products may be important for the function of the protein. Different host cells have characteristic and specific mechanisms for the post-translational processing and modification of proteins. Appropriate cell lines or host systems can be chosen to ensure the correct modification and processing of the foreign protein expressed. To this end, eukaryotic host cells which possess the cellular machinery for proper processing of the primary transcript, glycosylation, and phosphorylation of the gene product may be used. Such mammalian host cells include but are not limited to CHO, VERO, BHK, HeLa, COS, MDCK, 293, 3T3, WI38, etc.

[0401] For long-term, high-yield production of recombinant proteins, stable expression is preferred. For example, cell lines which stably express the novel nucleotide sequence encoded protein may be engineered. Rather than using expression vectors which contain viral origins of replication, host cells can be transformed with DNA controlled by appropriate expression control elements (e.g., promoter, enhancer, sequences, transcription terminators, polyadenylation sites, etc.), and a selectable marker. Following the introduction of the foreign DNA, engineered cells may be allowed to grow for 1–2 days in an enriched media, and then are switched to a selective media. The selectable marker in the recombinant plasmid confers resistance to the selection and allows cells to stably integrate the plasmid into their chromosomes and grow to form foci which in turn can be cloned and expanded into cell lines. This method may advantageously be used to engineer cell lines which express novel nucleotide sequence encoded protein. Such engineered cell lines may be particularly useful in screening and evaluation of compounds that affect the endogenous activity of the novel nucleotide sequence encoded protein.

[0402] A number of selection systems may be used, including but not limited to the herpes simplex virus thymidine kinase (Wigler, et al., 1977, Cell 11:223), hypoxanthine-guanine phosphoribosyltransferase (Szybalska & Szybalski, 1962, Proc. Natl. Acad. Sci. USA 48:2026), and adenine phosphoribosyltransferase (Lowy, et al., 1980, Cell 22:817) genes can be employed in tk-, hgprt- or aprt-cells, respectively. Also, antimetabolite resistance can be used as the basis of selection for dhfr, which confers resistance to methotrexate (Wigler, et al., 1980, Natl. Acad. Sci. USA 77:3567; O'Hare, et al., 1981, Proc. Natl. Acad. Sci. USA 78:1527); gpt, which confers resistance to mycophenolic acid (Mulligan & Berg, 1981, Proc. Natl. Acad. Sci. USA 78:2072); neo, which confers resistance to the aminoglycoside G-418 (Colberre-Garapin, et al., 1981, J. Mol. Biol. 150:1); and hygro, which confers resistance to hygromycin (Santerre, et al., 1984, Gene 30:147) genes.

[0403] An alternative fusion protein system allows for the ready purification of non-denatured fusion proteins expressed in human cell lines (Janknecht, et al., 1991, Proc. Natl. Acad. Sci. USA 88: 8972–8976). In this system, the nucleotide sequence of interest is subcloned into a vaccinia recombination plasmid such that the nucleotide sequence's open reading frame is translationally fused to an amino-terminal tag consisting of six histidine residues. Extracts from cells infected with recombinant vaccinia virus are loaded onto Ni.sup.2+-nitriloacetic acid-agarose columns and histidine-tagged proteins are selectively eluted with imidazole-containing buffers.

[0404] Where recombinant DNA technology is used to produce the protein encoded by the novel nucleotide sequence for such assay systems, it may be advantageous to engineer fusion proteins that can facilitate labeling, immobilization and/or detection.

Therapeutic Proteins

[0405] These proteins may be administered to patients. The proteins may be administered in a physiologically acceptable medium, e.g. deionized water, phosphate buffered saline (PBS), saline, aqueous ethanol or other alcohol, plasma, proteinaceous solutions, mannitol, aqueous glucose, alcohol, vegetable oil, or the like. Other additives which may be included include buffers, where the media are generally buffered at a pH in the range of about 5 to 10, where the buffer will generally range in concentration from about 50 to 250 mM, salt, where the concentration of salt will generally range from about 5 to 500 mM, physiologically acceptable stabilizers, and the like. The compositions may be lyophilized for convenient storage and transport.

[0406] The proteins will for the most part be administered orally, parenterally, such as intravascularly (IV), intraarterially (IA), intramuscularly (IM), subcutaneously (SC), or the like. Administration may in appropriate situations be by transfusion. In some instances, where reaction of the functional group is relatively slow, administration may be oral, nasal, rectal, transdermal or aerosol, where the nature of the conjugate allows for transfer to the vascular system. Usually a single injection will be employed although more than one injection may be used, if desired. The proteins may be administered by any convenient means, including syringe, trocar, catheter, or the like. The particular manner of administration will vary depending upon the amount to be administered, whether a single bolus or continuous administration, or the like. Preferably, the administration will be intravascularly, where the site of introduction is not critical to this invention, preferably at a site where there is rapid blood flow, e.g., intravenously, peripheral or central vein. Other routes may find use where the administration is coupled with slow release techniques or a protective matrix. The intent is that the proteins be effectively distributed in the blood, so as to be able to react with the blood components. The concentration of the conjugate will vary widely, generally ranging from about 1 pg/ml to 50 mg/ml. The total administered intravascularly will generally be in the range of about 0.1 mg/ml to about 10 mg/ml, more usually about 1 mg/ml to about 5 mg/ml.

Antibodies

[0407] Indirect labeling involves the use of a protein, such as a labeled antibody, which specifically binds to the protein encoded by the novel nucleotide sequence. Such antibodies include but are not limited to polyclonal, monoclonal, chimeric, single chain, Fab fragments and fragments produced by an Fab expression library.

[0408] The invention also provides for antibodies to the protein encoded by the novel nucleotide sequences. Described herein are methods for the production of antibodies capable of specifically recognizing one or more novel nucleotide sequence epitopes. Such antibodies may include, but are not limited to polyclonal antibodies, monoclonal antibodies (mAbs), humanized or chimeric antibodies, single chain antibodies, Fab fragments, F(ab′)2 fragments, fragments produced by a Fab expression library, anti-idiotypic (anti-Id) antibodies, and epitope-binding fragments of any of the above. Such antibodies may be used, for example, in the detection of a novel nucleotide sequence in a biological sample, or, alternatively, as a method for the inhibition of abnormal gene activity, for example, the inhibition of a disease target nucleotide sequence, as further described below. Thus, such antibodies may be utilized as part of cardiovascular or other disease treatment method, and/or may be used as part of diagnostic techniques whereby patients may be tested for abnormal levels of novel nucleotide sequence encoded proteins, or for the presence of abnormal forms of the such proteins.

[0409] For the production of antibodies to a novel nucleotide sequence, various host animals may be immunized by injection with a novel protein encoded by the novel nucleotide sequence, or a portion thereof. Such host animals may include but are not limited to rabbits, mice, and rats, to name but a few. Various adjuvants may be used to increase the immunological response, depending on the host species, including but not limited to Freund's (complete and incomplete), mineral gels such as aluminum hydroxide, surface active substances such as lysolecithin, pluronic polyols, polyanions, peptides, oil emulsions, keyhole limpet hemocyanin, dinitrophenol, and potentially useful human adjuvants such as BCG (bacille Calmette-Guerin) and Corynebacterium parvum.

[0410] Polyclonal antibodies are heterogeneous populations of antibody molecules derived from the sera of animals immunized with an antigen, such as novel gene product, or an antigenic functional derivative thereof. For the production of polyclonal antibodies, host animals such as those described above, may be immunized by injection with novel gene product supplemented with adjuvants as also described above.

[0411] Monoclonal antibodies, which are homogeneous populations of antibodies to a particular antigen, may be obtained by any technique which provides for the production of antibody molecules by continuous cell lines in culture. These include, but are not limited to the hybridoma technique of Kohler and Milstein, (1975, Nature 256:495–497; and U.S. Pat. No. 4,376,110), the human B-cell hybridoma technique (Kosbor et al., 1983, Immunology Today 4:72; Cole et al., 1983, Proc. Natl. Acad. Sci. USA 80:2026–2030), and the EBV-hybridoma technique (Cole et al., 1985, Monoclonal Antibodies And Cancer Therapy, Alan R. Liss, Inc., pp. 77–96). Such antibodies may be of any immunoglobulin class including IgG, IgM, IgE, IgA, IgD and any subclass thereof. The hybridoma producing the mAb of this invention may be cultivated in vitro or in vivo. In addition, techniques developed for the production of “chimeric antibodies” (Morrison et al., 1984, Proc. Natl. Acad. Sci., 81:6851–6855; Neuberger et al., 1984, Nature, 312:604–608; Takeda et al., 1985, Nature, 314:452–454) by splicing the genes from a mouse antibody molecule of appropriate antigen specificity together with genes from a human antibody molecule of appropriate biological activity can be used. A chimeric antibody is a molecule in which different portions are derived from different animal species, such as those having a variable region derived from a murine mAb and a human immunoglobulin constant region.

[0412] Alternatively, techniques described for the production of single chain antibodies (U.S. Pat. No. 4,946,778; Bird, 1988, Science 242:423–426; Huston et al., 1988, Proc. Natl. Acad. Sci. USA 85:5879–5883; and Ward et al., 1989, Nature 334:544–546) can be adapted to produce novel nucleotide sequence-single chain antibodies. Single chain antibodies are formed by linking the heavy and light chain fragments of the Fv region via an amino acid bridge, resulting in a single chain polypeptide.

[0413] Antibody fragments which recognize specific epitopes may be generated by known techniques For example, such fragments include but are not limited to: the F(ab′)2 fragments which can be produced by pepsin digestion of the antibody molecule and the Fab fragments which can be generated by reducing the disulfide bridges of the F(ab′)2 fragments. Alternatively, Fab expression libraries may be constructed (Huse et al., 1989, Science, 246:1275–1281) to allow rapid and easy identification of monoclonal Fab fragments with the desired specificity.

Disease Specific Target Nucleotide Sequences

[0414] The invention also provides disease specific target nucleotide sequences, and sets of disease specific target nucleotide sequences. The diagnostic nucleotide sets, subsets thereof, novel nucleotide sequences, and individual members of the diagnostic nucleotide sets identified as described above are also disease specific target nucleotide sequences.

[0415] In particular, individual nucleotide sequences that are differentially regulated or have predictive value that is strongly correlated with a disease or disease criterion are especially favorable as disease specific target nucleotide sequences. Sets of genes that are co-regulated may also be identified as disease specific target nucleotide sets. Such nucleotide sequences and/or nucleotide sequence products are targets for modulation by a variety of agents and techniques. For example, disease specific target nucleotide sequences (or the products of such nucleotide sequences, or sets of disease specific target nucleotide sequences) can be inhibited or activated by, e.g., target specific monoclonal antibodies or small molecule inhibitors, or delivery of the nucleotide sequence or gene product of the nucleotide sequence to patients. Also, sets of genes can be inhibited or activated by a variety of agents and techniques. The specific usefulness of the target nucleotide sequence(s) depends on the subject groups from which they were discovered, and the disease or disease criterion with which they correlate.

Imaging

[0416] The invention also provides for imaging reagents. The differentially expressed leukocyte nucleotide sequences, diagnostic nucleotide sets, or portions thereof, and novel nucleotide sequences of the invention are nucleotide sequences expressed in cells with or without disease. Leukocytes expressing a nucleotide sequence(s) that is differentially expressed in a disease condition may localize within the body to sites that are of interest for imaging purposes. For example, a leukocyte expressing a nucleotide sequence(s) that are differentially expressed in an individual having atherosclerosis may localize or accumulate at the site of an atherosclerotic plaque. Such leukocytes, when labeled, may provide a detection reagent for use in imaging regions of the body where labeled leukocyte accumulate or localize, for example, at the atherosclerotic plaque in the case of atherosclerosis. For example, leukocytes are collected from a subject, labeled in vitro, and reintroduced into a subject. Alternatively, the labeled reagent is introduced into the subject individual, and leukocyte labeling occurs within the patient.

[0417] Imaging agents that detect the imaging targets of the invention are produced by well-known molecular and immunological methods (for exemplary protocols, see, e.g., Ausubel, Berger, and Sambrook, as well as Harlow and Lane, supra).

[0418] For example, a full-length nucleic acid sequence, or alternatively, a gene fragment encoding an immunogenic peptide or polypeptide fragments, is cloned into a convenient expression vector, for example, a vector including an in-frame epitope or substrate binding tag to facilitate subsequent purification. Protein is then expressed from the cloned cDNA sequence and used to generate antibodies, or other specific binding molecules, to one or more antigens of the imaging target protein. Alternatively, a natural or synthetic polypeptide (or peptide) or small molecule that specifically binds (or is specifically bound to) the expressed imaging target can be identified through well established techniques (see, e.g., Mendel et al. (2000) Anticancer Drug Des 15:29–41; Wilson (2000) Curr Med Chem 7:73–98; Hamby and Showwalter (1999) Pharmacol Ther 82:169–93; and Shimazawa et al. (1998) Curr Opin Struct Biol 8:451–8). The binding molecule, e.g., antibody, small molecule ligand, etc., is labeled with a contrast agent or other detectable label, e.g., gadolinium, iodine, or a gamma-emitting source. For in-vivo imaging of a disease process that involved leukocytes, the labeled antibody is infused into a subject, e.g., a human patient or animal subject, and a sufficient period of time is passed to permit binding of the antibody to target cells. The subject is then imaged with appropriate technology such as MRI (when the label is gadolinium) or with a gamma counter (when the label is a gamma emitter).

Identification of Nucleotide Sequence Involved in Leukocyte Adhesion

[0419] The invention also encompasses a method of identifying nucleotide sequences involved in leukocyte adhesion. The interaction between the endothelial cell and leukocyte is a fundamental mechanism of all inflammatory disorders, including the diseases listed in Table 1. For example, the first visible abnormality in atherosclerosis is the adhesion to the endothelium and diapedesis of mononuclear cells (e.g., T-cell and monocyte). Insults to the endothelium (for example, cytokines, tobacco, diabetes, hypertension and many more) lead to endothelial cell activation. The endothelium then expresses adhesion molecules, which have counter receptors on mononuclear cells. Once the leukocyte receptors have bound the endothelial adhesion molecules, they stick to the endothelium, roll a short distance, stop and transmigrate across the endothelium. A similar set of events occurs in both acute and chronic inflammation. When the leukocyte binds the endothelial adhesion molecule, or to soluble cytokines secreted by endothelial or other cells, a program of gene expression is activated in the leukocyte. This program of expression leads to leukocyte rolling, firm adhesion and transmigration into the vessel wall or tissue parenchyma. Inhibition of this process is highly desirable goal in anti-inflammatory drug development. In addition, leukocyte nucleotide sequences and epithelial cell nucleotide sequences, that are differentially expressed during this process may be disease-specific target nucleotide sequences.

[0420] Human endothelial cells, e.g. derived from human coronary arteries, human aorta, human pulmonary artery, human umbilical vein or microvascular endothelial cells, are cultured as a confluent monolayer, using standard methods. Some of the endothelial cells are then exposed to cytokines or another activating stimuli such as oxidized LDL, hyperglycemia, shear stress, or hypoxia (Moser et al. 1992). Some endothelial cells are not exposed to such stimuli and serve as controls. For example, the endothelial cell monolayer is incubated with culture medium containing 5 U/ml of human recombinant IL-1alpha or 10 ng/ml TNF (tumor necrosis factor), for a period of minutes to overnight. The culture medium composition is changed or the flask is sealed to induce hypoxia. In addition, tissue culture plate is rotated to induce sheer stress.

[0421] Human T-cells and/or monocytes are cultured in tissue culture flasks or plates, with LGM-3 media from Clonetics. Cells are incubated at 37 degree C., 5% CO2 and 95% humidity. These leukocytes are exposed to the activated or control endothelial layer by adding a suspension of leukocytes on to the endothelial cell monolayer. The endothelial cell monolayer is cultured on a tissue culture treated plate/flask or on a microporous membrane. After a variable duration of exposures, the endothelial cells and leukocytes are harvested separately by treating all cells with trypsin and then sorting the endothelial cells from the leukocytes by magnetic affinity reagents to an endothelial cell specific marker such as PECAM-1 (Stem Cell Technologies). RNA is extracted from the isolated cells by standard techniques. Leukocyte RNA is labeled as described above, and hybridized to leukocyte candidate nucleotide library. Epithelial cell RNA is also labeled and hybridized to the leukocyte candidate nucleotide library. Alternatively, the epithelial cell RNA is hybridized to a epithelial cell candidate nucleotide library, prepared according to the methods described for leukocyte candidate libraries, above.

[0422] Hybridization to candidate nucleotide libraries will reveal nucleotide sequences that are up-regulated or down-regulated in leukocyte and/or epithelial cells undergoing adhesion. The differentially regulated nucleotide sequences are further characterized, e.g. by isolating and sequencing the full-length sequence, analysis of the DNA and predicted protein sequence, and functional characterization of the protein product of the nucleotide sequence, as described above. Further characterization may result in the identification of leukocyte adhesion specific target nucleotide sequences, which may be candidate targets for regulation of the inflammatory process. Small molecule or antibody inhibitors can be developed to inhibit the target nucleotide sequence function. Such inhibitors are tested for their ability to inhibit leukocyte adhesion in the in vitro test described above.

Integrated Systems

[0423] Integrated systems for the collection and analysis of expression profiles, and molecular signatures, as well as for the compilation, storage and access of the databases of the invention, typically include a digital computer with software including an instruction set for sequence searching and analysis, and, optionally, high-throughput liquid control software, image analysis software, data interpretation software, a robotic control armature for transferring solutions from a source to a destination (such as a detection device) operably linked to the digital computer, an input device (e.g., a computer keyboard) for entering subject data to the digital computer, or to control analysis operations or high throughput sample transfer by the robotic control armature. Optionally, the integrated system further comprises an image scanner for digitizing label signals from labeled assay components, e.g., labeled nucleic acid hybridized to a candidate library microarray. The image scanner can interface with image analysis software to provide a measurement of the presence or intensity of the hybridized label, i.e., indicative of an on/off expression pattern or an increase or decrease in expression.

[0424] Readily available computational hardware resources using standard operating systems are fully adequate, e.g., a PC (Intel x86 or Pentium chip-compatible DOS,™ OS2,™ WINDOWS,™ WINDOWS NT,™ WINDOWS95,™ WINDOWS98,™ LINUX, or even Macintosh, Sun or PCs will suffice) for use in the integrated systems of the invention. Current art in software technology is similarly adequate (i.e., there are a multitude of mature programming languages and source code suppliers) for design, e.g., of an upgradeable open-architecture object-oriented heuristic algorithm, or instruction set for expression analysis, as described herein. For example, software for aligning or otherwise manipulating molecular signatures can be constructed by one of skill using a standard programming language such as Visual basic, Fortran, Basic, Java, or the like, according to the methods herein.

[0425] Various methods and algorithms, including genetic algorithms and neural networks, can be used to perform the data collection, correlation, and storage functions, as well as other desirable functions, as described herein. In addition, digital or analog systems such as digital or analog computer systems can control a variety of other functions such as the display and/or control of input and output files.

[0426] For example, standard desktop applications such as word processing software (e.g., Corel WordPerfect™ or Microsoft Word™) and database software (e.g., spreadsheet software such as Corel Quattro Pro™, Microsoft Excel™, or database programs such as Microsoft Access™ or Paradox™) can be adapted to the present invention by inputting one or more character string corresponding, e.g., to an expression pattern or profile, subject medical or historical data, molecular signature, or the like, into the software which is loaded into the memory of a digital system, and carrying out the operations indicated in an instruction set, e.g., as exemplified in FIG. 2. For example, systems can include the foregoing software having the appropriate character string information, e.g., used in conjunction with a user interface in conjunction with a standard operating system such as a Windows, Macintosh or LINUX system. For example, an instruction set for manipulating strings of characters, either by programming the required operations into the applications or with the required operations performed manually by a user (or both). For example, specialized sequence alignment programs such as PILEUP or BLAST can also be incorporated into the systems of the invention, e.g., for alignment of nucleic acids or proteins (or corresponding character strings).

[0427] Software for performing the statistical methods required for the invention, e.g., to determine correlations between expression profiles and subsets of members of the diagnostic nucleotide libraries, such as programmed embodiments of the statistical methods described above, are also included in the computer systems of the invention. Alternatively, programming elements for performing such methods as principle component analysis (PCA) or least squares analysis can also be included in the digital system to identify relationships between data. Exemplary software for such methods is provided by Partek, Inc., St. Peter, Mo.; at the web site partek.com.

[0428] Any controller or computer optionally includes a monitor which can include, e.g., a flat panel display (e.g., active matrix liquid crystal display, liquid crystal display), a cathode ray tube (“CRT”) display, or another display system which serves as a user interface, e.g., to output predictive data. Computer circuitry, including numerous integrated circuit chips, such as a microprocessor, memory, interface circuits, and the like, is often placed in a casing or box which optionally also includes a hard disk drive, a floppy disk drive, a high capacity removable drive such as a writeable CD-ROM, and other common peripheral elements.

[0429] Inputting devices such as a keyboard, mouse, or touch sensitive screen, optionally provide for input from a user and for user selection, e.g., of sequences or data sets to be compared or otherwise manipulated in the relevant computer system. The computer typically includes appropriate software for receiving user instructions, either in the form of user input into a set parameter or data fields (e.g., to input relevant subject data), or in the form of preprogrammed instructions, e.g., preprogrammed for a variety of different specific operations. The software then converts these instructions to appropriate language for instructing the system to carry out any desired operation.

[0430] The integrated system may also be embodied within the circuitry of an application specific integrated circuit (ASIC) or programmable logic device (PLD). In such a case, the invention is embodied in a computer readable descriptor language that can be used to create an ASIC or PLD. The integrated system can also be embodied within the circuitry or logic processors of a variety of other digital apparatus, such as PDAs, laptop computer systems, displays, image editing equipment, etc.

[0431] The digital system can comprise a learning component where expression profiles, and relevant subject data are compiled and monitored in conjunction with physical assays, and where correlations, e.g., molecular signatures with predictive value for a disease, are established or refined. Successful and unsuccessful combinations are optionally documented in a database to provide justification/preferences for user-base or digital system based selection of diagnostic nucleotide sets with high predictive accuracy for a specified disease or condition.

[0432] The integrated systems can also include an automated workstation. For example, such a workstation can prepare and analyze leukocyte RNA samples by performing a sequence of events including: preparing RNA from a human blood sample; labeling the RNA with an isotopic or non-isotopic label; hybridizing the labeled RNA to at least one array comprising all or part of the candidate library; and detecting the hybridization pattern. The hybridization pattern is digitized and recorded in the appropriate database.

Automated RNA Preparation Tool

[0433] The invention also includes an automated RNA preparation tool for the preparation of mononuclear cells from whole blood samples, and preparation of RNA from the mononuclear cells. In a preferred embodiment, the use of the RNA preparation tool is fully automated, so that the cell separation and RNA isolation would require no human manipulations. Full automation is advantageous because it minimizes delay, and standardizes sample preparation across different laboratories. This standardization increases the reproducibility of the results.

[0434] FIG. 2 depicts the processes performed by the RNA preparation tool of the invention. A primary component of the device is a centrifuge (A). Tubes of whole blood containing a density gradient solution, transcription/translation inhibitors, and a gel barrier that separates erythrocytes from mononuclear cells and serum after centrifugation are placed in the centrifuge (B). The barrier is permeable to erythrocytes and granulocytes during centrifugation, but does not allow mononuclear cells to pass through (or the barrier substance has a density such that mononuclear cells remain above the level of the barrier during the centrifugation). After centrifugation, the erythrocytes and granulocytes are trapped beneath the barrier, facilitating isolation of the mononuclear cell and serum layers. A mechanical arm removes the tube and inverts it to mix the mononuclear cell layer and the serum (C). The arm next pours the supernatant into a fresh tube (D), while the erythrocytes and granulocytes remained below the barrier. Alternatively, a needle is used to aspirate the supernatant and transfer it to a fresh tube. The mechanical arms of the device opens and closes lids, dispenses PBS to aid in the collection of the mononuclear cells by centrifugation, and moves the tubes in and out of the centrifuge. Following centrifugation, the supernatant is poured off or removed by a vacuum device (E), leaving an isolated mononuclear cell pellet. Purification of the RNA from the cells is performed automatically, with lysis buffer and other purification solutions (F) automatically dispensed and removed before and after centrifugation steps. The result is a purified RNA solution. In another embodiment, RNA isolation is performed using a column or filter method. In yet another embodiment, the invention includes an on-board homogenizer for use in cell lysis.

Other Automated Systems

[0435] Automated and/or semi-automated methods for solid and liquid phase high-throughput sample preparation and evaluation are available, and supported by commercially available devices. For example, robotic devices for preparation of nucleic acids from bacterial colonies, e.g., to facilitate production and characterization of the candidate library include, for example, an automated colony picker (e.g., the Q-bot, Genetix, U.K.) capable of identifying, sampling, and inoculating up to 10,000/4 hrs different clones into 96 well microtiter dishes. Alternatively, or in addition, robotic systems for liquid handling are available from a variety of sources, e.g., automated workstations like the automated synthesis apparatus developed by Takeda Chemical Industries, LTD. (Osaka, Japan) and many robotic systems utilizing robotic arms (Zymate II, Zymark Corporation, Hopkinton, Mass.; Orca, Beckman Coulter, Inc. (Fullerton, Calif.)) which mimic the manual operations performed by a scientist. Any of the above devices are suitable for use with the present invention, e.g., for high-throughput analysis of library components or subject leukocyte samples. The nature and implementation of modifications to these devices (if any) so that they can operate as discussed herein will be apparent to persons skilled in the relevant art.

[0436] High throughput screening systems that automate entire procedures, e.g., sample and reagent pipetting, liquid dispensing, timed incubations, and final readings of the microplate in detector(s) appropriate for the relevant assay are commercially available. (see, e.g., Zymark Corp., Hopkinton, Mass.; Air Technical Industries, Mentor, Ohio; Beckman Instruments, Inc. Fullerton, Calif.; Precision Systems, Inc., Natick, Mass., etc.). These configurable systems provide high throughput and rapid start up as well as a high degree of flexibility and customization. Similarly, arrays and array readers are available, e.g., from Affymetrix, PE Biosystems, and others.

[0437] The manufacturers of such systems provide detailed protocols the various high throughput. Thus, for example, Zymark Corp. provides technical bulletins describing screening systems for detecting the modulation of gene transcription, ligand binding, and the like.

[0438] A variety of commercially available peripheral equipment, including, e.g., optical and fluorescent detectors, optical and fluorescent microscopes, plate readers, CCD arrays, phosphorimagers, scintillation counters, phototubes, photodiodes, and the like, and software is available for digitizing, storing and analyzing a digitized video or digitized optical or other assay results, e.g., using PC (Intel x86 or pentium chip—compatible DOS™, OS2™ WINDOWS™, WINDOWS NT™ or WINDOWS95™ based machines), MACINTOSH™, or UNIX based (e.g., SUN™ work station) computers.

Embodiment in a Web Site.

[0439] The methods described above can be implemented in a localized or distributed computing environment. For example, if a localized computing environment is used, an array comprising a candidate nucleotide library, or diagnostic nucleotide set, is configured in proximity to a detector, which is, in turn, linked to a computational device equipped with user input and output features.

[0440] In a distributed environment, the methods can be implemented on a single computer with multiple processors or, alternatively, on multiple computers. The computers can be linked, e.g. through a shared bus, but more commonly, the computer(s) are nodes on a network. The network can be generalized or dedicated, at a local level or distributed over a wide geographic area. In certain embodiments, the computers are components of an intra-net or an internet.

[0441] The predictive data corresponding to subject molecular signatures (e.g., expression profiles, and related diagnostic, prognostic, or monitoring results) can be shared by a variety of parties. In particular, such information can be utilized by the subject, the subject's health care practitioner or provider, a company or other institution, or a scientist. An individual subject's data, a subset of the database or the entire database recorded in a computer readable medium can be accessed directly by a user by any method of communication, including, but not limited to, the internet. With appropriate computational devices, integrated systems, communications networks, users at remote locations, as well as users located in proximity to, e.g., at the same physical facility, the database can access the recorded information. Optionally, access to the database can be controlled using unique alphanumeric passwords that provide access to a subset of the data. Such provisions can be used, e.g., to ensure privacy, anonymity, etc.

[0442] Typically, a client (e.g., a patient, practitioner, provider, scientist, or the like) executes a Web browser and is linked to a server computer executing a Web server. The Web browser is, for example, a program such as IBM's Web Explorer, Internet explorer, NetScape or Mosaic, or the like. The Web server is typically, but not necessarily, a program such as IBM's HTTP Daemon or other WWW daemon (e.g., LINUX-based forms of the program). The client computer is bi-directionally coupled with the server computer over a line or via a wireless system. In turn, the server computer is bi-directionally coupled with a website (server hosting the website) providing access to software implementing the methods of this invention.

[0443] A user of a client connected to the Intranet or Internet may cause the client to request resources that are part of the web site(s) hosting the application(s) providing an implementation of the methods described herein. Server program(s) then process the request to return the specified resources (assuming they are currently available). A standard naming convention has been adopted, known as a Uniform Resource Locator (“URL”). This convention encompasses several types of location names, presently including subclasses such as Hypertext Transport Protocol (“http”), File Transport Protocol (“ftp”), gopher, and Wide Area Information Service (“WAIS”). When a resource is downloaded, it may include the URLs of additional resources. Thus, the user of the client can easily learn of the existence of new resources that he or she had not specifically requested.

[0444] Methods of implementing Intranet and/or Intranet embodiments of computational and/or data access processes are well known to those of skill in the art and are documented, e.g., in ACM Press, pp. 383–392; ISO-ANSI, Working Draft, “Information Technology-Database Language SQL”, Jim Melton, Editor, International Organization for Standardization and American National Standards Institute, July 1992; ISO Working Draft, “Database Language SQL-Part 2:Foundation (SQL/Foundation)”, CD9075-2:199.chi.SQL, Sep. 11, 1997; and Cluer et al. (1992) A General Framework for the Optimization of Object-Oriented Queries, Proc SIGMOD International Conference on Management of Data, San Diego, Calif., Jun. 2–5, 1992, SIGMOD Record, vol. 21, Issue 2, June, 1992; Stonebraker, M., Editor. Other resources are available, e.g., from Microsoft, IBM, Sun and other software development companies.

[0445] Using the tools described above, users of the reagents, methods and database as discovery or diagnostic tools can query a centrally located database with expression and subject data. Each submission of data adds to the sum of expression and subject information in the database. As data is added, a new correlation statistical analysis is automatically run that incorporates the added clinical and expression data. Accordingly, the predictive accuracy and the types of correlations of the recorded molecular signatures increases as the database grows.

[0446] For example, subjects, such as patients, can access the results of the expression analysis of their leukocyte samples and any accrued knowledge regarding the likelihood of the patient's belonging to any specified diagnostic (or prognostic, or monitoring, or risk group), i.e., their expression profiles, and/or molecular signatures. Optionally, subjects can add to the predictive accuracy of the database by providing additional information to the database regarding diagnoses, test results, clinical or other related events that have occurred since the time of the expression profiling. Such information can be provided to the database via any form of communication, including, but not limited to, the internet. Such data can be used to continually define (and redefine) diagnostic groups. For example, if 1000 patients submit data regarding the occurrence of myocardial infarction over the 5 years since their expression profiling, and 300 of these patients report that they have experienced a myocardial infarction and 700 report that they have not, then the 300 patients define a new “group A.” As the algorithm is used to continually query and revise the database, a new diagnostic nucleotide set that differentiates groups A and B (i.e., with and without myocardial infarction within a five year period) is identified. This newly defined nucleotide set is then be used (in the manner described above) as a test that predicts the occurrence of myocardial infarction over a five-year period. While submission directly by the patient is exemplified above, any individual with access and authority to submit the relevant data e.g., the patient's physician, a laboratory technician, a health care or study administrator, or the like, can do so.

[0447] As will be apparent from the above examples, transmission of information via the internet (or via an intranet) is optionally bi-directional. That is, for example, data regarding expression profiles, subject data, and the like are transmitted via a communication system to the database, while information regarding molecular signatures, predictive analysis, and the like, are transmitted from the database to the user. For example, using appropriate configurations of an integrated system including a microarray comprising a diagnostic nucleotide set, a detector linked to a computational device can directly transmit (locally or from a remote workstation at great distance, e.g., hundreds or thousands of miles distant from the database) expression profiles and a corresponding individual identifier to a central database for analysis according to the methods of the invention. According to, e.g., the algorithms described above, the individual identifier is assigned to one or more diagnostic (or prognostic, or monitoring, etc.) categories. The results of this classification are then relayed back, via, e.g., the same mode of communication, to a recipient at the same or different internet (or intranet) address.

Kits

[0448] The present invention is optionally provided to a user as a kit. Typically, a kit contains one or more diagnostic nucleotide sets of the invention. Alternatively, the kit contains the candidate nucleotide library of the invention. Most often, the kit contains a diagnostic nucleotide probe set, or other subset of a candidate library, e.g., as a cDNA or antibody microarray packaged in a suitable container. The kit may further comprise, one or more additional reagents, e.g., substrates, labels, primers, for labeling expression products, tubes and/or other accessories, reagents for collecting blood samples, buffers, e.g., erythrocyte lysis buffer, leukocyte lysis buffer, hybridization chambers, cover slips, etc., as well as a software package, e.g., including the statistical methods of the invention, e.g., as described above, and a password and/or account number for accessing the compiled database. The kit optionally further comprises an instruction set or user manual detailing preferred methods of using the diagnostic nucleotide sets in the methods of the invention. In one embodiment, the kit may include contents useful for the discovery of diagnostic nucleotide sets using microarrays. The kit may include sterile, endotoxin and RNAse free blood collection tubes. The kit may also include alcohol swabs, tourniquet, blood collection set, and/or PBS (phosphate buffer saline; needed when method of example 8 is used to derived mononuclear RNA). The kit may also include cell lysis buffer. The kit may include RNA isolation kit, substrates for labeling of RNA (may vary for various expression profiling techniques). The kit may also include materials for fluorescence microarray expression profiling, including one or more of the following: reverse transcriptase and 10×RT buffer, T7(dT)24 primer (primer with T7 promoter at 5′ end), DTT, deoxynucleotides, optionally 100 mM each, RNAse inhibitor, second strand cDNA buffer, DNA polymerase, Rnase H, T7 RNA polymerase ribonucleotides, in vitro transcription buffer, and/or Cy3 and Cy5 labeled ribonucleotides. The kit may also include microarrays containing candidate gene libraries, cover slips for slides, and/or hybridization chambers. The kit may further include software package for identification of diagnostic gene set from data, that contains statistical methods, and/or allows alteration in desired sensitivity and specificity of gene set. The software may further facilitate access to and data analysis by centrally a located database server. The software may further include a password and account number to access central database server. In addition, the kit may include a kit user manual.

[0449] In another embodiment, the kit may include contents useful for the application of diagnostic nucleotide sets using microarrays. The kit may include sterile, endotoxin and/or RNAse free blood collection tubes. The kit may also include, alcohol swabs, tourniquet, and/or a blood collection set. The kit may further include PBS (phosphate buffer saline; needed when method of example 7 is used to derived mononuclear RNA), cell lysis buffer, and/or an RNA isolation kit. In addition, the kit may include substrates for labeling of RNA (may vary for various expression profiling techniques). For fluorescence microarray expression profiling, components may include reverse transcriptase and 10×RT buffer, T7(dT)24 primer (primer with T7 promoter at 5′ end), DTT, deoxynucleotides (optionally 100 mM each), RNAse inhibitor, second strand cDNA buffer, DNA polymerase, Rnase H, T7 RNA polymerase, ribonucleotides, in vitro transcription buffer, and/or Cy3 and Cy5 labeled ribonucleotides. The kit may further include microarrays containing candidate gene libraries. The kit may also include cover slips for slides, and/or hybridization chambers. The kit may include a software package for identification of diagnostic gene set from data. The software package may contain statistical methods, allow alteration in desired sensitivity and specificity of gene set, and/or facilitate access to and data analysis by centrally located database server. The software package may include a password and account number to access central database server. In addition, the kit may include a kit user manual.

[0450] In another embodiment, the kit may include contents useful for the application of diagnostic nucleotide sets using real-time PCR. This kit may include terile, endotoxin and/or RNAse free blood collection tubes. The kit may further include alcohol swabs, tourniquet, and/or a blood collection set. The kit may also include PBS (phosphate buffer saline; needed when method of example 7 is used to derived mononuclear RNA). In addition, the kit may include cell lysis buffer and/or an RNA isolation kit. The kit may also include substrates for real time RT-PCR, which may vary for various real-time PCR techniques, including poly dT primers, random hexamer primers, reverse Transcriptase and RT buffer, DTT, deoxynucleotides 100 mM, RNase H, primer pairs for diagnostic and control gene set, 10×PCR reaction buffer, and/or Taq DNA polymerase. The kit may also include fluorescent probes for diagnostic and control gene set (alternatively, fluorescent dye that binds to only double stranded DNA). The kit may further include reaction tubes with or without barcode for sample tracking, 96-well plates with barcode for sample identification, one barcode for entire set, or individual barcode per reaction tube in plate. The kit may also include a software package for identification of diagnostic gene set from data, and/or statistical methods. The software package may allow alteration in desired sensitivity and specificity of gene set, and/or facilitate access to and data analysis by centrally located database server. The kit may include a password and account number to access central database server. Finally, the kit may include a kit user manual.

[0451] This invention will be better understood by reference to the following non-limiting Examples:

List of Example Titles

[0452] Example 1: Generation of subtracted leukocyte candidate nucleotide library

[0453] Example 2: Identification of nucleotide sequences for candidate library using data mining techniques

[0454] Example 3: DNA Sequencing and Processing of raw sequence data.

[0455] Example 4: Further sequence analysis of novel nucleotide sequences identified by subtractive hybridization screening

[0456] Example 5: Further sequence analysis of novel Clone 596H6

[0457] Example 6: Further sequence analysis of novel Clone 486E11

[0458] Example 7: Preparation of a leukocyte cDNA array comprising a candidate gene library

[0459] Example 8: Preparation of RNA from mononuclear cells for expression profiling

[0460] Example 9: Preparation of Universal Control RNA for use in leukocyte expression profiling

[0461] Example 10. RNA Labeling and hybridization to a leukocyte cDNA array of candidate nucleotide sequences.

[0462] Example 11: Clinical study for the Identification of diagnostic gene sets useful in diagnosis and treatment of Cardiac allograft rejection

[0463] Example 12: Identification of diagnostic nucleotide sets for kidney and liver allograft rejection

[0464] Example 13: Identification of diagnostic nucleotide sequences sets for use in the diagnosis and treatment of Atherosclerosis, Stable Angina Pectoris, and acute coronary syndrome.

[0465] Example 14: Identification of diagnostic nucleotide sets for use in diagnosing and treating Restenosis

[0466] Example 15: Identification of diagnostic nucleotide sets for use in monitoring treatment and/or progression of Congestive Heart Failure

[0467] Example 16: Identification of diagnostic nucleotide sets for use in diagnosis of rheumatoid arthritis.

[0468] Example 17: Identification of diagnostic nucleotide sets for diagnosis of cytomegalovirus

[0469] Example 18: Identification of diagnostic nucleotide sets for diagnosis of Epstein Barr Virus

[0470] Example 19: Identification of diagnostic nucleotides sets for monitoring response to statin drugs.

[0471] Example 20: Probe selection for a 24,000 feature Array.

[0472] Example 21: Design of oligonucleotide probes.

[0473] Example 22: Production of an array of 8,000 spotted 50 mer oligonucleotides.

[0474] Example 23: Amplification, labeling and hybridization of total RNA to an oligonucleotide microarray.

[0475] Example 24: Analysis of Human Transplant Patient Mononuclear cell RNA Hybridized to a 24,000 Feature Microarray.

[0476] Example 25: Real-time PCR validation of array expression results

[0477] Example 26: Correlation and Classification Analysis

[0478] Example 27—Amplification, labeling, and hybridization of total RNA to an oligonucleotide microarray

[0479] Example 28: Real-time PCR validation of array expression results

[0480] Example 29: Real-time PCR expression markers of acute allograft rejection

[0481] Example 30: Identification of diagnostic nucleotide sets for diagnosis of Cardiac Allograft Rejection using microarrays

[0482] Example 31: Correlation and Classification Analysis

[0483] Example 32: Acute allograft rejection: biopsy tissue gene expression profiling

[0484] Example 33: Microarray and PCR gene expression panels for diagnosis and monitoring of acute allograft rejection

[0485] Example 34: Assay sample preparation

[0486] Example 35: Allograft rejection diagnostic gene sequence analysis

[0487] To produce a candidate nucleotide library with representatives from the spectrum of nucleotide sequences that are differentially expressed in leukocytes, subtracted hybridization libraries were produced from the following cell types and conditions:

[0488] 1. Buffy Coat leukocyte fractions—stimulated with ionomycin and PMA

[0489] 2. Buffy Coat leukocyte fractions—un-stimulated

[0490] 3. Peripheral blood mononuclear cells—stimulated with ionomycin and PMA

[0491] 4. Peripheral blood mononuclear cells—un-stimulated

[0492] 5. T lymphocytes—stimulated with PMA and ionomycin

[0493] 6. T lymphocytes—resting

[0494] Cells were obtained from multiple individuals to avoid introduction of bias by using only one person as a cell source.

[0495] Buffy coats (platelets and leukocytes that are isolated from whole blood) were purchased from Stanford Medical School Blood Center. Four buffy coats were used, each of which was derived from about 350 ml of whole blood from one donor individual 10 ml of buffy coat sample was drawn from the sample bag using a needle and syringe. 40 ml of Buffer EL (Qiagen) was added per 10 ml of buffy coat to lyse red blood cells. The sample was placed on ice for 15 minutes, and cells were collected by centrifugation at 2000 rpm for 10 minutes. The supernatant was decanted and the cell pellet was re-suspended in leukocyte growth media supplemented with DNase (LGM-3 from Clonetics supplemented with Dnase at a final concentration of 30 U/ml). Cell density was determined using a hemocytometer. Cells were plated in media at a density of 1×10⁶ cells/ml in a total volume of 30 ml in a T-75 flask (Corning). Half of the cells were stimulated with ionomycin and phorbol myristate acetate (PMA) at a final concentration of 1 μg/ml and 62 ng/ml, respectively. Cells were incubated at 37° C. and at 5% CO₂ for 3 hours, then cells were scraped off the flask and collected into 50 ml tubes. Stimulated and resting cell populations were kept separate. Cells were centrifuged at 2000 rpm for 10 minutes and the supernatant was removed. Cells were lysed in 6 ml of phenol/guanidine isothiocyanate (Trizol reagent, GibcoBRL), homogenized using a rotary homogenizer, and frozen at 80°. Total RNA and mRNA were isolated as described below.

[0496] Two frozen vials of 5×10⁶ human peripheral blood mononuclear cells (PBMCs) were purchased from Clonetics (catalog number cc-2702). The cells were rapidly thawed in a 37° C. water bath and transferred to a 15 ml tube containing 10 ml of leukocyte growth media supplemented with DNase (prepared as described above). Cells were centrifuged at 200 μg for 10 minutes. The supernatant was removed and the cell pellet was resuspended in LGM-3 media supplemented with DNase. Cell density was determined using a hemocytometer. Cells were plated at a density of 1×10⁶ cells/ml in a total volume of 30 ml in a T-75 flask (Corning). Half of the cells were stimulated with ionomycin and PMA at a final concentration of 1 μg/ml and 62 ng/ml, respectively. Cells were incubated at 37° C. and at 5% CO₂ for 3 hours, then cells were scraped off the flask and collected into 50 ml tubes. Stimulated and resting cell populations were kept separate. Cells were centrifuged at 2000 rpm and the supernatant was removed. Cells were lysed in 6 ml of phenol/guanidine isothiocyanate solution (TRIZOL reagent, GibcoBRL)), homogenized using a rotary homogenizer, and frozen at 80°. Total RNA and mRNA were isolated from these samples using the protocol described below.

[0497] 45 ml of whole blood was drawn from a peripheral vein of four healthy human subjects into tubes containing anticoagulant. 50 μl RosetteSep (Stem Cell Technologies) cocktail per ml of blood was added, mixed well, and incubated for 20 minutes at room temperature. The mixture was diluted with an equal volume of PBS+2% fetal bovine serum (FBS) and mixed by inversion. 30 ml of diluted mixture sample was layered on top of 15 ml DML medium (Stem Cell Technologies). The sample tube was centrifuged for 20 minutes at 1200×g at room temperature. The enriched T-lymphocyte cell layer at the plasma: medium interface was removed. Enriched cells were washed with PBS+2% FBS and centrifuged at 1200×g. The cell pellet was treated with 5 ml of erythrocyte lysis buffer (EL buffer, Qiagen) for 10 minutes on ice. The sample was centrifuged for 5 min at 1200 g. Cells were plated at a density of 1×10⁶ cells/ml in a total volume of 30 ml in a T-75 flask (Corning). Half of the cells were stimulated with ionomycin and PMA at a final concentration of 1 μg/ml and 62 ng/ml, respectively. Cells were incubated at 37° C. and at 5% CO₂ for 3 hours, then cells were scraped off the flask and collected into 50 ml tubes. Stimulated and resting cell populations were kept separate. Cells were centrifuged at 2000 rpm and the supernatant was removed. Cells were lysed in 6 ml of phenol/guanidine isothiocyanate solution (TRIZOL reagent, GibcoBRL), homogenized using a rotary homogenizer, and frozen at 80°. Total RNA and mRNA were isolated as described below.

[0498] Total RNA and mRNA were isolated using the following procedure: the homogenized samples were thawed and mixed by vortexing. Samples were lysed in a 1:0.2 mixture of Trizol and chloroform, respectively. For some samples, 6 ml of Trizol-chloroform was added. Variable amounts of Trizol-chloroform was added to other samples. Following lysis, samples were centrifuged at 3000 g for 15 min at 4° C. The aqueous layer was removed into a clean tube and 4 volumes of Buffer RLT Qiagen) was added for every volume of aqueous layer. The samples were mixed thoroughly and total RNA was prepared from the sample by following the Qiagen Rneasy midi protocol for RNA cleanup (October 1999 protocol, Qiagen). For the final step, the RNA was eluted from the column twice with 250 μl Rnase-free water. Total RNA was quantified using a spectrophotometer. Isolation of mRNA from total RNA sample was done using The Oligotex mRNA isolation protocol (Qiagen) was used to isolate mRNA from total RNA, according to the manufacturer's instructions (Qiagen, 7/99 version). mRNA was quantified by spectrophotometry.

[0499] Subtracted cDNA libraries were prepared using Clontech's PCR-Select cDNA Subtraction Kit (protocol number PT-1117-1) as described in the manufacturer's protocol. The protocol calls for two sources of RNA per library, designated “Driver” and “Tester.” The following 6 libraries were made:

[0500]

Library	Driver RNA	Tester RNA
Buffy Coat Stimulated	Un-stimulated Buffy	Stimulated Buffy Coat
	Coat
Buffy Coat Resting	Stimulated Buffy	Un-stimulated Buffy Coat
	Coat
PBMC Stimulated	Un-stimulated	Stimulated PBMCs
	PBMCs
PBMC Resting	Stimulated PBMCs	Un-stimulated PBMCs
T-cell Stimulated	Un-stimulated	Stimulated T-cells
	T-cells
T-cell Resting	Stimulated T-cells	Un-stimulated T-cells

[0501] The Clontech protocol results in the PCR amplification of cDNA products. The PCR products of the subtraction protocol were ligated to the pGEM T-easy bacterial vector as described by the vector manufacturer (Promega 6/99 version). Ligated vector was transformed into competent bacteria using well-known techniques, plated, and individual clones are picked, grown and stored as a glycerol stock at −80 C. Plasmid DNA was isolated from these bacteria by standard techniques and used for sequence analysis of the insert. Unique cDNA sequences were searched in the Unigene database (build 133), and Unigene cluster numbers were identified that corresponded to the DNA sequence of the cDNA. Unigene cluster numbers were recorded in an Excel spreadsheet.

[0502] Existing and publicly available gene sequence databases were used to identify candidate nucleotide sequences for leukocyte expression profiling. Genes and nucleotide sequences with specific expression in leukocytes, for example, lineage specific markers, or known differential expression in resting or activated leukocytes were identified. Such nucleotide sequences are used in a leukocyte candidate nucleotide library, alone or in combination with nucleotide sequences isolated through cDNA library construction, as described above.

[0503] Leukocyte candidate nucleotide sequences were identified using three primary methods. First, the publicly accessible publication database PubMed was searched to identify nucleotide sequences with known specific or differential expression in leukocytes. Nucleotide sequences were identified that have been demonstrated to have differential expression in peripheral blood leukocytes between subjects with and without particular disease(s) selected from Table 1. Additionally, genes and gene sequences that were known to be specific or selective for leukocytes or sub-populations of leukocytes were identified in this way.

[0504] Next, two publicly available databases of DNA sequences, Unigenea at the NCBI web site and BodyMap at the University of Tokyo, were searched for sequenced DNA clones that showed specificity to leukocyte lineages, or subsets of leukocytes, or resting or activated leukocytes.

[0505] The human Unigene database (build 133) was used to identify leukocyte candidate nucleotide sequences that were likely to be highly or exclusively expressed in leukocytes. We used the Library Differential Display utility of Unigene, which uses statistical methods (The Fisher Exact Test) to identify nucleotide sequences that have relative specificity for a chosen library or group of libraries relative to each other. We compared the following human libraries from Unigene release 133:

[0506] 546 NCI_CGAP_HSC1 (399)

[0507] 848 Human_mRNA_from_cd34+_stem_cells (122)

[0508] 105 CD34+DIRECTIONAL (150)

[0509] 3587 KRIBB_Human_CD4_intrathymic_T-cell_cDNA_library (134)

[0510] 3586 KRIBB_Human_DP_intrathymic_T-cell_cDNA_library (179)

[0511] 3585 KRIBB_Human_TN_intrathymic_T-cell_cDNA_library (127)

[0512] 3586 323 Activated_T-cells_I (740)

[0513] 376 Activated_T-cells_XX (1727)

[0514] 327 Monocytes,_stimulated_II (110)

[0515] 824 Proliferating_Erythroid_Cells_(LCB:ad_library) (665)

[0516] 825 429 Macrophage_II (105)

[0517] 387 Macrophage_I (137)

[0518] 669 NCI_CGAP_CLL1 (11626)

[0519] 129 Human_White_blood_cells (922)

[0520] 1400 NIH_MGC_—2 (422)

[0521] 55 Human_promyelocyte (1220)

[0522] 1010 NCI_CGAP_CML1 (2541)

[0523] 2217 NCI_CGAP_Sub7 (218)

[0524] 1395 NCI_CGAP_Sub6 (2764)

[0525] 4874 NIH_MGC_—48 (2524)

[0526] BodyMap, like Unigene, contains cell-specific libraries that contain potentially useful information about genes that may serve as lineage-specific or leukocyte specific markers (Okubo et al. 1992). We compared three leukocyte specific libraries, Granulocyte, CD4 T cell, and CD8 T cell, with the other libraries. Nucleotide sequences that were found in one or more of the leukocyte-specific libraries, but absent in the others, were identified.

[0527] Clones that were found exclusively in one of the three leukocyte libraries were also included in a list of nucleotide sequences that could serve as lineage-specific markers. Next, the sequence of the nucleotide sequences identified in PubMed or BodyMap were searched in Unigene (version 133), and a human Unigene cluster number was identified for each nucleotide sequence. The cluster number was recorded in a Microsoft Excel™ spreadsheet, and a non-redundant list of these clones was made by sorting the clones by UniGene number, and removing all redundant clones using Microsoft Excel™ tools. The non-redundant list of UniGene cluster numbers was then compared to the UniGene cluster numbers of the cDNAs identified using differential cDNA hybridization, as described above in Example 1 (listed in Table 3, Tables 8, 11–12, 14, 15, 19 and the sequence listing). Only UniGene clusters that were not contained in the cDNA libraries were retained. Unigene clusters corresponding to 1911 candidate nucleotide sequences for leukocyte expression profiling were identified in this way and are listed in Table 3 and the sequence listing.

[0528] DNA clones corresponding to each UniGene cluster number are obtained in a variety of ways. First, a cDNA clone with identical sequence to part of, or all of the identified UniGene cluster is bought from a commercial vendor or obtained from the IMAGE (Integrated Molecular Analysis of Genomes and their Expression) consortium. Alternatively, PCR primers are designed to amplify and clone any portion of the nucleotide sequence from cDNA or genomic DNA using well-known techniques. Alternatively, the sequences of the identified UniGene clusters are used to design and synthesize oligonucleotide probes for use in microarray based expression profiling.

[0529] Clones of differentially expressed cDNAs (identified by subtractive hybridization, described above) were sequenced on an MJ Research BaseStation™ slab gel based fluorescent detection system, using BigDye™ (Applied Biosystems, Foster City, Calif.) terminator chemistry was used (Heiner et al., Genome Res 1998 May;8(5):557–61). The fluorescent profiles were analyzed using the Phred sequence analysis program (Ewing et al, (1998), Genome Research 8: 175–185). Analysis of each clone results in a one pass nucleotide sequence and a quality file containing a number for each base pair with a score based on the probability that the determined base is correct. Each sequence files and its respective quality files were initially combined into single fasta format (Pearson, W R. Methods Mol. Biol. 2000;132:185–219), multi-sequence file with the appropriate labels for each clone in the headers for subsequent automated analysis. Initially, known sequences were analyzed by pair wise similarity searching using the blastn option of the blastall program obtained from the National Center for Biological Information, National Library of Medicine, National Institutes of Health (NCBI) to determine the quality score that produced accurate matching (Altschul S F, et al. J Mol Biol. 1990 Oct. 5;215(3):403–10.). Empirically, it was determined that a raw score of 8 was the minimum that contained useful information. Using a sliding window average for 16 base pairs, an average score was determined. The sequence was removed (trimmed) when the average score fell below 8. Maximum reads were 950 nucleotides long.

[0530] Next, the sequences were compared by similarity matching against a database file containing the flanking vector sequences used to clone the cDNA, using the blastall program with the blastn option. All regions of vector similarity were removed, or “trimmed” from the sequences of the clones using scripts in the GAWK programming language, a variation of AWK (Aho A V et al, The Awk Programming Language (Addison-Wesley, Reading Mass., 1988); Robbins, A D, “Effective AWK Programming” (Free Software Foundation, Boston Mass., 1997). It was found that the first 45 base pairs of all the sequences were related to vector; these sequences were also trimmed and thus removed from consideration. The remaining sequences were then compared against the NCBI vector database (Kitts, P. A. et al. National Center for Biological Information, National Library of Medicine, National Institutes of Health, Manuscript in preparation (2001) using blastall with the blastn option. Any vector sequences that were found were removed from the sequences.

[0531] Messenger RNA contains repetitive elements that are found in genomic DNA. These repetitive elements lead to false positive results in similarity searches of query mRNA sequences versus known mRNA and EST databases. Additionally, regions of low information content (long runs of the same nucleotide, for example) also result in false positive results. These regions were masked using the program RepeatMasker2 found at the University of Washington website. The trimmed and masked files were then subjected to further sequence analysis.

[0532] cDNA sequences were further characterized using BLAST analysis. The BLASTN program was used to compare the sequence of the fragment to the UniGene, dbEST, and nr databases at NCBI (GenBank release 123.0; see Table 5). In the BLAST algorithm, the expect value for an alignment is used as the measure of its significance. First, the cDNA sequences were compared to sequences in Unigene. If no alignments were found with an expect value less than 10⁻²⁵, the sequence was compared to the sequences in the dbEST database using BLASTN. If no alignments were found with an expect value less than 10⁻²⁵, the sequence was compared to sequences in the nr database.

[0533] The BLAST analysis produced the following categories of results: a) a significant match to a known or predicted human gene, b) a significant match to a nonhuman DNA sequence, such as vector DNA or E. coli DNA, c) a significant match to an unidentified GenBank entry (a sequence not previously identified or predicted to be an expressed sequence or a gene), such as a cDNA clone, mRNA, or cosmid, or d) no significant alignments. If a match to a known or predicted human gene was found, analysis of the known or predicted protein product was performed as described below. If a match to an unidentified GenBank entry was found, or if no significant alignments were found, the sequence was searched against all known sequences in the human genome database see Table 5).

[0534] If many unknown sequences were to be analyzed with BLASTN, the clustering algorithm CAP2 (Contig Assembly Program, version 2) was used to cluster them into longer, contiguous sequences before performing a BLAST search of the human genome. Sequences that can be grouped into contigs are likely to be cDNA from expressed genes rather than vector DNA, E. coli DNA or human chromosomal DNA from a noncoding region, any of which could have been incorporated into the library. Clustered sequences provide a longer query sequence for database comparisons with BLASTN, increasing the probability of finding a significant match to a known gene. When a significant alignment was found, further analysis of the putative gene was performed, as described below. Otherwise, the sequence of the original cDNA fragment or the CAP2 contig is used to design a probe for expression analysis and further approaches are taken to identify the gene or predicted gene that corresponds to the cDNA sequence, including similarity searches of other databases, molecular cloning, and Rapid Amplification of cDNA Ends (RACE).

[0535] In some cases, the process of analyzing many unknown sequences with BLASTN was automated by using the BLAST network-client program blastcl3, which was downloaded from the NCBI website.

[0536] When a cDNA sequence aligned to the sequence of one or more chromosomes, a large piece of the genomic region around the loci was used to predict the gene containing the cDNA. To do this, the contig corresponding to the mapped locus, as assembled by the RefSeq project at NCBI, was downloaded and cropped to include the region of alignment plus 100,000 bases preceding it and 100,000 bases following it on the chromosome. The result was a segment 200 kb in length, plus the length of the alignment. This segment, designated a putative gene, was analyzed using an exon prediction algorithm to determine whether the alignment area of the unknown sequence was contained within a region predicted to be transcribed (see Table 6).

[0537] This putative gene was characterized as follows: all of the exons comprising the putative gene and the introns between them were taken as a unit by noting the residue numbers on the 200 kb+ segment that correspond to the first base of the first exon and the last base of the last exon, as given in the data returned by the exon prediction algorithm. The truncated sequence was compared to the UniGene, dbEST, and nr databases to search for alignments missed by searching with the initial fragment.

[0538] The predicted amino acid sequence of the gene was also analyzed. The peptide sequence of the gene predicted from the exons was used in conjunction with numerous software tools for protein analysis (see Table 7). These were used to classify or identify the peptide based on similarities to known proteins, as well as to predict physical, chemical, and biological properties of the peptides, including secondary and tertiary structure, flexibility, hydrophobicity, antigenicity (hydrophilicity), common domains and motifs, and localization within the cell or tissues. The peptide sequence was compared to protein databases, including SWISS-PROT, TrEMBL, GenPept, PDB, PIR, PROSITE, ProDom, PROSITE, Blocks, PRINTS, and Pfam, using BLASTP and other algorithms to determine similarities to known proteins or protein subunits.

[0539] The sequence of clone 596H6 is provided below:

[0540]

	ACTATATTTA GGCACCACTG CCATAAACTA CCAAAAAAAA AATGTAATTC	50
	CTAGAAGCTG TGAAGAATAG TAGTGTAGCT AAGCACGGTG TGTGGACAGT	100
	GGGACATCTG CCACCTGCAG TAGGTCTCTG CACTCCCAAA AGCAAATTAC	150
	ATTGGCTTGA ACTTCAGTAT GCCCGGTTCC ACCCTCCAGA AACTTTTGTG	200
	TTCTTTGTAT AGAATTTAGG AACTTCTGAG GGCCACAAAT ACACACATTA	250
	AAAAAGGTAG AATTTTTGAA GATAAGATTC TTCTAAAAAA GCTTCCCAAT	300
	GCTTGAGTAG AAAGTATCAG TAGAGGTATC AAGGGAGGAG AGACTAGGTG	350
	ACCACTAAAC TCCTTCAGAC TCTTAAAATT ACGATTCTTT TCTCAAAGGG	400
	GAAGAACGTC AGTGCAGCGA TCCCTTCACC TTTAGCTAAA GAATTGGACT	450
	GTGCTGCTCA AAATAAAGAT CAGTTGGAGG TANGATGTCC AAGACTGAAG	500
	GTAAAGGACT AGTGCAAACT GAAAGTGATG GGGAAACAGA CCTACGTATG	550
	GAAGCCATGT AGTGTTCTTC ACAGGCTGCT GTTGACTGAA ATTCCTATCC	600
	TCAAATTACT CTAGACTGAA GCTGCTTCCC TTCAGTGAGC AGCCTCTCCT	650
	TCCAAGATTC TGGAAAGCAC ACCTGACTCC AAACAAAGAC TTAGAGCCCT	700
	GTGTCAGTGC TGCTGCTGCT TTTACCAGAT TCTCTAACCT TCCGGGTAGA	750
	AGAG (SEQ ID NO: 8767)

[0541] This sequence was used as input for a series of BLASTN searches. First, it was used to search the UniGene database, build 132 at the NCBI BLAST web site. No alignments were found with an expect value less than the threshold value of 10⁻²²⁵. A BLASTN search of the database dbEST, release 041001, was then performed on the sequence and 21 alignments were found at the NCBI BLAST web site. Ten of these had expect values less than 10⁻²⁵, but all were matches to unidentified cDNA clones. Next, the sequence was used to run a BLASTN search of the nr database, release 123.0. No significant alignment to any sequence in nr was found. Finally, a BLASTN search of the human genome was performed on the sequence located at the NCBI BLASTN web site.

[0542] A single alignment to the genome was found on contig NT_—004698.3 (e=0.0). The region of alignment on the contig was from base 1,821,298 to base 1,822,054, and this region was found to be mapped to chromosome 1, from base 105,552,694 to base 105,553,450. The sequence containing the aligned region, plus 100 kilobases on each side of the aligned region, was downloaded. Specifically, the sequence of chromosome 1 from base 105,452,694 to 105,653,450 was downloaded from the NCBI web site.

[0543] This 200,757 bp segment of the chromosome was used to predict exons and their peptide products as follows. The sequence was used as input for the Genscan algorithm found at the Massachusetts Institute of Technology (MIT), using the following Genscan settings:

[0544] Organism: vertebrate

[0545] Suboptimal exon cutoff: 1.00 (no suboptimal exons)

[0546] Print options: Predicted CDS and peptides

[0547] The region matching the sequence of clone 596H6 was known to span base numbers 100,001 to 100,757 of the input sequence. An exon was predicted by the algorithm, with a probability of 0.695, covering bases 100,601 to 101,094 (designated exon 4.14 of the fourth predicted gene). This exon was part of a predicted cistron that is 24,195 bp in length. The sequence corresponding to the cistron was noted and saved separately from the 200,757 bp segment. BLASTN searches of the Unigene, dbEST, and nr databases were performed on it.

[0548] At least 100 significant alignments to various regions of the sequence were found in the dbEST database, although most appeared to be redundant representations of a few exons. All matches were to unnamed cDNAs and mRNAs (unnamed cDNAs and mRNAs are cDNAs and mRNAs not previously identified, or shown to correspond to a known or predicted human gene) from various tissue types. Most aligned to a single region on the sequence and spanned 500 bp or less, but several consisted of five or six regions separated by gaps, suggesting the locations of exons in the gene. Several significant matches to entries in the UniGene database were found, as well, even after masking low-complexity regions and short repeats in the sequence. All matches were to unnamed cDNA clones.

[0549] At least 100 significant alignments were found in the nr database, as well. A similarity to hypothetical protein FLJ22457 (UniGene cluster Hs.238707) was found (e=0.0). The cDNA of this predicted protein has been isolated from B lymphocytes located at the NCBI web site.

[0550] Other significant alignments were to unnamed cDNAs and mRNAs.

[0551] Using Genscan, the following 730 residue peptide sequence was predicted from the putative gene:

[0552]

	MDGLGRRLRA SLRLKRGHGG HWRLNEMPYM KHEFDGGPPQ DNSGEALKEP	50
	ERAQEHSLPN FAGGQHFFEY LLVVSLKKKR SEDDYEPIIT YQFPKRENLL	100
	RGQQEEEERL LKAIPLFCFP DGNEWASLTE YPSLSCKTPG LLAALVVEKA	150
	QPRTCCHASA PSAAPQARGP DAPSPAAGQA LPAGPGPRLP KVYCIISCIG	200
	CFGLFSKILD EVEKRHQISM AVIYPFMQGL REAAFPAPGK TVTLKSFIPD	250
	SGTEFISLTR PLDSHLEHVD FSSLLHCLSF EQILQIFASA VLERKIIFLA	300
	EGLREEEKDV RDSTEVRGAG ECHGFQRKGN LGKQWGLCVE DSVKMGDNQR	350
	GTSCSTLSQC IHAAAALLYP FSWAHTYIPV VPESLLATVC CPTPFMVGVQ	400
	MRFQQEVMDS PMEEIQPQAE IKTVNPLGVY EERGPEKASL CLFQVLLVNL	450
	CEGTFLMSVG DEKDILPPKL QDDILDSLGQ GINELKTAEQ INEHVSGPFV	500
	QFFVKIVGHY ASYIKREANG QGHFQERSFC KALTSKTNRR FVKKFVKTQL	550
	FSLFIQEAEK SKNPPAEVTQ VGNSSTCVVD TWLEAAATAL SHHYNIFNTE	600
	HTLWSKGSAS LHEVCGHVRT RVKRKILFLY VSLAFTMGKS IFLVENKAMN	650
	MTIKWTTSGR PGHGDMFGVI ESWGAAALLL LTGRVRDTGK SSSSTGHRAS	700
	KSLVWSQVCF PESWEERLLT EGKQLQSRVI SEQ ID NO: 8768

[0553] Multiple analyses were performed using this prediction. First, a pairwise comparison of the sequence above and the sequence of FLJ22457, the hypothetical protein mentioned above, using BLASTP version 2.1.2 at the NCBI website, resulted in a match with an expect value of 0.0. The peptide sequence predicted from clone 596H6 was longer and 19% of the region of alignment between the two resulted from gaps in hypothetical protein FLJ22457. The cause of the discrepancy might be alternative mRNA splicing, alternative post-translational processing, or differences in the peptide-predicting algorithms used to create the two sequences, but the homology between the two is significant.

[0554] BLASTP and TBLASTN were also used to search for sequence similarities in the SWISS-PROT, TrEMBL, GenBank Translated, and PDB databases. Matches to several proteins were found, among them a tumor cell suppression protein, HTS1. No matches aligned to the full length of the peptide sequence, however, suggesting that similarity is limited to a few regions of the peptide.

[0555] TBLASTN produced matches to several proteins—both identified and theoretical—but again, no matches aligned to the full length of the peptide sequence. The best alignment was to the same hypothetical protein found in GenBank before (FLJ22457).

[0556] To discover similarities to protein families, comparisons of the domains (described above) were carried out using the Pfam and Blocks databases. A search of the Pfam database identified two regions of the peptide domains as belonging the DENN protein family (e=2.1×10−⁻³³). The human DENN protein possesses an RGD cellular adhesion motif and a leucine-zipper-like motif associated with protein dimerization, and shows partial homology to the receptor binding domain of tumor necrosis factor alpha. DENN is virtually identical to MADD, a human MAP kinase-activating death domain protein that interacts with type I tumor necrosis factor receptor found at the European Bioinformatics website. The search of the Blocks database also revealed similarities between regions of the peptide sequence and known protein groups, but none with a satisfactory degree of confidence. In the Blocks scoring system, scores over 1,100 are likely to be relevant. The highest score of any match to the predicted peptide was 1,058.

[0557] The Prosite, ProDom, PRINTS databases (all publicly available) were used to conduct further domain and motif analysis. The Prosite search generated many recognized protein domains. A BLASTP search was performed to identify areas of similarity between the protein query sequence and PRINTS, a protein database of protein fingerprints, groups of motifs that together form a characteristic signature of a protein family. In this case, no groups were found to align closely to any section of the submitted sequence. The same was true when the ProDom database was searched with BLASTP.

[0558] A prediction of protein structure was done by performing a BLAST search of the sequence against PDB, a database in which every member has tertiary structure information. No significant alignments were found by this method. Secondary and super-secondary structure was examined using the Garnier algorithm. Although it is only considered to be 60–65% accurate, the algorithm provided information on the locations and lengths of alpha-helices, beta-sheets, turns and coils.

[0559] The antigenicity of the predicted peptide was modeled by graphing hydrophilicity vs. amino acid number. This produced a visual representation of trends in hydrophilicity along the sequence. Many locations in the sequence showed antigenicity and five sites had antigenicity greater than 2. This information can be used in the design of affinity reagents to the protein.

[0560] Membrane-spanning regions were predicted by graphing hydrophobicity vs. amino acid number. Thirteen regions were found to be somewhat hydrophobic. The algorithm TMpred predicted a model with 6 strong transmembrane heliceslocated at the embnet web site.

[0561] NNPSL is a neural network algorithm developed by the Sanger Center. It uses amino acid composition and sequence to predict cellular location. For the peptide sequence submitted, its first choice was mitochondrial (51.1% expected accuracy). Its second choice was cytoplasmic (91.4% expected accuracy).

[0562] The sequence of clone 486E11 is provided below:

[0563]

	TAAAAGCAGG CTGTGCACTA GGGACCTAGT GACCTTACTA GAAAAAACTC	50
	AAATTCTCTG AGCCACAAGT CCTCATGGGC AAAATGTAGA TACCACCACC	100
	TAACCCTGCC AATTTCCTAT CATTGTGACT ATCAAATTAA ACCACAGGCA	150
	GGAAGTTGCC TTGAAAACTT TTTATAGTGT ATATTACTGT TCACATAGAT	200
	NAGCAATTAA CTTTACATAT ACCCGTTTTT AAAAGATCAG TCCTGTGATT	250
	AAAAGTCTGG CTGCCCTAAT TCACTTCGAT TATACATTAG GTTAAAGCCA	300
	TATAAAAGAG GCACTACGTC TTCGGAGAGA TGAATGGATA TTACAAGCAG	350
	TAATGTTGGC TTTGGAATAT ACACATAATG TCCACTTGAC CTCATCTATT	400
	TGACACAAAA TGTAAACTAA ATTATGAGCA TCATTAGATA CCTTGGCCTT	450
	TTCAAATCAC ACAGGGTCCT AGATCTNNNN NNNNNNNNNN NNNNNNNNNN	500
	NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNAC TTTGGGATTC	550
	CTATATCTTT GTCAGCTGTC AACTTCAGTG TTTTCAGGTT AAATTCTATC	600
	CATAGTCATC CCAATATACC TGCTTTAGAT GATACAACCT TCAAAAGATC	650
	CGCTCTTCCT CGTAAAAAGT GGAG SEQ ID NO: 8769

[0564] The BLASTN program was used to compare the sequence to the UniGene and dbEST databases. No significant alignments were found in either. It was then searched against the nr database and only alignments to unnamed genomic DNA clones were found.

[0565] CAP2 was used to cluster a group of unknowns, including clone 486E11. The sequence for 486E11 was found to overlap others. These formed a contig of 1,010 residues, which is shown below:

[0566]

CGGACAGGTA CCTAAAAGCA GGCTGTGCAC TAGGGACCTA GTGACCTTAC	50
TAGAAAAAAC TCAAATTCTC TGAGCCACAA GTCCTCATGG GCAAAATGTA	100
GATACCACCA CCTAACCCTG CCAATTTCCT ATCATTGTGA CTATCAAATT	150
AAACCACAGG CAGGAAGTTG CCTTGAAAAC TTTTTATAGT GTATATTACT	200
GTTCACATAG ATNAGCAATT AACTTTACAT ATACCCGTTT TTAAAAGATC	250
AGTCCTGTGA TTAAAAGTCT GGCTGCCCTA ATTCACTTCG ATTATACATT	300
AGGTTAAAGC CATATAAAAG AGGCACTACG TCTTCGGAGA GATGAATGGA	350
TATTACAAGC AGTAATTTTG GCTTTGGAAT ATACACATAA TGTCCACTTG	400
ACCTCATCTA TTTGACACAA AATGTAAACT AAATTATGAG CATCATTAGA	450
TACCTTGGGC CTTTTCAAAT CACACAGGGT CCTAGATCTG NNNNNNNNNN	500
NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN NNNNNNNNNN	550
NACTTTGGAT TCTTATATCT TTGTCAGCTG TCAACTTCAG TGTTTTCAGG	600
NTAAATTCTA TCCATAGTCA TCCCAATATA CCTGCTTTAG ATGATACAAA	650
CTTCAAAAGA TCCGGCTCTC CCTCGTAAAA CGTGGAGGAC AGACATCAAG	700
GGGGTTTTCT GAGTAAAGAA AGGCAACCGC TCGGCAAAAA CTCACCCTGG	750
CACAACAGGA NCGAATATAT ACAGACGCTG ATTGAGCGTT TTGCTCCATC	800
TTCACTTCTG TTAAATGAAG ACATTGATAT CTAAAATGCT ATGAGTCTAA	850
CTTTGTAAAA TTAAAATAGA TTTGTAGTTA TTTTTCAAAA TGAAATCGAA	900
AAGATACAAG TTTTGAAGGC AGTCTCTTTT TCCACCCTGC CCCTCTAGTG	950
TGTTTTACAC ACTTCTCTGG CCACTCCAAC AGGGAAGCTG GTCCAGGGCC	1000
ATTATACAGG SEQ ID NO: 8832

[0567] The sequence of the CAP2 contig was used in a BLAST search of the human genome. 934 out of 1,010 residues aligned to a region of chromosome 21. A gap of 61 residues divided the aligned region into two smaller fragments. The sequence of this region, plus 100 kilobases on each side of it, was downloaded and analyzed using the Genscan site at MIT, with the following settings:

[0568] Organism: vertebrate

[0569] Suboptimal exon cutoff: 1.00 (no suboptimal exons)

[0570] Print options: Predicted CDS and peptides

[0571] The fragment was found to fall within one of several predicted genes in the chromosome region. The bases corresponding to the predicted gene, including its predicted introns, were saved as a separate file and used to search GenBank again with BLASTN to find any ESTs or UniGene clusters identified by portions of the sequence not included in the original unknown fragment. The nr database contained no significant matches. At least 100 significant matches to various parts of the predicted gene were found in the dbEST database, but all of them were to unnamed cDNA clones. Comparison to UniGene produced fewer significant matches, but all matches were to unnamed cDNAs.

[0572] The peptide sequence predicted by Genscan was also saved. Multiple types of analyses were performed on it using the resources mentioned in Table 3. BLASTP and TBLASTN were used to search the TrEMBL protein database at the Expasy Molecular Biology Server web site and the GenBank nr database located at the NCBI web site, which includes data from the SwissProt, PIR, PRF, and PDB databases. No significant matches were found in any of these, so no gene identity or tertiary structure was discovered.

[0573] The peptide sequence was also searched for similarity to known domains and motifs using BLASTP with the Prosite, Blocks, Pfam, and ProDom databases. The searches produced no significant alignments to known domains. BLASTP comparison to the PRINTS database produced an alignment to the P450 protein family, but with a low probability of accuracy (e=6.9).

[0574] Two methods were used to predict secondary structure—the Garnier/Osguthorpe/Robson model and the Chou-Fasman model. The two methods differed somewhat in their results, but both produced representations of the peptide sequence with helical and sheet regions and locations of turns.

[0575] Antigenicity was plotted as a graph with amino acid number in the sequence on the x-axis and hydrophilicity on the y-axis. Several areas of antigenicity were observed, but only one with antigenicity greater than 2. Hydrophobicity was plotted in the same way. Only one region, from approximately residue 135 to residue 150, had notable hydrophobicity. TMpred, accessed through ExPASy, was used to predict transmembrane helices. No regions of the peptide sequence were predicted with reasonable confidence to be membrane-spanning helices.

[0576] NNPSL predicted that the putative protein would be found either in the nucleus (expected prediction accuracy=51.1%) or secreted from the cell (expected prediction accuracy=91.4%).

[0577] Candidate genes and gene sequences for leukocyte expression profiling were identified through methods described elsewhere in this document. Candidate genes are used to obtain or design probes for peripheral leukocyte expression profiling in a variety of ways. A cDNA microarray carrying 384 probes was constructed using sequences selected from the cDNA libraries described in example 1. cDNAs were selected from T-cell libraries, PBMC libraries and buffy coat libraries. A listing of the cDNA fragments used is given in Table 8.

96-Well PCR

[0578] Plasmids were isolated in 96-well format and PCR was performed in 96-well format. A master mix was made that contain the reaction buffer, dNTPs, forward and reverse primer and DNA polymerase was made. 99 ul of the master mix was aliquoted into 96-well plate. 1 ul of plasmid (1–2 ng/ul) of plasmid was added to the plate. The final reaction concentration was 10 mM Tris pH 8.3, 3.5 mM MgCl2, 25 mM KCl, 0.4 mM dNTPs, 0.4 uM M13 forward primer, 0.4 M13 reverse primer, and 10 U of Taq Gold (Applied Biosystems). The PCR conditions were:

[0579] Step 1 95 C for 10 min

[0580] Step 2 95 C for 15 sec

[0581] Step 3 56 C for 30 sec

[0582] Step 4 72 C for 2 min 15 seconds

[0583] Step 5 go to Step 2 39 times

[0584] Step 6 72 C for 10 minutes

[0585] Step 7 4 C for ever.

PCR Purification

[0586] PCR purification was done in a 96-well format. The ArrayIt (Telechem International, Inc.) PCR purification kit was used and the provided protocol was followed without modification. Before the sample was evaporated to dryness, the concentration of PCR products was determined using a spectrophotometer. After evaporation, the samples were re-suspended in 1× Micro Spotting Solution (ArrayIt) so that the majority of the samples were between 0.2–1.0 ug/ul.

Array Fabrication

[0587] Spotted cDNA microarrays were then made from these PCR products by ArrayIt using their protocols, which may be found at the ArrayIt website. Each fragment was spotted 3 times onto each array.

[0588] Candidate genes and gene sequences for leukocyte expression profiling were identified through methods described elsewhere in this document. Those candidate genes are used for peripheral leukocyte expression profiling. The candidate libraries can used to obtain or design probes for expression profiling in a variety of ways.

[0589] Oligonucleotide probes are also prepared using the DNA sequence information for the candidate genes identified by differential hybridization screening (listed in Table 3 and the sequence listing) and/or the sequence information for the genes identified by database mining (listed in Table 2) is used to design complimentary oligonucleotide probes. Oligo probes are designed on a contract basis by various companies (for example, Compugen, Mergen, Affymetrix, Telechem), or designed from the candidate sequences using a variety of parameters and algorithms as indicated at located at the MIT web site. Briefly, the length of the oligonucleotide to be synthesized is determined, preferably greater than 18 nucleotides, generally 18–24 nucleotides, 24–70 nucleotides and, in some circumstances, more than 70 nucleotides. The sequence analysis algorithms and tools described above are applied to the sequences to mask repetitive elements, vector sequences and low complexity sequences. Oligonucleotides are selected that are specific to the candidate nucleotide sequence (based on a Blast n search of the oligonucleotide sequence in question against gene sequences databases, such as the Human Genome Sequence, UniGene, dbEST or the non-redundant database at NCBI), and have <50% G content and 25–70% G+C content. Desired oligonucleotides are synthesized using well-known methods and apparatus, or ordered from a company (for example Sigma). Oligonucleotides are spotted onto microarrays. Alternatively, oligonucleotides are synthesized directly on the array surface, using a variety of techniques (Hughes et al. 2001, Yershov et al. 1996, Lockhart et al 1996).

[0590] Blood was isolated from the subject for leukocyte expression profiling using the following methods:

[0591] Two tubes were drawn per patient. Blood was drawn from either a standard peripheral venous blood draw or directly from a large-bore intra-arterial or intravenous catheter inserted in the femoral artery, femoral vein, subclavian vein or internal jugular vein. Care was taken to avoid sample contamination with heparin from the intravascular catheters, as heparin can interfere with subsequent RNA reactions.

[0592] For each tube, 8 ml of whole blood was drawn into a tube (CPT, Becton-Dickinson order #362753) containing the anticoagulant Citrate, 25° C. density gradient solution (e.g. Ficoll, Percoll) and a polyester gel barrier that upon centrifugation was permeable to RBCs and granulocytes but not to mononuclear cells. The tube was inverted several times to mix the blood with the anticoagulant. The tubes were centrifuged at 1750×g in a swing-out rotor at room temperature for 20 minutes. The tubes were removed from the centrifuge and inverted 5–10 times to mix the plasma with the mononuclear cells, while trapping the RBCs and the granulocytes beneath the gel barrier. The plasma/mononuclear cell mix was decanted into a 15 ml tube and 5 ml of phosphate-buffered saline (PBS) is added. The 15 ml tubes were spun for 5 minutes at 1750×g to pellet the cells. The supernatant was discarded and 1.8 ml of RLT lysis buffer is added to the mononuclear cell pellet. The buffer and cells were pipetted up and down to ensure complete lysis of the pellet. The cell lysate was frozen and stored until it is convenient to proceed with isolation of total RNA.

[0593] Total RNA was purified from the lysed mononuclear cells using the Qiagen Rneasy Miniprep kit, as directed by the manufacturer (10/99 version) for total RNA isolation, including homogenization (Qiashredder columns) and on-column DNase treatment. The purified RNA was eluted in 50 ul of water. The further use of RNA prepared by this method is described in Example 11, 24, and 23.

[0594] Some samples were prepared by a different protocol, as follows:

[0595] Two 8 ml blood samples were drawn from a peripheral vein into a tube (CPT, Becton-Dickinson order #362753) containing anticoagulant (Citrate), 25° C. density gradient solution (Ficoll) and a polyester gel barrier that upon centrifugation is permeable to RBCs and granulocytes but not to mononuclear cells. The mononuclear cells and plasma remained above the barrier while the RBCs and granulocytes were trapped below. The tube was inverted several times to mix the blood with the anticoagulant, and the tubes were subjected to centrifugation at 1750×g in a swing-out rotor at room temperature for 20 min. The tubes were removed from the centrifuge, and the clear plasma layer above the cloudy mononuclear cell layer was aspirated and discarded. The cloudy mononuclear cell layer was aspirated, with care taken to rinse all of the mononuclear cells from the surface of the gel barrier with PBS (phosphate buffered saline). Approximately 2 mls of mononuclear cell suspension was transferred to a 2 ml microcentrifuge tube, and centrifuged for 3 min. at 16,000 rpm in a microcentrifuge to pellet the cells. The supernatant was discarded and 1.8 ml of RLT lysis buffer (Qiagen) were added to the mononuclear cell pellet, which lysed the cells and inactivated Rnases. The cells and lysis buffer were pipetted up and down to ensure complete lysis of the pellet. Cell lysate was frozen and stored until it was convenient to proceed with isolation of total RNA.

[0596] RNA samples were isolated from 8 mL of whole blood. Yields ranged from 2 ug to 20 ug total RNA for 8 mL blood. A260/A280 spectrophotometric ratios were between 1.6 and 2.0, indicating purity of sample. 2 ul of each sample were run on an agarose gel in the presence of ethidium bromide. No degradation of the RNA sample and no DNA contamination was visible.

[0597] In some cases, specific subsets of mononuclear cells were isolated from peripheral blood of human subjects. When this was done, the StemSep cell separation kits (manual version 6.0.0) were used from StemCell Technologies (Vancouver, Canada). This same protocol can be applied to the isolation of T cells, CD4 T cells, CD8 T cells, B cells, monocytes, NK cells and other cells. Isolation of cell types using negative selection with antibodies may be desirable to avoid activation of target cells by antibodies.

[0598] Control RNA was prepared using total RNA from Buffy coats and/or total RNA from enriched mononuclear cells isolated from Buffy coats, both with and without stimulation with ionomycin and PMA. The following control RNAs were prepared:

[0599] Control 1: Buffy Coat Total RNA

[0600] Control 2: Mononuclear cell Total RNA

[0601] Control 3: Stimulated buffy coat Total RNA

[0602] Control 4: Stimulated mononuclear Total RNA

[0603] Control 5: 50% Buffy coat Total RNA/50% Stimulated buffy coat Total RNA

[0604] Control 6: 50% Mononuclear cell Total RNA/50% Stimulated Mononuclear Total RNA

[0605] Some samples were prepared using the following protocol: Buffy coats from 38 individuals were obtained from Stanford Blood Center. Each buffy coat is derived from ˜350 mL whole blood from one individual. 10 ml buffy coat was removed from the bag, and placed into a 50 ml tube. 40 ml of Buffer EL (Qiagen) was added, the tube was mixed and placed on ice for 15 minutes, then cells were pelleted by centrifugation at 2000×g for 10 minutes at 4° C. The supernatant was decanted and the cell pellet was re-suspended in 10 ml of Qiagen Buffer EL. The tube was then centrifuged at 2000×g for 10 minutes at 4° C. The cell pellet was then re-suspended in 20 ml TRIZOL (GibcoBRL) per Buffy coat sample, the mixture was shredded using a rotary homogenizer, and the lysate was then frozen at −80° C. prior to proceeding to RNA isolation.

[0606] Other control RNAs were prepared from enriched mononuclear cells prepared from Buffy coats. Buffy coats from Stanford Blood Center were obtained, as described above. 10 ml buffy coat was added to a 50 ml polypropylene tube, and 10 ml of phosphate buffer saline (PBS) was added to each tube. A polysucrose (5.7 g/dL) and sodium diatrizoate (9.0 g/dL) solution at a 1.077+/−0.0001 g/ml density solution of equal volume to diluted sample was prepared (Histopaque 1077, Sigma cat. no 1077-1). This and all subsequent steps were performed at room temperature. 15 ml of diluted buffy coat/PBS was layered on top of 15 ml of the histopaque solution in a 50 ml tube. The tube was centrifuged at 400×g for 30 minutes at room temperature. After centrifugation, the upper layer of the solution to within 0.5 cm of the opaque interface containing the mononuclear cells was discarded. The opaque interface was transferred into a clean centrifuge tube.

[0607] An equal volume of PBS was added to each tube and centrifuged at 350×g for 10 minutes at room temperature. The supernatant was discarded. 5 ml of Buffer EL (Qiagen) was used to resuspend the remaining cell pellet and the tube was centrifuged at 2000×g for 10 minutes at room temperature. The supernatant was discarded. The pellet was resuspended in 20 ml of TRIZOL (GibcoBRL) for each individual buffy coat that was processed. The sample was homogenized using a rotary homogenizer and frozen at −80 C until RNA was isolated.

[0608] RNA was isolated from frozen lysed Buffy coat samples as follows: frozen samples were thawed, and 4 ml of chloroform was added to each buffy coat sample. The sample was mixed by vortexing and centrifuged at 2000×g for 5 minutes. The aqueous layer was moved to new tube and then repurified by using the RNeasy Maxi RNA clean up kit, according to the manufacturer's instruction (Qiagen, PN 75162). The yield, purity and integrity were assessed by spectrophotometer and gel electrophoresis.

[0609] Some samples were prepared by a different protocol, as follows. The further use of RNA prepared using this protocol is described in Example 23.

[0610] 50 whole blood samples were randomly selected from consented blood donors at the Stanford Medical School Blood Center. Each buffy coat sample was produced from ˜350 mL of an individual's donated blood. The whole blood sample was centrifuged at ˜4,400×g for 8 minutes at room temperature, resulting in three distinct layers: a top layer of plasma, a second layer of buffy coat, and a third layer of red blood cells. 25 ml of the buffy coat fraction was obtained and diluted with an equal volume of PBS (phosphate buffered saline). 30 ml of diluted buffy coat was layered onto 15 ml of sodium diatrizoate solution adjusted to a density of 1.077+/−0.001 g/ml (Histopaque 1077, Sigma) in a 50 mL plastic tube. The tube was spun at 800 g for 10 minutes at room temperature. The plasma layer was removed to the 30 ml mark on the tube, and the mononuclear cell layer removed into a new tube and washed with an equal volume of PBS, and collected by centrifugation at 2000 g for 10 minutes at room temperature. The cell pellet was resuspended in 10 ml of Buffer EL (Qiagen) by vortexing and incubated on ice for 10 minutes to remove any remaining erthythrocytes. The mononuclear cells were spun at 2000 g for 10 minutes at 4 degrees Celsius. The cell pellet was lysed in 25 ml of a phenol/guanidinium thiocyanate solution (TRIZOL Reagent, Invitrogen). The sample was homogenized using a PowerGene 5 rotary homogenizer (Fisher Scientific) and Omini disposable generator probes (Fisher Scientific). The Trizol lysate was frozen at −80 degrees C. until the next step.

[0611] The samples were thawed out and incubated at room temperature for 5 minutes. 5 ml chloroform was added to each sample, mixed by vortexing, and incubated at room temperature for 3 minutes. The aqueous layers were transferred to new 50 ml tubes. The aqueous layer containing total RNA was further purified using the Qiagen RNeasy Maxi kit (PN 75162), per the manufacturer's protocol (October 1999). The columns were eluted twice with 1 ml Rnase-free water, with a minute incubation before each spin.

[0612] Quantity and quality of RNA was assessed using standard methods. Generally, RNA was isolated from batches of 10 buffy coats at a time, with an average yield per buffy coat of 870 μg, and an estimated total yield of 43.5 mg total RNA with a 260/280 ratio of 1.56 and a 28S/18S ratio of 1.78.

[0613] Quality of the RNA was tested using the Agilent 2100 Bioanalyzer using RNA 6000 microfluidics chips. Analysis of the electrophorgrams from the Bioanalyzer for five different batches demonstrated the reproducibility in quality between the batches.

[0614] Total RNA from all five batches were combined and mixed in a 50 ml tube, then aliquoted as follows: 2×10 ml aliquots in 15 ml tubes, and the rest in 100 μl aliquots in 1.5 ml microcentrifuge tubes. The aliquots gave highly reproducible results with respect to RNA purity, size and integrity. The RNA was stored at −80° C.

Test Hybridization of Reference RNA.

[0615] When compared with BC38 and Stimulated mononuclear reference samples, the R50 performed as well, if not better than the other reference samples as shown in FIG. 3. In an analysis of hybridizations, where the R50 targets were fluorescently labeled with Cy-5 using methods described herein and the amplified and labeled aRNA was hybridized (as in example 23) to the oligonucleotide array described in examples 20–22. The R50 detected 97.3% of probes with a Signal to Noise ratio (S/N) of greater than three and 99.9% of probes with S/N greater than one.

Comparison of Guanine-Silica to Acid-Phenol RNA Purification (GSvsAP)

[0616] These data are from a set of 12 hybridizations designed to identify differences between the signal strength from two different RNA purification methods. The two RNA methods used were guanidine-silica (GS, Qiagen) and acid-phenol (AP, Trizol, Gibco BRL). Ten tubes of blood were drawn from each of four people. Two were used for the AP prep, the other eight were used for the GS prep. The protocols for the leukocyte RNA preps using the AP and GS techniques were completed as described here:

Guanidine-Silica (GS) Method:

[0617] For each tube, 8 ml blood was drawn into a tube containing the anticoagulant Citrate, 25° C. density gradient solution and a polyester gel barrier that upon centrifugation is permeable to RBCs and granulocytes but not to mononuclear cells. The mononuclear cells and plasma remained above the barrier while the RBCs and granulocytes were trapped below. CPT tubes from Becton-Dickinson (#362753) were used for this purpose. The tube was inverted several times to mix the blood with the anticoagulant. The tubes were immediately centrifuged @1750×g in a swinging bucket rotor at room temperature for 20 min. The tubes were removed from the centrifuge and inverted 5–10 times. This mixed the plasma with the mononuclear cells, while the RBCs and the granulocytes remained trapped beneath the gel barrier. The plasma/mononuclear cell mix was decanted into a 15 ml tube and 5 ml of phosphate-buffered saline (PBS) was added. The 15 ml tubes are spun for 5 minutes at 1750×g to pellet the cells. The supernatant was discarded and 1.8 ml of RLT lysis buffer (guanidine isothiocyanate) was added to the mononuclear cell pellet. The buffer and cells were pipetted up and down to ensure complete lysis of the pellet. The cell lysate was then processed exactly as described in the Qiagen Rneasy Miniprep kit protocol (10/99 version) for total RNA isolation (including steps for homogenization (Qiashredder columns) and on-column DNase treatment. The purified RNA was eluted in 50 ul of water.

Acid-Phenol (AP) Method:

[0618] For each tube, 8 ml blood was drawn into a tube containing the anticoagulant Citrate, 25° C. density gradient solution and a polyester gel barrier that upon centrifugation is permeable to RBCs and granulocytes but not to mononuclear cells. The mononuclear cells and plasma remained above the barrier while the RBCs and granulocytes were trapped below. CPT tubes from Becton-Dickinson (#362753) were used for this purpose. The tube was inverted several times to mix the blood with the anticoagulant. The tubes were immediately centrifuged @1750×g in a swinging bucket rotor at room temperature for 20 min. The tubes were removed from the centrifuge and inverted 5–10 times. This mixed the plasma with the mononuclear cells, while the RBCs and the granulocytes remained trapped beneath the gel barrier. The plasma/mononuclear cell mix was decanted into a 15 ml tube and 5 ml of phosphate-buffered saline (PBS) was added. The 15 ml tubes are spun for 5 minutes @1750×g to pellet the cells. The supernatant was discarded and the cell pellet was lysed using 0.6 mL Phenol/guanidine isothiocyanate (e.g. Trizol reagent, GibcoBRL). Subsequent total RNA isolation proceeded using the manufacturers protocol.

[0619] RNA from each person was labeled with either Cy3 or Cy5, and then hybridized in pairs to the mini-array. For instance, the first array was hybridized with GS RNA from one person (Cy3) and GS RNA from a second person (Cy5).

[0620] Techniques for labeling and hybridization for all experiments discussed here were completed as detailed above in example 10. Arrays were prepared as described in example 7.

[0621] RNA isolated from subject samples, or control Buffy coat RNA, were labeled for hybridization to a cDNA array. Total RNA (up to 100 μg) was combined with 2 μl of 100 μM solution of an Oligo (dT)12–18 (GibcoBRL) and heated to 70° C. for 10 minutes and place on ice. Reaction buffer was added to the tube, to a final concentration of 1×RT buffer (GibcoBRL), 10 mM DTT (GibcoBRL), 0.1 mM unlabeled dATP, dTTP, and dGTP, and 0.025 mM unlabeled dCTP, 200 μg of CAB (A. thaliana photosystem I chlorophyll a/b binding protein), 200 pg of RCA (A. thaliana RUBISCO activase), 0.25 mM of Cy-3 or Cy-5 dCTP, and 400 U Superscript II RT (GibcoBRL).

[0622] The volumes of each component of the labeling reaction were as follows: 20 μl of 5×RT buffer; 10 μl of 100 mM DTT; 1 μl of 10 mM dNTPs without dCTP; 0.5 μl of 5 mM CTP; 13 μl of H20; 0.02 μl of 10 ng/μl CAB and RCA; 1 μl of 40 Units/μl RNAseOUT Recombinant Ribonuclease Inhibitor (GibcoBRL); 2.5 μl of 1.0 mM Cy-3 or Cy-5 dCTP; and 2.0 μl of 200 Units/μl of Superscript II RT. The sample was vortexed and centrifuged. The sample was incubated at 4° C. for 1 hour for first strand cDNA synthesis, then heated at 70° C. for 10 minutes to quench enzymatic activity. 1 μl of 10 mg/ml of Rnase A was added to degrade the RNA strand, and the sample was incubated at 37° C. for 30 minutes.

[0623] Next, the Cy-3 and Cy-5 cDNA samples were combined into one tube. Unincorporated nucleotides were removed using QlAquick RCR purification protocol (Qiagen), as directed by the manufacturer. The sample was evaporated to dryness and resuspended in 5 μL of water. The sample was mixed with hybridization buffer containing 5×SSC, 0.2% SDS, 2 mg/ml Cot-1 DNA (GibcoBRL), 1 mg/ml yeast tRNA (GibcoBRL), and 1.6 ng/μl poly dA40–60 (Pharmacia). This mixture was placed on the microarray surface and a glass cover slip was placed on the array (Corning). The microarray glass slide was placed into a hybridization chamber (ArrayIt). The chamber was then submerged in a water bath overnight at 62° C. The microarray was removed from the cassette and the cover slip was removed by repeatedly submerging it to a wash buffer containing 1×SSC, and 0.1% SDS. The microarray slide was washed in 1×SSC/0.1% SDS for 5 minutes. The slide was then washed in 0.1% SSC/0.1% SDS for 5 minutes. The slide was finally washed in 0.1×SSC for 2 minutes. The slide was spun at 1000 rpm for 2 minutes to dry out the slide, then scanned on a microarray scanner (Axon Instruments, Union City, Calif.). Six hybridizations with 20 μg of RNA were performed for each type of RNA preparation (GS or AP). Since both the Cy3 and the Cy5 labeled RNA are from test preparations, there are six data points for each GS prepped, Cy3-labeled RNA and six for each GS-prepped, Cy5-labeled RNA. The mini array hybridizations were scanned on and Axon Instruments scanner using GenPix 3.0 software. The data presented were derived as follows. First, all features flagged as “not found” by the software were removed from the dataset for individual hybridizations. These features are usually due to high local background or other processing artifacts. Second, the median fluorescence intensity minus the background fluorescence intensity was used to calculate the mean background subtracted signal for each dye for each hybridization. In FIG. 3, the mean of these means across all six hybridizations is graphed (n=6 for each column). The error bars are the SEM. This experiment shows that the average signal from AP prepared RNA is 47% of the average signal from GS prepared RNA for both Cy3 and Cy5.

Generation of Expression Data for Leukocyte Genes from Peripheral Leukocyte Samples

[0624] Six hybridizations were performed with RNA purified from human blood leukocytes using the protocols given above. Four of the six were prepared using the GS method and 2 were prepared using the AP method. Each preparation of leukocyte RNA was labeled with Cy3 and 10 μg hybridized to the mini-array. A control RNA was batch labeled with Cy5 and 10 μg hybridized to each mini-array together with the Cy3-labeled experimental RNA.

[0625] The control RNA used for these experiments was Control 1: Buffy Coat RNA, as described above. The protocol for the preparation of that RNA is reproduced here:

Buffy Coat RNA Isolation:

[0626] Buffy coats were obtained from Stanford Blood Center (in total 38 individual buffy coats were used. Each buffy coat is derived from ˜350 mL whole blood from one individual. 10 ml buffy coat was taken and placed into a 50 ml tube and 40 ml of a hypochlorous acid (HOCl) solution (Buffer EL from Qiagen) was added. The tube was mixed and placed on ice for 15 minutes. The tube was then centrifuged at 2000×g for 10 minutes at 4° C. The supernatant was decanted and the cell pellet was re-suspended in 10 ml of hypochlorous acid solution (Qiagen Buffer EL). The tube was then centrifuged at 2000×g for 10 minutes at 4° C. The cell pellet was then re-suspended in 20 ml phenol/guanidine thiocyanate solution (TRIZOL from GibcoBRL) for each individual buffy coat that was processed. The mixture was then shredded using a rotary homogenizer. The lysate was then frozen at −80° C. prior to proceeding to RNA isolation.

[0627] The arrays were then scanned and analyzed on an Axon Instruments scanner using GenePix 3.0 software. The data presented were derived as follows. First, all features flagged as “not found” by the software were removed from the dataset for individual hybridizations. Second, control features were used to normalize the data for labeling and hybridization variability within the experiment. The control features are cDNA for genes from the plant, Arabidopsis thaliana, that were included when spotting the mini-array. Equal amounts of RNA complementary to two of these cDNAs were added to each of the samples before they were labeled. A third was pre-labeled and equal amounts were added to each hybridization solution before hybridization. Using the signal from these genes, we derived a normalization constant (L_j) according to the following formula:

[0628] $L_j=\frac{\frac{\sum _{i=1}^N{\mathrm{BGSS}}_{j,i}}{N}}{\frac{\sum _{j=1}^K\frac{\sum _{i=1}^N{\mathrm{BGSS}}_{j,i}}{N}}{K}}$
where BGSS_i is the signal for a specific feature as identified in the GenePix software as the median background subtracted signal for that feature, N is the number of A. thaliana control features, K is the number of hybridizations, and L is the normalization constant for each individual hybridization.

[0629] Using the formula above, the mean over all control features of a particular hybridization and dye (eg Cy3) was calculated. Then these control feature means for all Cy3 hybridizations were averaged. The control feature mean in one hybridization divided by the average of all hybridizations gives a normalization constant for that particular Cy3 hybridization.

[0630] The same normalization steps were performed for Cy3 and Cy5 values, both fluorescence and background. Once normalized, the background Cy3 fluorescence was subtracted from the Cy3 fluorescence for each feature. Values less than 100 were eliminated from further calculations since low values caused spurious results.

[0631] FIG. 4 shows the average background subtracted signal for each of nine leukocyte-specific genes on the mini array. This average is for 3–6 of the above-described hybridizations for each gene. The error bars are the SEM.

[0632] The ratio of Cy3 to Cy5 signal is shown for a number of genes. This ratio corrects for variability among hybridizations and allows comparison between experiments done at different times. The ratio is calculated as the Cy3 background subtracted signal divided by the Cy5 background subtracted signal. Each bar is the average for 3–6 hybridizations. The error bars are SEM.

[0633] Together, these results show that we can measure expression levels for genes that are expressed specifically in sub-populations of leukocytes. These expression measurements were made with only 10 μg of leukocyte total RNA that was labeled directly by reverse transcription. The signal strength can be increased by improved labeling techniques that amplify either the starting RNA or the signal fluorescence. In addition, scanning techniques with higher sensitivity can be used.

[0634] Genes in FIGS. 4 and 5:

[0635]

	GenBank
	Accession	Gene Name
Gene Name/Description	Number	Abbreviation
T cell-specific tyrosine kinase Mrna	L10717	TKTCS
Interleukin 1 alpha (IL 1) mRNA, complete	NM_000575	IL1A
cds
T-cell surface antigen CD2 (T11) mRNA,	M14362	CD2
complete cds
Interleukin-13 (IL-13) precursor gene,	U31120	IL-13
complete cds
Thymocyte antigen CD1a mRNA, complete	M28825	CD1a
cds
CD6 mRNA for T cell glycoprotein CDS	NM_006725	CD6
MHC class II HLA-DQA1 mRNA, complete	U77589	HLA-DQA1
cds
Granulocyte colony-stimulating factor	M28170	CD19
Homo sapiens CD69 antigen	NM_001781	CD69

[0636] An observational study was conducted in which a prospective cohort of cardiac transplant recipients were analyzed for associations between clinical events or rejection grades and expression of a leukocyte candidate nucleotide sequence library. Patients were identified at 4 cardiac transplantation centers while on the transplant waiting list or during their routing post-transplant care. All adult cardiac transplant recipients (new or re-transplants) who received an organ at the study center during the study period or within 3 months of the start of the study period were eligible. The first year after transplantation is the time when most acute rejection occurs and it is thus important to study patients during this period. Patients provided informed consent prior to study procedures. Peripheral blood leukocyte samples were obtained from all patients at the following time points: prior to transplant surgery (when able), the same day as routinely scheduled screening biopsies, upon evaluation for suspected acute rejection (urgent biopsies), on hospitalization for an acute complication of transplantation or immunosuppression, and when Cytomegalovirus (CMV) infection was suspected or confirmed. Samples were obtained through a standard peripheral vein blood draw or through a catheter placed for patient care (for example, a central venous catheter placed for endocardial biopsy). When blood was drawn from a intravenous line, care was taken to avoid obtaining heparin with the sample as it can interfere with downstream reactions involving the RNA. Mononuclear cells were prepared from whole blood samples as described in Example 8. Samples were processed within 2 hours of the blood draw and DNA and serum were saved in addition to RNA. Samples were stored at −80° C. or on dry ice and sent to the site of RNA preparation in a sealed container with ample dry ice. RNA was isolated from subject samples as described in Example 8 and hybridized to a candidate library of differentially expressed leukocyte nucleotide sequences, as further described in Examples 20–22. Methods used for amplification, labeling, hybridization and scanning are described in Example 23. Analysis of human transplant patient mononuclear cell RNA hybridized to a microarray and identification of diagnostic gene sets is shown in Example 24.

[0637] From each patient, clinical information was obtained at the following time points: prior to transplant surgery (when available), the same day as routinely scheduled screening biopsies, upon evaluation for suspected acute rejection (e.g., urgent biopsies), on hospitalization for an acute complication of transplantation or immunosuppression, and when Cytomegalovirus (CMV) infection was suspected or confirmed. Data was collected directly from the patient, from the patient's medical record, from diagnostic test reports or from computerized hospital databases. It was important to collect all information pertaining to the study clinical correlates (diagnoses and patient events and states to which expression data is correlated) and confounding variables (diagnoses and patient events and states that may result in altered leukocyte gene expression. Examples of clinical data collected are: patient sex, date of birth, date of transplant, race, requirement for prospective cross match, occurrence of pre-transplant diagnoses and complications, indication for transplantation, severity and type of heart disease, history of left ventricular assist devices, all known medical diagnoses, blood type, HLA type, viral serologies (including CMV, Hepatitis B and C, HIV and others), serum chemistries, white and red blood cell counts and differentials, CMV infections (clinical manifestations and methods of diagnosis), occurrence of new cancer, hemodynamic parameters measured by catheterization of the right or left heart (measures of graft function), results of echocardiography, results of coronary angiograms, results of intravascular ultrasound studies (diagnosis of transplant vasculopathy), medications, changes in medications, treatments for rejection, and medication levels. Information was also collected regarding the organ donor, including demographics, blood type, HLA type, results of screening cultures, results of viral serologies, primary cause of brain death, the need for inotropic support, and the organ cold ischemia time.

[0638] Of great importance was the collection of the results of endocardial biopsy for each of the patients at each visit. Biopsy results were all interpreted and recorded using the international society for heart and lung transplantation (ISHLT) criteria, described below. Biopsy pathological grades were determined by experienced pathologists at each center.

[0639]

ISHLT Criteria
		Rejection
Grade	Finding	Severity
0	No lymphocytic infiltrates	None
1A	Focal (perivascular or interstitial lymphocytic	Borderline
	infiltrates without necrosis)	mild
1B	Diffuse but sparse lymphocytic infiltrates without	Mild
	necrosis
2	One focus only with aggressive lymphocytic	Mild,
	infiltrate and/or myocyte damage	focal
		moderate
3A	Multifocal aggressive lymphocytic infiltrates and/or	Moderate
	myocardial damage
3B	Diffuse inflammatory lymphocytic infiltrates with	Borderline
	necrosis	Severe
4	Diffuse aggressive polymorphous lymphocytic	Severe
	infiltrates with edema hemorrhage and vasculitis,
	with necrosis

[0640] Because variability exists in the assignment of ISHLT grades, it was important to have a centralized and blinded reading of the biopsy slides by a single pathologist. This was arranged for all biopsy slides associated with samples in the analysis. Slides were obtained and assigned an encoded number. A single pathologist then read all slides from all centers and assigned an ISHLT grade. Grades from the single pathologist were then compared to the original grades derived from the pathologists at the study centers. For the purposes of correlation analysis of leukocyte gene expression to biopsy grades, the centralized reading information was used in a variety of ways (see Example 24 for more detail). In some analyses, only the original reading was used as an outcome. In other analyses, the result from the centralized reader was used as an outcome. In other analyses, the highest of the 2 grades was used. For example, if the original assigned grade was 0 and the centralized reader assigned a 1A, then 1A was the grade used as an outcome. In some analyses, the highest grade was used and then samples associated with a Grade 1A reading were excluded from the analysis. In some analyses, only grades with no disagreement between the 2 readings were used as outcomes for correlation analysis. Clinical data was entered and stored in a database. The database was queried to identify all patients and patient visits that meet desired criteria (for example, patients with >grade II biopsy results, no CMV infection and time since transplant <12 weeks).

[0641] The collected clinical data (disease criteria) is used to define patient or sample groups for correlation of expression data. Patient groups are identified for comparison, for example, a patient group that possesses a useful or interesting clinical distinction, versus a patient group that does not possess the distinction. Examples of useful and interesting patient distinctions that can be made on the basis of collected clinical data are listed here:

[0642] 1. Rejection episode of at least moderate histologic grade, which results in treatment of the patient with additional corticosteroids, anti-T cell antibodies, or total lymphoid irradiation.

[0643] 2. Rejection with histologic grade 2 or higher.

[0644] 3. Rejection with histologic grade <2.

[0645] 4. The absence of histologic rejection and normal or unchanged allograft function (based on hemodynamic measurements from catheterization or on echocardiographic data).

[0646] 5. The presence of severe allograft dysfunction or worsening allograft dysfunction during the study period (based on hemodynamic measurements from catheterization or on echocardiographic data).

[0647] 6. Documented CMV infection by culture, histology, or PCR, and at least one clinical sign or symptom of infection.

[0648] 7. Specific graft biopsy rejection grades

[0649] 8. Rejection of mild to moderate histologic severity prompting augmentation of the patient's chronic immunosuppressive regimen

[0650] 9. Rejection of mild to moderate severity with allograft dysfunction prompting plasmapheresis or a diagnosis of “humoral” rejection

[0651] 10. Infections other than CMV, esp. Epstein Barr virus (EBV)

[0652] 11. Lymphoproliferative disorder (also called, post-transplant lymphoma)

[0653] 12. Transplant vasculopathy diagnosed by increased intimal thickness on intravascular ultrasound (IVUS), angiography, or acute myocardial infarction.

[0654] 13. Graft Failure or Retransplantation

[0655] 14. All cause mortality

[0656] 15. Grade 1A or higher rejection as defined by the initial biopsy reading.

[0657] 16. Grade 1B or higher rejection as defined by the initial biopsy reading.

[0658] 17. Grade 1A or higher rejection as defined by the centralized biopsy reading.

[0659] 18. Grade 1B or higher rejection as defined by the centralized biopsy reading.

[0660] 19. Grade 1A or higher rejection as defined by the highest of the initial and centralized biopsy reading.

[0661] 20. Grade 1B or higher rejection as defined by the highest of the initial and centralized biopsy reading.

[0662] 21. Any rejection >Grade 2 occurring in patient at any time in the post-transplant course.

[0663] Expression profiles of subject samples are examined to discover sets of nucleotide sequences with differential expression between patient groups, for example, by methods describes above and below.

[0664] Non-limiting examples of patient leukocyte samples to obtain for discovery of various diagnostic nucleotide sets are as follows:

[0665] Leukocyte set to avoid biopsy or select for biopsy:

[0666] Samples: Grade 0 vs. Grades 1–4

[0667] Leukocyte set to monitor therapeutic response:

[0668] Examine successful vs. unsuccessful drug treatment.

[0669] Samples:

[0670] Successful: Time 1: rejection, Time 2: drug therapy Time 3: no rejection

• • [0671] Unsuccessful: Time 1: rejection, Time 2: drug therapy; Time 3: rejection

[0672] Leukocyte set to predict subsequent acute rejection.

[0673] Biopsy may show no rejection, but the patient may develop rejection shortly thereafter.

[0674] Look at profiles of patients who subsequently do and do not develop rejection.

[0675] Samples:

[0676] Group 1 (Subsequent rejection): Time 1: Grade 0; Time 2: Grade>0

[0677] Group 2 (No subsequent rejection): Time 1: Grade 0; Time 2: Grade 0

[0678] Focal rejection may be missed by biopsy. When this occurs the patient may have a Grade 0, but actually has rejection. These patients may go on to have damage to the graft etc.

[0679] Samples:

[0680] Non-rejectors: no rejection over some period of time

[0681] Rejectors: an episode of rejection over same period

[0682] Leukocyte set to diagnose subsequent or current graft failure:

[0683] Samples:

[0684] Echocardiographic or catheterization data to define worsening function over time and correlate to profiles.

[0685] Leukocyte set to diagnose impending active CMV:

[0686] Samples:

[0687] Look at patients who are CMV IgG positive. Compare patients with subsequent (to a sample) clinical CMV infection verses no subsequent clinical CMV infection.

[0688] Leukocyte set to diagnose current active CMV:

[0689] Samples:

[0690] Analyze patients who are CMV IgG positive. Compare patients with active current clinical CMV infection vs. no active current CMV infection.

[0691] Upon identification of a nucleotide sequence or set of nucleotide sequences that distinguish patient groups with a high degree of accuracy, that nucleotide sequence or set of nucleotide sequences is validated, and implemented as a diagnostic test. The use of the test depends on the patient groups that are used to discover the nucleotide set. For example, if a set of nucleotide sequences is discovered that have collective expression behavior that reliably distinguishes patients with no histological rejection or graft dysfunction from all others, a diagnostic is developed that is used to screen patients for the need for biopsy. Patients identified as having no rejection do not need biopsy, while others are subjected to a biopsy to further define the extent of disease. In another example, a diagnostic nucleotide set that determines continuing graft rejection associated with myocyte necrosis (>grade I) is used to determine that a patient is not receiving adequate treatment under the current treatment regimen. After increased or altered immunosuppressive therapy, diagnostic profiling is conducted to determine whether continuing graft rejection is progressing. In yet another example, a diagnostic nucleotide set(s) that determine a patient's rejection status and diagnose cytomegalovirus infection is used to balance immunosuppressive and anti-viral therapy.

[0692] The methods of this example are also applicable to cardiac xenograft monitoring.

[0693] Diagnostic tests for rejection are identified using patient leukocyte expression profiles to identify a molecular signature correlated with rejection of a transplanted kidney or liver.

[0694] Blood, or other leukocyte source, samples are obtained from patients undergoing kidney or liver biopsy following liver or kidney transplantation, respectively. Such results reveal the histological grade, i.e., the state and severity of allograft rejection. Expression profiles are obtained from the samples as described above, and the expression profile is correlated with biopsy results. In the case of kidney rejection, clinical data is collected corresponding to urine output, level of creatine clearance, and level of serum creatine (and other markers of renal function). Clinical data collected for monitoring liver transplant rejection includes, biochemical characterization of serum markers of liver damage and function such as SGOT, SGPT, Alkaline phosphatase, GGT, Bilirubin, Albumin and Prothrombin time.

[0695] Leukocyte nucleotide sequence expression profiles are collected and correlated with important clinical states and outcomes in renal or hepatic transplantation. Examples of useful clinical correlates are given here:

[0696] 1. Rejection episode of at least moderate histologic grade, which results in treatment of the patient with additional corticosteroids, anti-T cell antibodies, or total lymphoid irradiation.

[0697] 2. The absence of histologic rejection and normal or unchanged allograft function (based on tests of renal or liver function listed above).

[0698] 3. The presence of severe allograft dysfunction or worsening allograft dysfunction during the study period (based on tests of renal and hepatic function listed above).

[0699] 4. Documented CMV infection by culture, histology, or PCR, and at least one clinical sign or symptom of infection.

[0700] 5. Specific graft biopsy rejection grades

[0701] 6. Rejection of mild to moderate histologic severity prompting augmentation of the patient's chronic immunosuppressive regimen

[0702] 7. Infections other than CMV, esp. Epstein Barr virus (EBV)

[0703] 8. Lymphoproliferative disorder (also called, post-transplant lymphoma)

[0704] 9. Graft Failure or Retransplantation

[0705] 10. Need for hemodialysis or other renal replacement therapy for renal transplant patients.

[0706] 11. Hepatic encephalopathy for liver transplant recipients.

[0707] 12. All cause mortality

[0708] Subsets of the candidate library (or of a previously identified diagnostic nucleotide set), are identified, according to the above procedures, that have predictive and/or diagnostic value for kidney or liver allograft rejection.

Prediction of Complications of Atherosclerosis: Angina Pectoris.

[0709] Over 50 million in the US have atherosclerotic coronary artery disease (CAD). Almost all adults have some atherosclerosis. The most important question is who will develop complications of atherosclerosis. Patients with angiographically-confirmed atherosclerosis are enrolled in a study, and followed over time. Leukocyte expression profiles are taken at the beginning of the study, and routinely thereafter. Some patients develop angina and others do not. Expression profiles are correlated with development of angina, and subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures, that have predictive and/or diagnostic value for angina pectoris.

[0710] Alternatively, patients are followed by serial angiography. Profiles are collected at the first angiography, and at a repeat angiography at some future time (for example, after 1 year). Expression profiles are correlated with progression of disease, measured, for example, by decrease in vessel lumen diameter. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures, that have predictive and/or diagnostic value for progression of atherosclerosis.

Prediction and/or Diagnosis of Acute Coronary Syndrome

[0711] The main cause of death due to coronary atherosclerosis is the occurrence of acute coronary syndromes: myocardial infarction and unstable angina. Patients with at a very high risk of acute coronary syndrome (e.g., patients with a history of acute coronary syndrome, patients with atherosclerosis, patients with multiple traditional risk factors, clotting disorders or lupus) are enrolled in a prospective study. Leukocyte expression profiles are taken at the beginning of the study period and patients are monitored for the occurrence of unstable angina and/or myocardial infarction. Standard criteria for the occurrence of an event are used (serum enzyme elevation, EKG, nuclear imaging or other), and the occurrence of these events can be collected from the patient, the patient's physician, the medical record or medical database. Expression profiles (taken at the beginning of the study) are correlated with the occurrence of an acute event. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures, that have predictive value for occurrence of an acute event.

[0712] In addition, expression profiles (taken at the time that an acute event occurs) are correlated with the occurrence of an acute event. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures, that have diagnostic value for occurrence of an acute event.

Risk Stratification: Occurrence of Coronary Artery Disease

[0713] The established and classic risks for the occurrence of coronary artery disease and complications of that disease are: cigarette smoking, diabetes, hypertension, hyperlipidemia and a family history of early atherosclerosis. Obesity, sedentary lifestyle, syndrome X, cocaine use, chronic hemodialysis and renal disease, radiation exposure, endothelial dysfunction, elevated plasma homocysteine, elevated plasma lipoprotein a, and elevated CRP. Infection with CMV and chlamydia infection are less well established, controversial or putative risk factors for the disease. These risk factors can be assessed or measured in a population.

[0714] Leukocyte expression profiles are measured in a population possessing risk factors for the occurrence of coronary artery disease. Expression profiles are correlated with the presence of one or more risk factors (that may correlate with future development of disease and complications). Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures, that have predictive value for the development of coronary artery disease.

[0715] Additional examples of useful correlation groups in cardiology include:

• • [0716] 1. Samples from patients with a high risk factor burden (e.g., smoking, diabetes, high cholesterol, hypertension, family history) versus samples from those same patients at different times with fewer risks, or versus samples from different patients with fewer or different risks.

[0717] 2. Samples from patients during an episode of unstable angina or myocardial infarction versus paired samples from those same patients before the episode or after recovery, or from different patients without these diagnoses.

[0718] 3. Samples from patients (with or without documented atherosclerosis) who subsequently develop clinical manifestations of atherosclerosis such as stable angina, unstable angina, myocardial infarction, or stroke versus samples from patients (with or without atherosclerosis) who do not develop these manifestations over the same time period.

[0719] 4. Samples from patients who subsequently respond to a given medication or treatment regimen versus samples from those same or different patients who subsequently do not respond to a given medication or treatment regimen.

[0720] Restenosis is the re-narrowing of a coronary artery after an angioplasty. Patients are identified who are about to, or have recently undergone angioplasty. Leukocyte expression profiles are measured before the angioplasty, and at 1 day and 1–2 weeks after angioplasty or stent placement. Patients have a follow-up angiogram at 3 months and/or are followed for the occurrence of clinical restenosis, e.g., chest pain due to re-narrowing of the artery, that is confirmed by angiography. Expression profiles are compared between patients with and without restenosis, and candidate nucleotide profiles are correlated with the occurrence of restenosis. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures, that have predictive value for the development of restenosis.

[0721] CHF effects greater than 5 million individuals in the US and the prevalence of this disorder is growing as the population ages. The disease is chronic and debilitating. Medical expenditures are huge due to the costs of drug treatments, echocardiograms and other tests, frequent hospitalization and cardiac transplantation. The primary causes of CHF are coronary artery disease, hypertension and idiopathic cardiomyopathy.

[0722] Congestive heart failure is the number one indication for heart transplantation.

[0723] There is ample recent evidence that congestive heart failure is associated with systemic inflammation. A leukocyte test with the ability to determine the rate of progression and the adequacy of therapy is of great interest. Patients with severe CHF are identified, e.g. in a CHF clinic, an inpatient service, or a CHF study or registry (such as the cardiac transplant waiting list/registry). Expression profiles are taken at the beginning of the study and patients are followed over time, for example, over the course of one year, with serial assessments performed at least every three months. Further profiles are taken at clinically relevant end-points, for example: hospitalization for CHF, death, pulmonary edema, worsening of Ejection Fraction or increased cardiac chamber dimensions determined by echocardiography or another imaging test, and/or exercise testing of hemodynamic measurements. Clinical data is collected from patients if available, including:

[0724] Serial C-Reactive Protein (CRP), other serum markers, echocardiography (e.g., ejection fraction or another echocardiographic measure of cardiac function), nuclear imaging, NYHA functional classes, hospitalizations for CHF, quality of life measures, renal function, transplant listing, pulmonary edema, left ventricular assist device use, medication use and changes.

[0725] Expression profiles correlating with progression of CHF are identified. Expression profiles predicting disease progression, monitoring disease progression and response to treatment, and predicting response to a particular treatment(s) or class of treatment(s) are identified. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures, that have predictive value for the progression of CHF. Such diagnostic nucleotide sets are also useful for monitoring response to treatment for CHF.

[0726] Rheumatoid arthritis (hereinafter, “RA”) is a chronic and debilitating inflammatory arthritis. The diagnosis of RA is made by clinical criteria and radiographs. A new class of medication, TNF blockers, are effective, but the drugs are expensive, have side effects and not all patients respond to treatment. In addition, relief of disease symptoms does not always correlate with inhibition of joint destruction. For these reasons, an alternative mechanism for the titration of therapy is needed.

[0727] An observational study was conducted in which a cohort of patients meeting American College of Rheumatology (hereinafter “ARC”) criteria for the diagnosis of RA was identified. Arnett et al. (1988) Arthritis Rheum 31:315–24. Patients gave informed consent and a peripheral blood mononuclear cell RNA sample was obtained by the methods as described herein. When available, RNA samples were also obtained from surgical specimens of bone or synovium from effected joints, and synovial fluid.

[0728] From each patient, the following clinical information was obtained if available:

[0729] Demographic information; information relating to the ACR criteria for RA; presence or absence of additional diagnoses of inflammatory and non-inflammatory conditions; data from laboratory test, including complete blood counts with differentials, CRP, ESR, ANA, Serum IL6, Soluble CD40 ligand, LDL, HDL, Anti-DNA antibodies, rheumatoid factor, C3, C4, serum creatinine and any medication levels; data from surgical procedures such as gross operative findings and pathological evaluation of resected tissues and biopsies; information on pharmacological therapy and treatment changes; clinical diagnoses of disease “flare”; hospitalizations; quantitative joint exams; results from health assessment questionnaires (HAQs); other clinical measures of patient symptoms and disability; physical examination results and radiographic data assessing joint involvement, synovial thickening, bone loss and erosion and joint space narrowing and deformity.

[0730] From these data, measures of improvement in RA are derived as exemplified by the ACR 20% and 50% response/improvement rates (Felson et al. 1996). Measures of disease activity over some period of time is derived from these data as are measures of disease progression. Serial radiography of effected joints is used for objective determination of progression (e.g., joint space narrowing, peri-articular osteoporosis, synovial thickening). Disease activity is determined from the clinical scores, medical history, physical exam, lab studies, surgical and pathological findings.

[0731] The collected clinical data (disease criteria) is used to define patient or sample groups for correlation of expression data. Patient groups are identified for comparison, for example, a patient group that possesses a useful or interesting clinical distinction, verses a patient group that does not possess the distinction. Examples of useful and interesting patient distinctions that can be made on the basis of collected clinical data are listed here:

[0732] 1. Samples from patients during a clinically diagnosed RA flare versus samples from these same or different patients while they are asymptomatic.

[0733] 2. Samples from patients who subsequently have high measures of disease activity versus samples from those same or different patients who have low subsequent disease activity.

[0734] 3. Samples from patients who subsequently have high measures of disease progression versus samples from those same or different patients who have low subsequent disease progression.

[0735] 4. Samples from patients who subsequently respond to a given medication or treatment regimen versus samples from those same or different patients who subsequently do not respond to a given medication or treatment regimen (for example, TNF pathway blocking medications).

[0736] 5. Samples from patients with a diagnosis of osteoarthritis versus patients with rheumatoid arthritis.

[0737] 6. Samples from patients with tissue biopsy results showing a high degree of inflammation versus samples from patients with lesser degrees of histological evidence of inflammation on biopsy.

[0738] Expression profiles correlating with progression of RA are identified. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures, that have predictive value for the progression of RA. Diagnostic nucleotide set(s) are identified which predict respond to TNF blockade. Patients are profiled before and during treatment with these medications. Patients are followed for relief of symptoms, side effects and progression of joint destruction, e.g., as measured by hand radiographs. Expression profiles correlating with response to TNF blockade are identified. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures that have predictive value for response to TNF blockade.

[0739] Cytomegalovirus is a very important cause of disease in immunocompromised patients, for example, transplant patients, cancer patients, and AIDS patients. The virus can cause inflammation and disease in almost any tissue (particularly the colon, lung, bone marrow and retina). It is increasingly important to identify patients with current or impending clinical CMV disease, particularly when immunosuppressive drugs are to be used in a patient, e.g. for preventing transplant rejection.

[0740] Leukocytes are profiled in patients with active CMV, impending CMV, or no CMV. Expression profiles correlating with diagnosis of active or impending CMV are identified. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures that have predictive value for the diagnosis of active or impending CMV. Diagnostic nucleotide set(s) identified with predictive value for the diagnosis of active or impending CMV may be combined, or used in conjunction with, cardiac, liver and/or kidney allograft-related diagnostic gene set(s) (described in Examples 12 and 24).

[0741] In addition, or alternatively, CMV nucleotide sequences are obtained, and a diagnostic nucleotide set is designed using CMV nucleotide sequence. The entire sequence of the organism is known and all CMV nucleotide sequences can be isolated and added to the library using the sequence information and the approach described below. Known expressed genes are preferred. Alternatively, nucleotide sequences are selected to represent groups of CMV genes that are coordinately expressed (immediate early genes, early genes, and late genes) (Spector et al. 1990, Stamminger et al. 1990).

[0742] Oligonucleotides were designed for CMV genes using the oligo design procedures of Example 21. Probes were designed using the 14 gene sequences shown here and were included on the array described in examples 20–22:

[0743]

Cytomegalovirus	HCMVTRL2	1893. . .2240
(CMV)	(IRL2)
Accession	HCMVTRL7	complement(6595. . .6843)
#X17403	(IRL7)
	HCMVUL21	complement(26497. . .27024)
	HCMVUL27	complement(32831. . .34657)
	HCMVUL33	43251. . .44423
	HCMVUL54	complement(76903. . .80631)
	HCMVUL75	complement(107901. . .110132)
	HCMVUL83	complement(119352. . .121037)
	HCMVUL106	complement(154947. . .155324)
	HCMVUL109	complement(157514. . .157810)
	HCMVUL113	161503. . .162800
	HCMVUL122	complement(169364. . .170599)
	HCMVUL123	complement(171006. . .172225)
	(last exon at 3′-end)
	HCMVUS28	219200. . .220171

[0744] Diagnostic nucleotide set(s) for expression of CMV genes is used in combination with diagnostic leukocyte nucleotide sets for diagnosis of other conditions, e.g. organ allograft rejection.

[0745] Using the techniques described in example 8 mononuclear samples from 180 cardiac transplant recipients (enrolled in the study described in Example 11) were used for expression profiling with the leukocyte arrays. Of these samples 15 were associated with patients who had a diagnosis of primary or reactivation CMV made by culture, PCR or any specific diagnostic test.

[0746] After preparation of RNA, amplification, labeling, hybridization, scanning, feature extraction and data processing were done as described in Example 23 using the oligonucleotide microarrays described in Examples 20–22.

[0747] The resulting log ratio of expression of Cy3 (patient sample)/Cy5 (R50 reference RNA) was used for analysis. Significance analysis for microarrays (SAM, Tusher 2001, see Example 26) was applied to determine which genes were most significantly differentially expressed between these 15 CMV patients and the 165 non-CMV patients (Table 11A). 12 genes were identified with a 0% FDR and 6 with a 0.1% FDR. Some genes are represented by more than one oligonucleotide on the array and for 2 genes, multiple oligonucleotides from the same gene are called significant (SEQ IDs: 5559, 6308: eomesodermin and 1685, 2428, 4113, 6059: small inducible cytokine A4).

[0748] Clinical variables were also included in the significance analysis. For example, the white blood cell count and the number of weeks post transplant (for the patient at the time the sample was obtained) were available for most of the 180 samples. The log of these variables was taken and the variables were then used in the significance analysis described above with the gene expression data. Both the white blood cell count (0.1% FDR) and the weeks post transplant (0% FDR) appeared to correlate with CMV status. CMV patients were more likely to have samples associated with later post transplant data and the lower white blood cell counts.

[0749] These genes and variables can be used alone or in association with other genes or variables or with other genes to build a diagnostic gene set or a classification algorithm using the approaches described herein.

[0750] Primers for real-time PCR validation were designed for some of these genes as described in Example 25 and listed in Table 11B and the sequence listing. Using the methods described in example 25, primers for Granzyme B were designed and used to validate expression findings from the arrays. 6 samples were tested (3 from patients with CMV and 3 from patients without CMV). The gene was found to be differentially expressed between the patients with and without CMV (see example 25 for full description). This same approach can be used to validate other diagnostic genes by real-time PCR.

[0751] Diagnostic nucleotide sets can also be identified for a variety of other viral diseases (Table 1) using this same approach.

[0752] Cytomegalovirus is a very important cause of disease in immunosuppressed patients, for example, transplant patients, cancer patients, and AIDS patients. The virus can cause inflammation and disease in almost any tissue (particularly the colon, lung, bone marrow and retina). It is increasingly important to identify patients with current or impending clinical CMV disease, particularly when immunosuppressive drugs are to be used in a patient, e.g. for preventing transplant rejection.

[0753] Leukocytes are profiled in patients with active CMV, impending CMV, or no CMV. Expression profiles correlating with diagnosis of active or impending CMV are identified. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures, that have predictive value for the diagnosis of active or impending CMV. Diagnostic nucleotide set(s) identified with predictive value for the diagnosis of active or impending CMV may be combined, or used in conjunction with, cardiac, liver and/or kidney allograft-related diagnostic gene set(s) (described in Examples 11 and 12).

[0754] In addition, or alternatively, CMV nucleotide sequences are obtained, and a diagnostic nucleotide set is designed using CMV nucleotide sequence. The entire sequence of the organism is known and all CMV nucleotide sequences can be isolated and added to the library using the sequence information and the approach described below. Known expressed genes are preferred. Alternatively, nucleotide sequences are selected to represent groups of CMV genes that are coordinately expressed (immediate early genes, early genes, and late genes) (Spector et al. 1990, Stamminger et al. 1990).

[0755] CMV nucleotide sequences were isolated as follows: Primers were designed to amplify known expressed CMV genes, based on the publically available sequence of CMV strain AD 169 (Genbank LOCUS: HEHCMVCG 229354 bp; DEFINITION Human cytomegalovirus strain AD169 complete genome; ACCESSION X17403; VERSION X17403.1 GI:59591). The following primer were used to PCR amplify nucleotide sequences from 175 ng of AD 169 viral genomic DNA (Advance Biotechnologies Incorporated) as a template:

[0756]

	CMV GENE	PRIMER SEQUENCES	SEQ. ID. NO:
	UL21 5′	atgtggccgcttctgaaaaac	8771
	UL21 3′	tcatggggtggggacgggg	8772
	UL33 5′	gtacgcgctgctgggtcatg	8773
	UL33 3′	tcataccccgctgaggttatg	8774
	UL54 5′	cacggacgacgacgctgacg	8775
	UL54 3′	gtacggcagaaaagccggctc	8776
	UL55 5′	caccaaagacacgtcgttacag	8777
	UL55 3′	tcagacgttctcttcttcgtcg	8778
	UL75 5′	cagcggcgctcaacatttcac	8779
	UL75 3′	tcagcatgtcttgagcatgcgg	8780
	UL80 5′	cctccccaactactactaccg	8781
	UL80 3′	ttactcgagcttattgagcgcag	8782
	UL83 5′	cacgtcgggcgttatgacac	8783
	UL83 3′	tcaacctcggtgctttttggg	8784
	UL97 5′	ctgtctgctcattctggcgg	8785
	UL97 3′	ttactcggggaacagttggcg	8786
	UL106 5′	atgatgaccgaccgcacgga	8787
	UL106 3′	tcacggtggctcgatacactg	8788
	UL107 5′	aagcttccttacagcataactgt	8789
	UL107 3′	ccttataacatgtattttgaaaaattg	8790
	UL109 5′	atgatacacgactaccactgg	8791
	UL109 3′	ttacgagcaagagttcatcacg	8792
	UL112 5′	ctgcgtgtcctcgctgggt	8793
	UL112 3′	tcacgagtccactcggaaagc	8794
	UL113 5′	ctcgtcttcttcggctccac	8795
	UL113 3′	ttaatcgtcgaaaaacgccgcg	8796
	UL122 5′	gatgcttgtaacgaaggcgtc	8797
	UL122 3′	ttactgagacttgttcctcagg	8798
	UL123 5′	gtagcctacactttggccacc	8799
	UL123 3′	ttactggtcagccttgcttcta	8800
	IRL2 5′	acgtccctggtagacggg	8801
	IRL2 3′	ttataagaaaagaagcacaagctc	8802
	IRL3 5′	atgtattgttttcttttttacagaaag	8803
	IRL3 3′	ttatattattatcaaaacgaaaaacag	8804
	IRL4 5′	cttctcctttccttaatctcgg	8805
	IRL4 3′	ctatacggagatcgcggtcc	8806
	IRL5 5′	atgcatacatacacgcgtgcat	8807
	IRL5 3′	ctaccatataaaaacgcagggg	8808
	IRL7 5′	atgaaagcaagaggcagccg	8809
	IRL7 3′	tcataaggtaacgatgctacttt	8810
	IRL13 5′	atggactggcgatttacggtt	8811
	IRL13 3′	ctacattgtgccatttctcagt	8812
	US2 5′	atgaacaatctctggaaagcctg	8813
	US2 3′	tcagcacacgaaaaaccgcatc	8814
	US3 5′	atgaagccggtgttggtgctc	8815
	US3 3′	ttaaataaatcgcagacgggcg	8816
	US6 5′	atggatctcttgattcgtctcg	8817
	US6 3′	tcaggagccacaacgtcgaatc	8818
	US11 5′	cgcaaaacgctactggctcc	8819
	US11 3′	tcaccactggtccgaaaacatc	8820
	US18 5′	tacggctggtccgtcatcgt	8821
	US18 3′	ttacaacaagctgaggagactc	8822
	US27 5′	atgaccacctctacaaataatcaaac	8823
	US27 3′	gtagaaacaagcgttgagtccc	8824
	US28 5′	cgttgcggtgtctcagtcg	8825
	US28 3′	tcatgctgtggtaccaggata	8826

[0757] The PCR reaction conditions were 10 mM Tris pH 8.3, 3.5 mM MgCl2, 25 mM KCl, 200 uM dNTP's, 0.2 uM primers, and 5 Units of Taq Gold. The cycle parameters were as follows:

[0758] 1. 95° C. for 30 sec

[0759] 2. 95° C. for 15 sec

[0760] 3. 56° C. for 30 sec

[0761] 4. 72° C. for 2 min

[0762] 5. go to step 2, 29 times

[0763] 6. 72° C. for 2 min

[0764] 7. 4° C. forever

[0765] PCR products were gel purified, and DNA was extracted from the agarose using the QiaexII gel purification kit (Qiagen). PCR product was ligated into the T/A cloning vector p-GEM-T-Easy (Promega) using 3 ul of gel purified PCR product and following the Promega protocol. The products of the ligation reaction were transformed and plated as described in the p-GEM protocol. White colonies were picked and grow culture in LB-AMP medium. Plasmid was prepared from these cultures using Qiagen Miniprep kit (Qiagen). Restriction enzyme digested plasmid (Not I and EcoRI) was examined after agarose gel electrophoresis to assess insert size. When the insert was the predicted size, the plasmid was sequenced by well-known techniques to confirm the identity of the CMV gene. Using forward and reverse primers that are complimentary to sequences flanking the insert cloning site (M13F and M13R), the isolated CMV gene was amplified and purified as described above. Amplified cDNAs were used to create a microarray as described above. In addition, 50mer oligonucleotides corresponding the CMV genes listed above were designed, synthesized and placed on a microarray using methods described elsewhere in the specification.

[0766] Alternatively, oligonucleotide sequences a redesigned and synthesized for oligonucleotide array expression analysis from CMV genes as described in examples 20–22.

[0767] Diagnostic nucleotide set(s) for expression of CMV genes is used in combination with diagnostic leukocyte nucleotide sets for diagnosis of other conditions, e.g. organ allograft rejection.

[0768] HMG-CoA reductase inhibitors, called “Statins,” are very effective in preventing complications of coronary artery disease in either patients with coronary disease and high cholesterol (secondary prevention) or patients without known coronary disease and with high cholesterol (primary prevention). Examples of Statins are (generic names given) pravastatin, atorvastatin, and simvastatin. Monitoring response to Statin therapy is of interest. Patients are identified who are on or are about to start Statin therapy.

[0769] Leukocytes are profiled in patients before and after initiation of therapy, or in patients already being treated with Statins. Data is collected corresponding to cholesterol level, markers of inflammation (e.g., C-Reactive Protein and the Erythrocyte Sedimentation Rate), measures of endothelial function (e.g., improved forearm resistance or coronary flow reserve) and clinical endpoints (new stable angina, unstable angina, myocardial infarction, ventricular arrhythmia, claudication). Patient groups can be defined based on their response to Statin therapy (cholesterol, clinical endpoints, endothelial function). Expression profiles correlating with response to Statin treatment are identified. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, according to the above procedures, that have predictive value for the response to Statins. Members of candidate nucleotide sets with expression that is altered by Statins are disease target nucleotides sequences.

[0770] This Example describes the compilation of almost 8,000 unique genes and ESTs using sequences identified from the sources described below. The sequences of these genes and ESTs were used to design probes, as described in the following Example.

[0771] Tables 3A, 3B and 3C list the sequences identified in the subtracted leukocyte expression libraries. All sequences that were identified as corresponding to a known RNA transcript were represented at least once, and all unidentified sequences were represented twice—once by the sequence on file and again by the complementary sequence—to ensure that the sense (or coding) strand of the gene sequence was included.

[0772] Table 3A. Table 3A contained all those sequences in the subtracted libraries of example 1 that matched sequences in GenBank's nr, EST_Human, and UniGene databases with an acceptable level of confidence. All the entries in the table representing the sense strand of their genes were grouped together and all those representing the antisense strand were grouped. A third group contained those entries whose strand could not be determined. Two complementary probes were designed for each member of this third group.

[0773] Table 3B and 3C. Table 3B and 3C contained all those sequences in the leukocyte expression subtracted libraries of example 1 that did not match sequences in GenBank's nr, EST_Human, and UniGene databases with an acceptable level of confidence, but which had a high probability of representing real mRNA sequences. Sequences in Table 3B did not match anything in the databases above but matched regions of the human genome draft and were spatially clustered along it, suggesting that they were exons, rather than genomic DNA included in the library by chance. Sequences in Table 3C also aligned well to regions of the human genome draft, but the aligned regions were interrupted by genomic DNA, meaning they were likely to be spliced transcripts of multiple exon genes.

[0774] Table 3B lists 510 clones and Table 3C lists 48 clones that originally had no similarity with any sequence in the public databases. Blastn searches conducted after the initial filing have identified sequences in the public database with high similarity (E values less than 1e−40) to the sequences determined for these clones. Table 3B contained 272 clones and Table 3C contained 25 clones that were found to have high similarity to sequences in dbEST. The sequences of the similar dbEST clones were used to design probes. Sequences from clones that contained no similar regions to any sequence in the database were used to design a pair of complementary probes.

[0775] Probes were designed from database sequences that had the highest similarity to each of the sequenced clones in Tables 3A, 3B, and 3C. Based on BLASTn searches the most similar database sequence was identified by locus number and the locus number was submitted to GenBank using batch Entrez located at the web site ncbi.nlm.nih.gov/entrez/batchentrez.cgi?db=Nucleotide to obtain the sequence for that locus. The GenBank entry sequence was used because in most cases it was more complete or was derived from multi-pass sequencing and thus would likely have fewer errors than the single pass cDNA library sequences. When only UniGene cluster IDs were available for genes of interest, the respective sequences were extracted from the UniGene_unique database, build 137, downloaded from NCBI (ftp://ncbi.nlm.nih.gov/repository/UniGene/). This database contains one representative sequence for each cluster in UniGene.

[0776]

Summary of library clones used in array probe design
Table	Sense Strand	Antisnese Strand	Strand Undetermined
3A	3621	763	124
3B	142	130	238
3C	19	6	23
Totals	3782	899	385

Literature Searches

[0777] Example 2 describes searches of literature databases. We also searched for research articles discussing genes expressed only in leukocytes or involved in inflammation and particular disease conditions, including genes that were specifically expressed or down-regulated in a disease state. Searches included, but were not limited to, the following terms and various combinations of theses terms: inflammation, atherosclerosis, rheumatoid arthritis, osteoarthritis, lupus, SLE, allograft, transplant, rejection, leukocyte, monocyte, lymphocyte, mononuclear, macrophage, neutrophil, eosinophil, basophil, platelet, congestive heart failure, expression, profiling, microarray, inflammatory bowel disease, asthma, RNA expression, gene expression, granulocyte.

[0778] A UniGene cluster ID or GenBank accession number was found for each gene in the list. The strand of the corresponding sequence was determined, if possible, and the genes were divided into the three groups: sense (coding) strand, anti-sense strand, or strand unknown. The rest of the probe design process was carried out as described above for the sequences from the leukocyte subtracted expression library.

Database Mining

[0779] Database mining was performed as described in Example 2. In addition, the Library Browser at the NCBI UniGene web site ncbi.nlm.nih.gov/UniGene/lbrowse.cgi?ORG=Hs&DISPLAY=ALL was used to identify genes that are specifically expressed in leukocyte cell populations. All expression libraries available at the time were examined and those derived from leukocytes were viewed individually. Each library viewed through the Library Browser at the UniGene web site contains a section titled “Shown below are UniGene clusters of special interest only” that lists genes that are either highly represented or found only in that library. Only the genes in this section were downloaded from each library. Alternatively, every sequence in each library is downloaded and then redundancy between libraries is reduced by discarding all UniGene cluster IDs that are represented more than once. A total of 439 libraries were downloaded, containing 35,819 genes, although many were found in more than one library. The most important libraries from the remaining set were separated and 3,914 genes remained. After eliminating all redundancy between these libraries and comparing the remaining genes to those listed in Tables 3A, 3B and 3C, the set was reduced to 2,573 genes in 35 libraries as shown in Table 9. From these, all genes in first 30 libraries were used to design probes. A random subset of genes was used from Library Lib.376, “Activated_T-cells_XX”. From the last four libraries, a random subset of sequences listed as “ESTs, found only in this library” was used.

Angiogenesis Markers

[0780] 215 sequences derived from an angiogenic endothelial cell subtracted cDNA library obtained from Stanford University were used for probe design. Briefly, using well known subtractive hybridization procedures, (as described in, e.g., U.S. Pat. Nos. 5,958,738; 5,589,339; 5,827,658; 5,712,127; 5,643,761; 5,565,340) modified to normalize expression by suppressing over-representation of abundant RNA species while increasing representation of rare RNA species, a library was produced that is enriched for RNA species (messages) that are differentially expressed between test (stimulated) and control (resting) HUVEC populations. The subtraction/suppression protocol was performed as described by the kit manufacturer (Clontech, PCR-select cDNA Subtraction Kit).

[0781] Pooled primary HUVECs (Clonetics) were cultured in 15% FCS, M199 (GibcoBRL) with standard concentrations of Heparin, Penicillin, Streptomycin, Glutamine and Endothelial Cell Growth Supplement. The cells were cultured on 1% gelatin coated 10 cm dishes. Confluent HUVECs were photographed under phase contrast microscopy.

[0782] The cells formed a monolayer of flat cells without gaps. Passage 2–5 cells were used for all experiments. Confluent HUVECs were treated with trypsin/EDTA and seeded onto collagen gels. Collagen gels were made according to the protocol of the Collagen manufacturer (Becton Dickinson Labware). Collagen gels were prepared with the following ingredients: Rat tail collagen type I (Collaborative Biomedical) 1.5 mg/mL, mouse laminin (Collaborative Biomedical) 0.5 mg/mL, 10% 10× media 199 (Gibco BRL). 1N NaOH, 10×PBS and sterile water were added in amounts recommended in the protocol. Cell density was measured by microscopy. 1.2×10⁶ cells were seeded onto gels in 6-well, 35 mm dishes, in 5% FCS M199 media. The cells were incubated for 2 hrs at 37 C with 5% CO2. The media was then changed to the same media with the addition of VEGF (Sigma) at 30 ng/mL media. Cells were cultured for 36 hrs. At 12, 24 and 36 hrs, the cells were observed with phase contrast microscopy. At 36 hours, the cells were observed elongating, adhering to each other and forming lumen structures. At 12 and 24 hrs media was aspirated and refreshed. At 36 hrs, the media was aspirated, the cells were rinsed with PBS and then treated with Collagenase (Sigma) 2.5 mg/mL PBS for 5 min with active agitation until the collagen gels were liquefied. The cells were then centrifuged at 4 C, 2000 g for 10 min. The supernatant was removed and the cells were lysed with 1 mL Trizol Reagent (Gibco) per 5×10⁶ cells. Total RNA was prepared as specified in the Trizol instructions for use. mRNA was then isolated as described in the micro-fast track mRNA isolation protocol from Invitrogen. This RNA was used as the tester RNA for the subtraction procedure.

[0783] Ten plates of resting, confluent, p4 HUVECs, were cultured with 15% FCS in the M199 media described above. The media was aspirated and the cells were lysed with 1 mL Trizol and total RNA was prepared according to the Trizol protocol. mRNA was then isolated according to the micro-fast track mRNA isolation protocol from Invitrogen. This RNA served as the control RNA for the subtraction procedure.

[0784] The entire subtraction cloning procedure was carried out as per the user manual for the Clontech PCR Select Subtraction Kit. The cDNAs prepared from the test population of HUVECs were divided into “tester” pools, while cDNAs prepared from the control population of HUVECs were designated the “driver” pool. cDNA was synthesized from the tester and control RNA samples described above. Resulting cDNAs were digested with the restriction enzyme RsaI. Unique double-stranded adapters were ligated to the tester cDNA. An initial hybridization was performed consisting of the tester pools of cDNA (with its corresponding adapter) and an excess of the driver cDNA. The initial hybridization results in a partial normalization of the cDNAs such that high and low abundance messages become more equally represented following hybridization due to a failure of driver/tester hybrids to amplify.

[0785] A second hybridization involved pooling unhybridized sequences from the first hybridization together with the addition of supplemental driver cDNA. In this step, the expressed sequences enriched in the two tester pools following the initial hybridization can hybridize. Hybrids resulting from the hybridization between members of each of the two tester pools are then recovered by amplification in a polymerase chain reaction (PCR) using primers specific for the unique adapters. Again, sequences originating in a tester pool that form hybrids with components of the driver pool are not amplified.

[0786] Hybrids resulting between members of the same tester pool are eliminated by the formation of “panhandles” between their common 5′ and 3′ ends. The subtraction was done in both directions, producing two libraries, one with clones that are upregulated in tube-formation and one with clones that are down-regulated in the process.

[0787] The resulting PCR products representing partial cDNAs of differentially expressed genes were then cloned (i.e., ligated) into an appropriate vector according to the manufacturer's protocol (pGEM-Teasy from Promega) and transformed into competent bacteria for selection and screening. Colonies (2180) were picked and cultured in LB broth with 50 ug/mL ampicillin at 37 C overnight. Stocks of saturated LB+50 ug/mL ampicillin and 15% glycerol in 96-well plates were stored at −80 C. Plasmid was prepared from 1.4 mL saturated LB broth containing 50 ug/mL ampicillin. This was done in a 96 well format using commercially available kits according to the manufacturer's recommendations (Qiagen 96-turbo prep).

[0788] 2 probes to represent 22 of these sequences required, therefore, a total of 237 probes were derived from this library.

Viral Genes

[0789] Several viruses may play a role in a host of disease including inflammatory disorders, atherosclerosis, and transplant rejection. Table 10 lists the viral genes represented by oligonucleotide probes on the microarray. Low-complexity regions in the sequences were masked using RepeatMasker before using them to design probes.

Strand Selection

[0790] It was necessary to design sense oligonucleotide probes because the labeling and hybridization protocol to be used with the microarray results in fluorescently-labeled antisense cRNA. All of the sequences we selected to design probes could be divided into three categories:

[0791] (1) Sequences known to represent the sense strand

[0792] (2) Sequences known to represent the antisense strand

[0793] (3) Sequences whose strand could not be easily determined from their descriptions

[0794] It was not known whether the sequences from the leukocyte subtracted expression library were from the sense or antisense strand. GenBank sequences are reported with sequence given 5′ to 3′, and the majority of the sequences we used to design probes came from accession numbers with descriptions that made it clear whether they represented sense or antisense sequence. For example, all sequences containing “mRNA” in their descriptions were understood to be the sequences of the sense mRNA, unless otherwise noted in the description, and all IMAGE Consortium clones are directionally cloned and so the direction (or sense) of the reported sequence can be determined from the annotation in the GenBank record.

[0795] For accession numbers representing the sense strand, the sequence was downloaded and masked and a probe was designed directly from the sequence. These probes were selected as close to the 3′ end as possible. For accession numbers representing the antisense strand, the sequence was downloaded and masked, and a probe was designed complementary to this sequence. These probes were designed as close to the 5′ end as possible (i.e., complementary to the 3′ end of the sense strand).

Minimizing Probe Redundancy.

[0796] Multiple copies of certain genes or segments of genes were included in the sequences from each category described above, either by accident or by design. Reducing redundancy within each of the gene sets was necessary to maximize the number of unique genes and ESTs that could be represented on the microarray.

[0797] Three methods were used to reduce redundancy of genes, depending on what information was available. First, in gene sets with multiple occurrences of one or more UniGene numbers, only one occurrence of each UniGene number was kept. Next, each gene set was searched by GenBank accession numbers and only one occurrence of each accession number was conserved. Finally, the gene name, description, or gene symbol were searched for redundant genes with no UniGene number or different accession numbers. In reducing the redundancy of the gene sets, every effort was made to conserve the most information about each gene.

[0798] We note, however, that the UniGene system for clustering submissions to GenBank is frequently updated and UniGene cluster IDs can change. Two or more clusters may be combined under a new cluster ID or a cluster may be split into several new clusters and the original cluster ID retired. Since the lists of genes in each of the gene sets discussed were assembled at different times, the same sequence may appear in several different sets with a different UniGene ID in each.

[0799] Sequences from Table 3A were treated differently. In some cases, two or more of the leukocyte subtracted expression library sequences aligned to different regions of the same GenBank entry, indicating that these sequences were likely to be from different exons in the same gene transcript. In these cases, one representative library sequence corresponding to each presumptive exon was individually listed in Table 3A.

Compilation.

[0800] After redundancy within a gene set was sufficiently reduced, a table of approximately 8,000 unique genes and ESTs was compiled in the following manner. All of the entries in Table 3A were transferred to the new table. The list of genes produced by literature and database searches was added, eliminating any genes already contained in Table 3A. Next, each of the remaining sets of genes was compared to the table and any genes already contained in the table were deleted from the gene sets before appending them to the table.

[0801]

	Probes
Subtracted Leukocyte Expression Library
Table 3A	4,872
Table 3B	796
Table 3C	85
Literature Search Results	494
Database Mining	1,607
Viral genes
CMV	14
EBV	6
HHV 6	14
Adenovirus	8
Angiogenesis markers: 215, 22 of which needed two probes	237
Arabidopsis thaliana genes	10
Total sequences used to design probes	8,143

[0802] By way of example, this section describes the design of four oligonucleotide probes using Array Designer Ver 1.1 (Premier Biosoft International, Palo Alto, Calif.). The major steps in the process are given first.

[0803] Obtain best possible sequence of mRNA from GenBank. If a full-length sequence reference sequence is not available, a partial sequence is used, with preference for the 3′ end over the 5′ end. When the sequence is known to represent the antisense strand, the reverse complement of the sequence is used for probe design. For sequences represented in the subtracted leukocyte expression library that have no significant match in GenBank at the time of probe design, our sequence is used.

[0804] Mask low complexity regions and repetitive elements in the sequence using an algorithm such as RepeatMasker.

[0805] Use probe design software, such as Array Designer, version 1.1, to select a sequence of 50 residues with specified physical and chemical properties. The 50 residues nearest the 3′ end constitute a search frame. The residues it contains are tested for suitability. If they don't meet the specified criteria, the search frame is moved one residue closer to the 5′ end, and the 50 residues it now contains are tested. The process is repeated until a suitable 50-mer is found.

[0806] If no such 50-mer occurs in the sequence, the physical and chemical criteria are adjusted until a suitable 50-mer is found.

[0807] Compare the probe to dbEST, the UniGene cluster set, and the assembled human genome using the BLASTn search tool at NCBI to obtain the pertinent identifying information and to verify that the probe does not have significant similarity to more than one known gene.

Clone 40H12

[0808] Clone 40H12 was sequenced and compared to the nr, dbEST, and UniGene databases at NCBI using the BLAST search tool. The sequence matched accession number NM_—002310, a ‘curated RefSeq project’ sequence, see Pruitt et al. (2000) Trends Genet. 16:44–47, encoding leukemia inhibitory factor receptor (LIFR) mRNA with a reported E value of zero. An E value of zero indicates there is, for all practical purposes, no chance that the similarity was random based on the length of the sequence and the composition and size of the database. This sequence, cataloged by accession number NM_—002310, is much longer than the sequence of clone 40H12 and has a poly-A tail. This indicated that the sequence cataloged by accession number NM_—002310 is the sense strand and a more complete representation of the mRNA than the sequence of clone 40H12, especially at the 3′ end. Accession number “NM_—002310” was included in a text file of accession numbers representing sense strand mRNAs, and sequences for the sense strand mRNAs were obtained by uploading a text file containing desired accession numbers as an Entrez search query using the Batch Entrez web interface and saving the results locally as a FASTA file. The following sequence was obtained, and the region of alignment of clone 40H12 is outlined:

[0809]

[0810] The FASTA file, including the sequence of NM_—002310, was masked using the RepeatMasker web interface (Smit, A F A & Green, P RepeatMasker at http://ftp.genome.washington.edu/RM/RepeatMasker.html, Smit and Green).

[0811] Specifically, during masking, the following types of sequences were replaced with “N's”: SINE/MIR & LINE/L2, LINE/L1, LTR/MaLR, LTR/Retroviral, Alu, and other low informational content sequences such as simple repeats. Below is the sequence following masking:

[0812]

[0813] The length of this sequence was determined using batch, automated computational methods and the sequence, as sense strand, its length, and the desired location of the probe sequence near the 3′ end of the mRNA was submitted to Array Designer Ver 1.1 (Premier Biosoft International, Palo Alto, Calif.). Search quality was set at 100%, number of best probes set at 1, length range set at 50 base pairs, Target Tm set at 75 C. degrees plus or minus 5 degrees, Hairpin max deltaG at 6.0-kcal/mol., Self dimmer max deltaG at 6.0-kcal/mol, Run/repeat (dinucleotide) max length set at 5, and Probe site minimum overlap set at 1. When none of the 49 possible probes met the criteria, the probe site would be moved 50 base pairs closer to the 5′ end of the sequence and resubmitted to Array Designer for analysis. When no possible probes met the criteria, the variation on melting temperature was raised to plus and minus 8 degrees and the number of identical basepairs in a run increased to 6 so that a probe sequence was produced.

[0814] In the sequence above, using the criteria noted above, Array Designer Ver 1.1 designed a probe corresponding to oligonucleotide number 2280 in Table 8 and is indicated by underlining in the sequence above. It has a melting temperature of 68.4 degrees Celsius and a max run of 6 nucleotides and represents one of the cases where the criteria for probe design in Array Designer Ver 1.1 were relaxed in order to obtain an oligonucleotide near the 3′ end of the mRNA (Low melting temperature was allowed).

Clone 463D12

[0815] Clone 463D12 was sequenced and compared to the nr, dbEST, and UniGene databases at NCBI using the BLAST search tool. The sequence matched accession number AI184553, an EST sequence with the definition line “qd60a05.x1 Soares_testis_NHT Homo sapiens cDNA clone IMAGE:1733840 3′ similar to gb:M29550 PROTEIN PHOSPHATASE 2B CATALYTIC SUBUNIT 1 (HUMAN); mRNA sequence.” The E value of the alignment was 1.00×10⁻¹¹⁸. The GenBank sequence begins with a poly-T region, suggesting that it is the antisense strand, read 5′ to 3′. The beginning of this sequence is complementary to the 3′ end of the mRNA sense strand. The accession number for this sequence was included in a text file of accession numbers representing antisense sequences. Sequences for antisense strand mRNAs were obtained by uploading a text file containing desired accession numbers as an Entrez search query using the Batch Entrez web interface and saving the results locally as a FASTA file. The following sequence was obtained, and the region of alignment of clone 463D12 is outlined:

[0816]

[0817] The FASTA file, including the sequence of AA184553, was then masked using the RepeatMasker web interface, as shown below. The region of alignment of clone 463D12 is outlined.

[0818]

[0819] The sequence was submitted to Array Designer as described above, however, the desired location of the probe was indicated at base pair 50 and if no probe met the criteria, moved in the 3′ direction. The complementary sequence from Array Designer was used, because the original sequence was antisense. The oligonucleotide designed by Array Designer corresponds to oligonucleotide number 4342 in Table 8 and is complementary to the underlined sequence above. The probe has a melting temperature of 72.7 degrees centigrade and a max run of 4 nucleotides.

Clone 72D4

[0820] Clone 72D4 was sequenced and compared to the nr, dbEST, and UniGene databases at NCBI using the BLAST search tool. No significant matches were found in any of these databases. When compared to the human genome draft, significant alignments were found to three consecutive regions of the reference sequence NT_—008060, as depicted below, suggesting that the insert contains three spliced exons of an unidentified gene.

[0821]

Residue numbers on Matching residue
	clone 72D4 sequence	numbers on NT_008060
	1–198	478646–478843
	197–489	479876–480168
	491–585	489271–489365

[0822] Because the reference sequence contains introns and may represent either the coding or noncoding strand for this gene, BioCardia's own sequence file was used to design the oligonucleotide. Two complementary probes were designed to ensure that the sense strand was represented. The sequence of the insert in clone 72D4 is shown below, with the three putative exons outlined.

[0823]

[0824] The sequence was submitted to RepeatMasker, but no repetitive sequences were found. The sequence shown above was used to design the two 50-mer probes using Array Designer as described above. The probes are shown in bold typeface in the sequence depicted below. The probe in the sequence is oligonucleotide number 6415 (SEQ ID NO: 6415) in Table 8 and the complementary probe is oligonucleotide number 6805 (SEQ ID NO:6805).

[0825]

CAGGTCACACAGCACATCAGTGGCTACATGTGAGCTCAGACCTGGGTCTGCTGCTGTCT
GTCTTCCCAATATCCATGACCTTGACTGATGCAGGTGTCTAGGGATACGTCCATCCCCG
TCCTGCTGGAGCCCAGAGCACGGAAGCCTGGCCCTCCGAGGAGACAGAAGGGAGTGTCG
GACACCATGACGAGAGCTTGGCAGAATAAATAACTTCTTTAAACAATTTTACGGCATGA
AGAAATCTGGACCAGTTTATTAAATGGGATTTCTGCCACAAACCTTGGAAGAATCACAT
CATCTTANNCCCAAGTGAAAACTGTGTTGCGTAACAAAGAACATGACTGCGCTCCACAC
ATACATCATTGCCCGGCGAGGCGGGACACAAGTCAACGACGGAACACTTGAGACAGGCC
TACAACTGTGCACGGGTCAGAAGCAAGTTTAAGCCATACTTGCTGCAGTGAGACTACAT
TTCTGTCTATAGAAGATACCTGACTTGATCTGTTTTTCAGCTCCAGTTCCCAGATGTGC
←----GTCAAGGGTCTACACG
GTGTTGTGGTCCCCAAGTATCACCTTCCAATTTCTGGGAG--→
CACAACACCAGGGGTTCATAGTGGAAGGTTAAAG-5′
CAGTGCTCTGGCCGGATCCTTGCCGCGCGGATAAAAACT---→

Confirmation of Probe Sequence

[0826] Following probe design, each probe sequence was confirmed by comparing the sequence against dbEST, the UniGene cluster set, and the assembled human genome using BLASTn at NCBI. Alignments, accession numbers, gi numbers, UniGene cluster numbers and names were examined and the most common sequence used for the probe. The final probe set was compiled into Table 8.

[0827] We produced an array of 8000 spotted 50mer oligonucleotides. Examples 20 and 21 exemplify the design and selection of probes for this array.

[0828] Sigma-Genosys (The Woodlands, Tex.) synthesized unmodified 50-mer oligonucleotides using standard phosphoramidite chemistry, with a starting scale of synthesis of 0.05 μmole (see, e.g., R. Meyers, ed. (1995) Molecular Biology and Biotechnology: A Comprehensive Desk Reference). Briefly, to begin synthesis, a 3′ hydroxyl nucleoside with a dimethoxytrityl (DMT) group at the 5′ end was attached to a solid support. The DMT group was removed with trichloroacetic acid (TCA) in order to free the 5′-hydroxyl for the coupling reaction. Next, tetrazole and a phosphoramidite derivative of the next nucleotide were added. The tetrazole protonates the nitrogen of the phosphoramidite, making it susceptible to nucleophilic attack. The DMT group at the 5′-end of the hydroxyl group blocks further addition of nucleotides in excess. Next, the inter-nucleotide linkage was converted to a phosphotriester bond in an oxidation step using an oxidizing agent and water as the oxygen donor. Excess nucleotides were filtered out and the cycle for the next nucleotide was started by the removal of the DMT protecting group. Following the synthesis, the oligo was cleaved from the solid support. The oligonucleotides were desalted, resuspended in water at a concentration of 100 or 200 μM, and placed in 96-deep well format. The oligonucleotides were re-arrayed into Whatman Uniplate 384-well polyproylene V bottom plates. The oligonucleotides were diluted to a final concentration 30 μM in 1× Micro Spotting Solution Plus (Telechem/arrayit.com, Sunnyvale, Calif.) in a total volume of 15 μl. In total, 8,031 oligonucleotides were arrayed into twenty-one 384-well plates.

[0829] Arrays were produced on Telechem/arrayit.com Super amine glass substrates (Telechem/arrayit.com), which were manufactured in 0.1 mm filtered clean room with exact dimensions of 25×76×0.96 mm. The arrays were printed using the Virtek Chipwriter with a Telechem 48 pin Micro Spotting Printhead. The Printhead was loaded with 48 Stealth SMP3B TeleChem Micro Spotting Pins, which were used to print oligonucleotides onto the slide with the spot size being 110–115 microns in diameter.

Amplification, Labeling, Hybridization and Scanning

[0830] Samples consisting of at least 2 μg of intact total RNA were further processed for array hybridization. Amplification and labeling of total RNA samples was performed in three successive enzymatic reactions. First, a single-stranded DNA copy of the RNA was made (hereinafter, “ss-cDNA”). Second, the ss-cDNA was used as a template for the complementary DNA strand, producing double-stranded cDNA (hereinafter, “ds-cDNA, or cDNA”). Third, linear amplification was performed by in vitro transcription from a bacterial T₇ promoter. During this step, fluorescent-conjugated nucleotides were incorporated into the amplified RNA (hereinafter, “aRNA”).

[0831] The first strand cDNA was produced using the Invitrogen kit (Superscript II). The first strand cDNA was produced in a reaction composed of 50 mM Tris-HCl (pH 8.3), 75 mM KCl, and 3 mM MgCl₂ (1× First Strand Buffer, Invitrogen), 0.5 mM dGTP, 0.5 mM dATP, 0.5 mM dTTP, 0.5 mM dCTP, 10 mM DTT, 10 U reverse transcriptase (Superscript II, Invitrogen, #18064014), 15 U RNase inhibitor (RNAGuard, Amersham Pharmacia, #27-0815-01), 5 μM T7T24 primer (5′-GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGGTTTTTTTTTTTT TTTTTTTTTTTT-3′), (SEQ ID NO:8831) and 2 μg of selected sample total RNA. Several purified, recombinant control mRNAs from the plant Arabidopsis thaliana were added to the reaction mixture: 2–20 pg of the following genes CAB, RCA, LTP4, NAC1, RCP1,XCP2, RBCL, LTP6, TIM, and PRKase (Stratagene, #252201, #252202, #252204, #252208, #252207, #252206, #252203, #252205, #252209, #252210 respectively). The control RNAs allow the estimate of copy numbers for individual mRNAs in the clinical sample because corresponding sense oligonucleotide probes for each of these plant genes are present on the microarray. The final reaction volume of 20 μl was incubated at 42° C. for 60 min.

[0832] For synthesis of the second cDNA strand, DNA polymerase and RNase were added to the previous reaction, bringing the final volume to 150 μl. The previous contents were diluted and new substrates were added to a final concentration of 20 mM Tris-HCl (pH 7.0) (Fisher Scientific, Pittsburgh, Pa. #BP1756-100), 90 mMKCl (Teknova, Half Moon Bay, Calif., #0313-500), 4.6 mM MgCl₂ (Teknova, Half Moon Bay, Calif., #0304-500), 10 mM(NH₄)₂SO₄ (Fisher Scientific #A702-500)(1× Second Strand buffer, Invitrogen), 0.266 mM dGTP, 0.266 mM dATP, 0.266 mM dTTP, 0.266 mM dCTP, 40 U E. coli DNA polymerase (Invitrogen, #18010-025), and 2 U RNaseH (Invitrogen, #18021-014). The second strand synthesis took place at 16° C. for 120 minutes.

[0833] Following second-strand synthesis, the ds-cDNA was purified from the enzymes, dNTPs, and buffers before proceeding to amplification, using phenol-chloroform extraction followed by ethanol precipitation of the cDNA in the presence of glycogen.

[0834] Alternatively, a silica-gel column is used to purify the cDNA (e.g. Qiaquick PCR cleanup from Qiagen, #28104). The cDNA was collected by centrifugation at >10,000×g for 30 minutes, the supernatant is aspirated, and 150 μl of 70% ethanol, 30% water was added to wash the DNA pellet. Following centrifugation, the supernatant was removed, and residual ethanol was evaporated at room temperature.

[0835] Linear amplification of the cDNA was performed by in vitro transcription of the cDNA. The cDNA pellet from the step described above was resuspended in 7.4 μl of water, and in vitro transcription reaction buffer was added to a final volume of 20 μl containing 7.5 mM GTP, 7.5 mM ATP, 7.5 mM TTP, 2.25 mM CTP, 1.025 mM Cy3-conjugated CTP (Perkin Elmer; Boston, Mass., #NEL-580), 1× reaction buffer (Ambion, Megascript Kit, Austin, Tex. and #1334) and 1% T₇ polymerase enzyme mix (Ambion, Megascript Kit, Austin, Tex. and #1334). This reaction was incubated at 37° C. overnight. Following in vitro transcription, the RNA was purified from the enzyme, buffers, and excess NTPs using the RNeasy kit from Qiagen (Valencia, Calif.; # 74106) as described in the vendor's protocol. A second elution step was performed and the two eluates were combined for a final volume of 60 μl. RNA is quantified using an Agilent 2100 bioanalyzer with the RNA 6000 nano LabChip.

[0836] Reference RNA was prepared as described above, except Cy5-CTP was incorporated instead of Cy3CTP. Reference RNA from five reactions, each reaction started with 2 ug total RNA, was pooled together and quantitated as described above.

Hybridization to an Array

[0837] RNA was prepared for hybridization as follows: for an 18 mm×55 mm array, 20 μg of amplified RNA (aRNA) was combined with 20 μg of reference aRNA. The combined sample and reference aRNA was concentrated by evaporating the water to 10 μl in a vacuum evaporator. The sample was fragmented by heating the sample at 95° C. for 30 minutes to fragment the RNA into 50–200 bp pieces. Alternatively, the combined sample and reference aRNA was concentrated by evaporating the water to 5 μl in a vacuum evaporator. Five μl of 20 mM zinc acetate was added to the aRNA and the mix incubated at 60° C. for 10 minutes. Following fragmentation, 40 μl of hybridization buffer was added to achieve final concentrations of 5×SSC and 0.20% SDS with 0.1 μg/ul of Cot-1 DNA (Invitrogen) as a competitor DNA. The final hybridization mix was heated to 98° C., and then reduced to 50° C. at 0.1° C. per second.

[0838] Alternatively, formamide is included in the hybridization mixture to lower the hybridization temperature.

[0839] The hybridization mixture was applied to a pre-heated 65° C. microarray, surface, covered with a glass coverslip (Corning, #2935-246), and placed on a pre-heated 65° C. hybridization chamber (Telechem, AHC-10). 15 ul of 5×SSC was placed in each of the reservoir in the hybridization chamber and the chamber was sealed and placed in a water bath at 62° C. for overnight (16–20 hrs). Following incubation, the slides were washed in 2×SSC, 0.1% SDS for five minutes at 30° C., then in 2×SSC for five minutes at 30° C., then in 2×SSC for another five minutes at 30° C., then in 0.2×SSC for two minutes at room temperature. The arrays were spun at 1000×g for 2 minutes to dry them. The dry microarrays are then scanned by methods described above.

[0840] The microarrays were imaged on the Agilent (Palo Alto, Calif.) scanner G2565AA. The scan settings using the Agilent software were as follows: for the PMT Sensitivity (100% Red and 100% Green); Scan Resolution (10 microns); red and green dye channels; used the default scan region for all slides in the carousel; using the largest scan region; scan date for Instrument ID; and barcode for Slide ID. The full image produced by the Agilent scanner was flipped, rotated, and split into two images (one for each signal channel) using TIFFSplitter (Agilent, Palo Alto, Calif.). The two channels are the output at 532 nm (Cy3-labeled sample) and 633 nm (Cy5-labeled R50). The individual images were loaded into GenePix 3.0 (Axon Instruments, Union City, Calif.) for feature extraction, each image was assigned an excitation wavelength corresponding the file opened; Red equals 633 nm and Green equals 532 nm. The setting file (gal) was opened and the grid was laid onto the image so that each spot in the grid overlapped with >50% of the feature. Then the GenePix software was used to find the features without setting minimum threshold value for a feature. For features with low signal intensity, GenePix reports “not found”. For all features, the diameter setting was adjusted to include only the feature if necessary. The GenePix software determined the median pixel intensity for each feature (F_i) and the median pixel intensity of the local background for each feature (B_i) in both channels. The standard deviation (SDF_{i and} SDB_i) for each is also determined. Features for which GenePix could not discriminate the feature from the background were “flagged” as described below.

[0841] Following feature extraction into a .gpr file, the header information of the .gpr file was changed to carry accurate information into the database. An Excel macro was written to change the headers. The steps in that macro were:

[0842] Open .gpr file.

[0843] Check the value in the first row, first column. If it is “ATF”, then the header has likely already been reformatted. The file is skipped and the user is alerted. Otherwise, proceed through the following steps.

[0844] Store the following values in variables.

[0845] Name of .tif image file: parsed from row 11.

[0846] SlideID: parsed from name of .tif image file.

[0847] Version of the feature extraction software: parsed from row 25

[0848] GenePix Array List file: parsed from row 6

[0849] GenePix Settings file: parsed from row 5

[0850] Delete rows 1–8, 10–12, 20, 22, and 25.

[0851] Arrange remaining values in rows 15–29.

[0852] Fill in rows 1–14 with the following:

[0853] Row 1: ScanID (date image file was last modified, formatted as yyyy.mm.dd-hh.mm.ss)

[0854] Row 2: SlideID, from stored value

[0855] Row 3: Name of person who scanned the slide, from user input

[0856] Row 4: Inage file name, from stored value

[0857] Row 5: Green PMT setting, from user input

[0858] Row 6: Red PMT setting, from user input

[0859] Row 7: ExtractID (date .gpr file was created, formatted as yyyy.mm.dd-hh.mm.ss)

[0860] Row 8: Name of person who performed the feature extraction, from user input

[0861] Row 9: Feature extraction software used, from stored value

[0862] Row 10: Results file name (same as the .gpr file name)

[0863] Row 11: GenePix Array List file, from stored value

[0864] Row 12: GenePix Settings file, from stored value

[0865] Row 13: StorageCD, currently left blank

[0866] Row 14: Extraction comments, from user input (anything about the scanning or feature extraction of the image the user feels might be relevant when selecting which hybridizations to include in an analysis)

Pre-Processing with Excel Templates

[0867] Following analysis of the image and extraction of the data, the data from each hybridization was pre-processed to extract data that was entered into the database and subsequently used for analysis. The complete GPR file produced by the feature extraction in GenePix was imported into an excel file pre-processing template. The same excel template was used to process each GPR file. The template performs a series of calculations on the data to differentiate poor features from others and to combine triplicate feature data into a single data point for each probe.

[0868] Each GPR file contained 31 rows of header information, followed by rows of data for 24093 features. The last of these rows was retained with the data. Rows 31 through the end of the file were imported into the excel template. Each row contained 43 columns of data. The only columns used in the pre-processing were: Oligo ID, F633 Median (median value from all the pixels in the feature for the Cy5 dye), B633 Median (the median value of all the pixels in the local background of the selected feature for Cy5), B633 SD (the standard deviation of the values for the pixels in the local background of the selected feature for Cy5), F532 Median (median value from all the pixels in the feature for the Cy3 dye), B532 Median (the median value of all the pixels in the local background of the selected feature for Cy3), B532 SD (the standard deviation of the values for the pixels in the local background of the selected feature for Cy3), and Flags. The GenePix Flags column contains the flags set during feature extraction. “−75” indicates there were no features printed on the array in that position, “−50” indicates that GenePix could not differentiate the feature signal from the local background, and “−100” indicates that the user marked the feature as bad.

[0869] Once imported, the rows with −75 flags were deleted. Then the median of B633 SD and B532 SD were calculated over all features with a flag value of “0”. The minimum values of B633 Median and B532 Median were identified, considering only those values associated with a flag value of “0”. For each feature, the signal to noise ratio (S/N) was calculated for both dyes by taking the fluorescence signal minus the local background (BGSS) and dividing it by the standard deviation of the local background:

[0870] $S/N=\frac{F_i-B_i}{{\mathrm{SDB}}_i}$

[0871] If the S/N was less than 3, then an adjusted background-subtracted signal was calculated as the fluorescence minus the minimum local background on the slide. An adjusted S/N was then calculated as the adjusted background subtracted signal divided by the median noise over all features for that channel. If the adjusted S/N was greater than three and the original S/N were less than three, a flag of 25 was set for the Cy5 channel, a flag of 23 was set for the Cy3 channel, and if both met these criteria, then a flag of 20 was set. If both the adjusted S/N and the original S/N were less than three, then a flag of 65 was set for Cy5, 63 set for Cy3, and 60 set if both dye channels had an adjusted S/N less than three. All signal to noise calculations, adjusted background-subtracted signal, and adjusted S/N were calculated for each dye channel. If the BGSS value was greater than or equal to 64000, a flag was set to indicate saturation; 55 for Cy5, 53 for Cy3, 50 for both.

[0872] The BGSS used for further calculations was the original BGSS if the original S/N was greater than or equal to three. If the original S/N ratio was less than three and the adjusted S/N ratio was greater than or equal to three, then the adjusted BGSS was used. If the adjusted S/N ratio was less than three, then the adjusted BGSS was used, but with knowledge of the flag status.

[0873] To facilitate comparison among arrays, the Cy3 and Cy5 data were scaled to have a median of 1. For each dye channel, the median value of all features with flags=0,20,23, or 25 was calculated. The BGSS for each dye in each feature was then divided by this median value. The Cy3/Cy5 ratio was calculated for each feature using the scaled BGSS:

[0874] $R_n=\frac{\mathrm{Cy}3S_i}{\mathrm{Cy}5S_i}$

[0875] The flag setting for each feature was used to determine the expression ratio for each probe, a combination of three features. If all three features had flag settings in the same category (categories=negatives, 0 to 25, 50–55, and 60–65), then the average and CV of the three feature ratios was calculated. If the CV of all three features was less than 15, the average was used. If the CV was greater than 15, then the CV of each combination of two of the features was calculated and the two features with the lowest CV were averaged. If none of the combinations of two features had a CV less than 15, then the median ratio of the three features was used as the probe feature.

[0876] If the three features do not have flags in the same category, then the features with the best quality flags were used (0>25>23>20>55>53>50>65>63>60). Features with negative flags were never used. When the best flags were two features in the same category, the average was used. If a single feature had a better flag category than the other two then that feature was used.

[0877] Once the probe expression ratio was calculated from the three features, the log of the ratio was taken as described below and stored for use in analyzing the data. Whichever features were used to calculate the probe value, the worst of the flags from those features was carried forward and stored as the flag value for that probe. 2 different data sets can be used for analysis. Flagged data uses all values, including those with flags. Filtered data sets are created by removing flagged data from the set before analysis.

[0878] Genes were identified which have expression patterns useful for the diagnosis and monitoring of cardiac allograft rejection. Further, sets of genes that work together in a diagnostic algorithm for allograft rejection were identified. Patients, patient clinical data and patient samples used in the discovery of markers below were derived from a clinical study described in example 11.

[0879] The collected clinical data is used to define patient or sample groups for correlation of expression data. Patient groups are identified for comparison, for example, a patient group that possesses a useful or interesting clinical distinction, verses a patient group that does not possess the distinction. Measures of cardiac allograft rejection are derived from the clinical data described above to divide patients (and patient samples) into groups with higher and lower rejection activity over some period of time or at any one point in time. Such data are rejection grade as determined from pathologist reading of the cardiac biopsies and data measuring progression of end-organ damage, including depressed left ventricular dysfunction (decreased cardiac output, decreased ejection fraction, clinical signs of low cardiac output) and usage of inotropic agents (Kobashigawa 1998).

[0880] Expression profiles correlating with occurrence of allograft rejection are identified, including expression profiles corresponding to end-organ damage and progression of end-organ damage. Expression profiles are identified predicting allograft rejection, and response to treatment or likelihood of response to treatment. Subsets of the candidate library (or a previously identified diagnostic nucleotide set) are identified, that have predictive value for the presence of allograft rejection or prediction of allograft rejection or end organ damage.

Identification of a Diagnostic Nucleotide Set for Diagnosis of Cardiac Allograft Acute Rejection

[0881] Mononuclear RNA samples were collected from patients who had recently undergone a cardiac allograft transplantation using the protocol described in example 8. The allograft rejection status at the time of sample collection was determined by examination of cardiac biopsies as described in example 11.

[0882] 180 samples were included in the analysis. Each patient sample was associated with a biopsy and clinical data collected at the time of the sample. The cardiac biopsies were graded by a pathologist at the local center and by a centralized pathologist who read the biopsy slides from all four local centers in a blinded manner. Biopsy grades included 0, 1A, 1B, 2, 3A, and 3B. No grade 4 rejection was identified. Dependent variables were developed based on these grades using either the local center pathology reading or the higher of the two readings, local or centralized. The dependent variables used for correlation of gene expression profiles with cardiac allograft rejection are shown in Table 13. Dependent variables are used to create classes of samples corresponding to the presence or absence of rejection.

[0883] Clinical data were also used to determine criteria for including samples in the analysis. The strictest inclusion criteria required that samples be from patients who did not have a bacterial or viral infection, were at least two weeks post cardiac transplant and were not currently admitted to the hospital. A second inclusion criteria (inclusion 2) reduced the post-transplant criteria to 1 week and eliminated the hospital admission criteria.

[0884] After preparation of RNA (example 8), amplification, labeling, hybridization, scanning, feature extraction and data processing were done as described in Example 23, using the oligonucleotide microarrays described in Examples 20–22. The resulting log ratio of expression of Cy3 (patient sample)/Cy5 (R50 reference RNA) was used for analysis. This dataset is called the “static” data. A second type of dataset, referenced, was derived from the first. These datasets compared the gene expression log ratio in each sample to a baseline sample from the same patient using the formula:
ref log ratio=(log ratio_sample)−(log ratio_baseline)

[0885] Two referenced datasets were used, named “0 HG” and “Best 0”. The baseline for 0 HG was a Grade 0 sample from the same patient as the sample, using the highest grade between the centralized and local pathologists. The baseline for Best 0 was a Grade 0 sample from the same patient as the sample, using both the local and centralized reader biopsy grade data. When possible a Grade 0 prior to the sample was used as the baseline in both referenced datasets.

[0886] The datasets were also divided into subsets to compare analysis between two subsets of roughly half of the data. The types of subsets constructed were as follows. First half/second half subsets were the first half of the samples and the second half of the samples from a dataset ordered by sample number. Odd/even subsets used the same source, a dataset ordered by sample number, but the odd subset consisted of every 2^nd sample starting with the first and the even subset consisted of every 2^nd sample starting with the second sample, Center 14/other subsets were the same datasets, divided by transplant hospital. The center 14 subset consisted of all samples from patients at center 14, while the other subset consisted of all samples from the other three centers (12,13, and 15).

[0887] Initially, significance analysis for microarrays (SAM, Tusher 2001, Example 26) was used to discover genes that were differentially expressed between the rejection and no-rejection groups. Ninety-six different combinations of dependent variables, inclusion criteria, static/referenced, and data subsets were used in SAM analysis to develop the primary lists of genes significantly differentially expressed between rejection and no-rejection. The most significant of these genes were chosen based on the following criteria. Tier 1 (A1, in Table 12A) genes were those which appeared with an FDR of less than 20% in identical analyses in two independent subsets. Tier 2 (A2 in Table 12A) genes were those which appeared in the top 20 genes on the list with an FDR less than 20% more than 50% of the time over all dependent variables with the inclusion criteria, and static/referenced constant. Tier 3 genes were those that appeared more than 50% of the time with an FDR less than 20% more than 50% of the time over all dependent variables with the inclusion criteria, and static/referenced constant. The genes that were identified by the analysis as statistically differentially expressed between rejection and no rejection are shown in Table 12A.

[0888] SAM chooses genes as significantly different based on the magnitude of the difference between the groups and the variation among the samples within each group. An example of the difference between some Grade 0 and some Grade 3A samples for 9 genes is show in FIG. 8A.

[0889] Additionally, many of these same combinations were used in the Supervised Harvesting of Expression Trees (SHET, Hastie et al. 2001) algorithm (see example 26) to identify markers that the algorithm chose as the best to distinguish between the rejection and no rejection classes using a bias factor of 0.01. The top 20 or 30 terms were taken from the SHET output and among all comparisons in either the static or referenced data the results were grouped. Any gene found in the top 5 terms in more than 50% of the analyses was selected to be in group B1 (Table 12A). The occurrences of each gene were tabulated over all SHET analysis (for either static or referenced data) and the 10 genes that occurred the most were selected to be in group B2 (Table 12A).

[0890] An additional classification method used was CART (Salford Systems, San Diego, example 26). Either the static or referenced dataset was reduced to only the genes for which expression values (log ratios) were present in at least 80% of the samples. These data were used in CART with the default settings, using the Symmetric Gini algorithm. Each of the dependent variables was used with both the full sample set and the strict inclusion criteria. Two groups of genes were identified. Group C1 were those genes that were a primary splitter (1^st decision node). Group C2 genes were the 10 genes that occurred as splitters the most often over all these analyses.

[0891] Two other classification models were developed and their best genes identified as markers of cardiac allograft rejection. Group D genes were identified from a set of 59 samples, referenced data, local biopsy reading grade, using logistic regression. Group E genes were identified from the primary static dataset using a K-nearest neighbor classification algorithm.

[0892] Both hierarchical clustering (Eisen et al. 1998) and CART were used to identify surrogates for each identified marker as shown in Table 12C. Hierarchical clustering surrogates are genes co-expressed in these and were chosen from the nearest branches of the dendrogram. CART surrogates were identified by CART as the surrogates for those genes chosen as primary splitters at decision nodes.

[0893] Primers for real-time PCR validation were designed for each of the marker genes as described in Example 25 and are listed in Table 12B.

[0894] CART was used to build a decision tree for classification of samples as rejection or no-rejection using the gene expression data from the arrays. The analysis identified sets of genes that can be used together to accurately identify samples derived from cardiac allograft transplant patients. The set of genes and the identified threshold expression levels for the decision tree are referred to as a “models”. This model can be used to predict the rejection state of an unknown sample. The input data were the static expression data (log ratio) and the referenced expression data (log ratio referenced to the best available grade 0 from either the centralized reader or the local reader) for 139 of our top marker genes.

[0895] These two types of expression data were entered into the CART software as independent variables. The dependent variable was rejection state, defined for this model as no rejection=grade 0 and rejection=grade 3A. Samples were eliminated from consideration in the training set if they were from patients with either bacterial or viral infection or were from patients who were less than two weeks post-transplant. The method used was Symmetric Gini, allowing linear combinations of independent variables. The costs were set to 1 for both false negatives and false positives and the priors were set equal for the two states. No penalties were assessed for missing data, however the marker genes selected have strong representation across the dataset. 10-fold cross validation was used to test the model. Settings not specified remained at the default values.

[0896] The model shown in FIG. 8B is based on decisions about expression values at three nodes, each a different marker gene. The cost assigned to this model is 0.292, based on the priors being equal, the costs set to 1 for each type of error, and the results from the 10-fold cross validation.

[0897] In the training set, no rejection samples were misclassified (sensitivity=100%) and only 1 no-rejection sample was misclassified (specificity=94.4%). Following 10-fold cross validation, 2 rejection samples were misclassified (sensitivity=87.5%) and 3 no-rejection samples were misclassified (specificity=83.3%). The CART software assigns surrogate markers for each decision node. For this model, the surrogates are shown in FIG. 8C and Table 12C.

[0898] These genes can be used alone or in association with other genes or variables to build a diagnostic gene set or a classification algorithm. These genes can be used in association with known gene markers for rejection (such as those identified in the prior art) to provide a diagnostic algorithm.

[0899] In examples 17 and 24, leukocyte gene expression was used to discover expression markers and diagnostic gene sets for clinical outcomes. It is desirable to validate the gene expression results for each gene using a more sensitive and quantitative technology such as real-time PCR. Further, it is possible for the diagnostic nucleotide sets to be implemented as a diagnostic test as a real-time PCR panel. Alternatively, the quantitative information provided by real-time PCR validation can be used to design a diagnostic test using any alternative quantitative or semi-quantitative gene expression technology.

[0900] To validate the results of the microarray experiments we used real-time, or kinetic, PCR. In this type of experiment the amplification product is measured during the PCR reaction. This enables the researcher to observe the amplification before any reagent becomes rate limiting for amplification. In kinetic PCR the measurement is of C_T (threshold cycle) or C_P (crossing point). This measurement (C_T=C_P) is the point at which an amplification curve crosses a threshold fluorescence value. The threshold is set to a point within the area where all of the reactions were in their linear phase of amplification. When measuring C_T, a lower C_T value is indicative of a higher amount of starting material since an earlier cycle number means the threshold was crossed more quickly.

[0901] Several fluorescence methodologies are available to measure amplification product in real-time PCR. Taqman (Applied BioSystems, Foster City, Calif.) uses fluorescence resonance energy transfer (FRET) to inhibit signal from a probe until the probe is degraded by the sequence specific binding and Taq 3′ exonuclease activity. Molecular Beacons (Stratagene, La Jolla, Calif.) also use FRET technology, whereby the fluorescence is measured when a hairpin structure is relaxed by the specific probe binding to the amplified DNA. The third commonly used chemistry is Sybr Green, a DNA-binding dye (Molecular Probes, Eugene, Oreg.). The more amplified product that is produced, the higher the signal. The Sybr Green method is sensitive to non-specific amplification products, increasing the importance of primer design and selection. Other detection chemistries can also been used, such as ethidium bromide or other DNA-binding dyes and many modifications of the fluorescent dye/quencher dye Taqman chemistry.

[0902] Initially, samples are chosen for validation, which have already been used for microarray based expression analysis. They are also chosen to represent important disease classes or disease criteria. For the first steps of this example (primer design, primer endpoint testing, and primer efficiency testing) we examined β-actin and β-GUS. These genes are considered “housekeeping” genes because they are required for maintenance in all cells. They are commonly used as a reference that is expected to not change with experimental treatment. We chose these two particular genes as references because they varied the least in expression across 5 mRNA samples examined by real-time PCR.

[0903] The inputs for real time PCR reaction are gene-specific primers, cDNA from specific patient samples, and the standard reagents. The cDNA was produced from mononuclear RNA (prepared as in example 8) by reverse transcription using OligodT primers (Invitrogen, 18418-012) and random hexamers (Invitrogen, 48190-011) at a final concentration of 0.5 ng/μl and 3 ng/μl respectively. For the first strand reaction mix, 1.45 μg/μl of total RNA (R50, universal leukocyte reference RNA as described in Example 9) and 1 μl of the Oligo dT/Random Hexamer Mix, were added to water to a final volume of 11.5 μl. The sample mix was then placed at 70° C. for 10 minutes.

[0904] Following the 70° C. incubation, the samples were chilled on ice, spun down, and 88.5 μl of first strand buffer mix dispensed into the reaction tube. The final first strand buffer mix produced final concentrations of 1× first strand buffer (Invitrogen, Y00146, Carlsbad, Calif.), 0.01 mM DTT (Invitrogen, Y00147), 0.1 mM dATP (NEB, N0440S, Beverly, Mass.), 0.1 mM dGTP (NEB, N0442S), 0.1 mM dTTP (NEB, N0443S), 0.1 mM dCTP (NEB, N0441 S), 2U of reverse transcriptase (Superscript II, Invitrogen, 18064-014), and 0.18U of RNase inhibitor (RNAGaurd Amersham Pharmacia, 27-0815-01, Piscataway, N.J.). The reaction was incubated at 42° C. for 1 hour. After incubation the enzyme was heat inactivated at 70° C. for 15 minutes, 1 μl of RNAse H added to the reaction tube, and incubated at 37° C. for 20 minutes.

Primer Design

[0905] Two methods were used to design primers. The first was to use the software, Primer Express™ and recommendations for primer design that are provided with the GeneAmp® 7700 Sequence Detection System supplied by Applied BioSystems (Foster City, Calif.). The second method used to design primers was the PRIMER3 ver 0.9 program that is available from the Whitehead Research Institute, Cambridge, Mass. at the web site genome.wi.mit.edu/genome_software/other/primer3.html. The program can also be accessed on the World Wide Web at the web site genome.wi.mit.edu/cgi-bin/primer/primer3_www.cgi. Primers and Taqman/hybridization probes were designed as described below using both programs.

[0906] The Primer Express literature explains that primers should be designed with a melting temperature between 58 and 60 degrees C. while the Taqman probes should have a melting temperature of 68 to 70 under the salt conditions of the supplied reagents. The salt concentration is fixed in the software. Primers should be between 15 and 30 basepairs long. The primers should produce and amplicon in size between 50 and 150 base pairs, have a C-G content between 20% and 80%, have no more than 4 identical base pairs next to one another, and no more than 2 C's and G's in the last 5 bases of the 3′ end. The probe cannot have a G on the 5′ end and the strand with the fewest G's should be used for the probe.

[0907] Primer3 has a large number of parameters. The defaults were used for all except for melting temperature and the optimal size of the amplicon was set at 100 bases. One of the most critical is salt concentration as it affects the melting temperature of the probes and primers. In order to produce primers and probes with melting temperatures equivalent to Primer Express, a number of primers and probes designed by Primer Express were examined using PRIMER3. Using a salt concentration of 50 mM these primers had an average melting temperature of 3.7 degrees higher than predicted by Primer Express. In order to design primers and probes with equivalent melting temperatures as Primer Express using PRIMER3, a melting temperature of 62.7 plus/minus 1.0 degree was used in PRIMER3 for primers and 72.7 plus/minus 1.0 degrees for probes with a salt concentration of 50 mM.

[0908] The C source code for Primer3 was downloaded and complied on a Sun Enterprise 250 server using the GCC complier. The program was then used from the command line using a input file that contained the sequence for which we wanted to design primers and probes along with the input parameters as described by help files that accompany the software. Using scripting it was possible to input a number of sequences and automatically generate a number of possible probes and primers.

[0909] Primers for β-Actin (Beta Actin, Genbank Locus: NM_—001101) and β-GUS: glucuronidase, beta, (GUSB, Genbank Locus: NM_—000181), two reference genes, were designed using both methods and are shown here as examples:

[0910] The first step was to mask out repetitive sequences found in the mRNA sequences using RepeatMasker program that can be accessed at: the web site repeatmasker.genome.washington.edu/cgi-bin/RepeatMasker (Smit, A F A & Green, P “RepeatMasker” at the web site ftp.genome.washington.edu/RM/RepeatMasker.html). The last 500 basepairs on the last 3′ end of masked sequence was then submitted to PRIMER3 using the following exemplary input file:

[0911]

PRIMER_SEQUENCE_ID=>ACTB Beta Actin

PRIMER_EXPLAIN_FLAG=1

PRIMER_MISPRIMING_LIBRARY=

SEQUENCE=TTGGCTTGACTCAGGATTTAAAAACTGGAACGGTGAAGGTGACAGCAGTCGGTTGGACGAGCATCCCCCAA

AGTTCACAATGTGGCCGAGGACTTTGATTGCACATTGTTGTTTTTTAATAGTCATTCCAAATATGAGATGCATTGTTACA

GGAAGTCCCTTGCCATCCTAAAAGCACCCCACTTCTCTCTAAGGAGAATGGCCCAGTCCTCTCCCAAGTCCACACAGGGG

AGGGATAGCATTGCTTTCGTGTAAATTATGTAATGCAAAATTTTTTTAATCTTCGCCTTAATCTTTTTTATTTTGTTTTA

TTTTGAATGATGAGCCTTCGTGCCCCCCCTTCCCCCTTTTTTCCCCCAACTTGAGATGTATGAAGGCTTTTGGTCTCCCT

GGGAGTGGGTGGAGGCAGCCGGGCTTACCTGTACACTGACTTGAGACCAGTTGAATAAAAGTGCACACCTTA (SEQ ID

NO: 9956)

PRIMER_PRODUCT_OPT_SIZE=100

PRIMER_NUM_RETURN=100

PRIMER_MAX_END_STABILITY=9.0

PRIMER_MAX_MISPRIMING=12.00

PRIMER_PAIR_MAX_MISPRIMING=24.00

PRIMER_MIN_SIZE=18

PRIMER_OPT_SIZE=20

PRIMER_MAX_SIZE=32

PRIMER_MIN_TM=61.7

PRIMER_OPT_TM=62.7

PRIMER_MAX_TM=63.7

PRIMER_MAX_DIFF_TM=100.0

PRIMER_MIN_GC=20.0

PRIMER_MAX_GC=80.0

PRIMER_SELF_ANY=8.00

PRIMER_SELF_END=3.00

PRIMER_NUM_NS_ACCEPTED=0

PRIMER_MAX_POLY_X=4

PRIMER_OUTSIDE_PENALTY=0

PRIMER_GC_CLAMP=0

PRIMER_SALT_CONC=50.0

PRIMER_DNA_CONC=50.0

PRIMER_LIBERAL_BASE=1

PRIMER_MIN_QUALITY=0

PRIMER_MIN_END_QUALITY=0

PRIMER_QUALITY_RANGE_MIN=0

PRIMER_QUALITY_RANGE_MAX=100

PRIMER_WT_TM_LT=1.0

PRIMER_WT_TM_GT=1.0

PRIMER_WT_SIZE_LT=1.0

PRIMER_WT_SIZE_GT=1.0

PRIMER_WT_GC_PERCENT_LT=0.0

PRIMER_WT_GC_PERCENT_GT=0.0

PRIMER_WT_COMPL_ANY=0.0

PRIMER_WT_COMPL_END=0.0

PRIMER_WT_NUM_NS=0.0

PRIMER_WT_REP_SIM=0.0

PRIMER_WT_SEQ_QUAL=0.0

PRIMER_WT_END_QUAL=0.0

PRIMER_WT_POS_PENALTY=0.0

PRIMER_WT_END_STABILITY=0.0

PRIMER_PAIR_WT_PRODUCT_SIZE_LT=0.05

PRIMER_PAIR_WT_PRODUCT_SIZE_GT=0.05

PRIMER_PAIR_WT_PRODUCT_TM_LT=0.0

PRIMER_PAIR_WT_PRODUCT_TM_GT=0.0

PRIMER_PAIR_WT_DIFF_TM=0.0

PRIMER_PAIR_WT_COMPL_ANY=0.0

PRIMER_PAIR_WT_COMPL_END=0.0

PRIMER_PAIR_WT_REP_SIM=0.0

PRIMER_PAIR_WT_PR_PENALTY=1.0

PRIMER_PAIR_WT_IO_PENALTY=0.0

PRIMER_INTERNAL_OLIGO_MIN_SIZE=18

PRIMER_INTERNAL_OLIGO_OPT_SIZE=20

PRIMER_INTERNAL_OLIGO_MAX_SIZE=35

PRIMER_INTERNAL_OLIGO_MIN_TM=71.7

PRIMER_INTERNAL_OLIGO_OPT_TM=72.7

PRIMER_INTERNAL_OLIGO_MAX_TM=73.7

PRIMER_INTERNAL_OLIGO_MIN_GC=20.0

PRIMER_INTERNAL_OLIGO_MAX_GC=80.0

PRIMER_INTERNAL_OLIGO_SELF_ANY=12.00

PRIMER_INTERNAL_OLIGO_SELF_END=12.00

PRIMER_INTERNAL_OLIGO_NUM_NS=0

PRIMER_INTERNAL_OLIGO_MAX_POLY_X=5

PRIMER_INTERNAL_OLIGO_MISHYB_LIBRARY=

PRIMER_INTERNAL_OLIGO_MAX_MISHYB=12.00

PRIMER_INTERNAL_OLIGO_MIN_QUALITY=0

PRIMER_INTERNAL_OLIGO_SALT_CONC=50.0

PRIMER_INTERNAL_OLIGO_DNA_CONC=50.0

PRIMER_IO_WT_TM_LT=1.0

PRIMER_IO_WT_TM_GT=1.0

PRIMER_IO_WT_SIZE_LT=1.0

PRIMER_IO_WT_SIZE_GT=1.0

PRIMER_IO_WT_GC_PERCENT_LT=0.0

PRIMER_IO_WT_GC_PERCENT_GT=0.0

PRIMER_IO_WT_COMPL_ANY=0.0

PRIMER_IO_WT_NUM_NS=0.0

PRIMER_IO_WT_REP_SIM=0.0

PRIMER_IO_WT_SEQ_QUAL=0.0

PRIMER_TASK=pick_pcr_primers_and_hyb_probe

PRIMER_PRODUCT_SIZE_RANGE=50–150

PRIMER_FIRST_BASE_INDEX=1

PRIMER_PICK_ANYWAY=1

=

PRIMER_SEQUENCE_ID=>GUSB

PRIMER_EXPLAIN_FLAG=1

PRIMER_MISPRIMING_LIBRARY=

SEQUENCE=GAAGAGTACCAGAAAAGTCTGCTAGAGCAGTACCATCTGGGTCTGGATCAAAAACGCAGAAAATATGTGGT

TGGAGAGCTCATTTGGAATTTTGCCGATTTCATGACTGAACAGTCACCGACGAGAGTGCTGGGGAATAAAAAGGGGATCT

TCACTCGGCAGAGACAACCAAAAAGTGCAGCGTTCCTTTTGCGAGAGAGATACTGGAAGATTGCCAATGAAACCAGGTAT

CCCCACTCAGTAGCCAAGTCACAATGTTTGGAAAACAGCCCGTTTACTTGAGCAAGACTGATACCACCTGCGTGTCCCTT

CCTCCCCGAGTCAGGGCGACTTCCACAGCAGCAGAACAAGTGCCTCCTGGACTGTTCACGGCAGACCAGAACGTTTCTGG

CCTGGGTTTTGTGGTCATCTATTCTAGCAGGGAACACTAAAGGTGGAAATAAAAGATTTTCTATTATGGAAATAAAGAGT

TGGCATGAAAGTCGCTACTG (SEQ ID NO: 9957)

PRIMER_PRODUCT_OPT_SIZE=100

PRIMER_NUM_RETURN=100

PRIMER_MAX_END_STABILITY=9.0

PRIMER_MAX_MISPRIMING=12.00

PRIMER_PAIR_MAX_MISPRIMING=24.00

PRIMER_MIN_SIZE=18

PRIMER_OPT_SIZE=20

PRIMER_MAX_SIZE=32

PRIMER_MIN_TM=61.7

PRIMER_OPT_TM=62.7

PRIMER_MAX_TM=63.7

PRIMER_MAX_DIFF_TM=100.0

PRIMER_MIN_GC=20.0

PRIMER_MAX_GC=80.0

PRIMER_SELF_ANY=8.00

PRIMER_SELF_END=3.00

PRIMER_NUM_NS_ACCEPTED=0

PRIMER_MAX_POLY_X=4

PRIMER_OUTSIDE_PENALTY=0

PRIMER_GC_CLAMP=0

PRIMER_SALT_CONC=50.0

PRIMER_DNA_CONC=50.0

PRIMER_LIBERAL_BASE=1

PRIMER_MIN_QUALITY=0

PRIMER_MIN_END_QUALITY=0

PRIMER_QUALITY_RANGE_MIN=0

PRIMER_QUALITY_RANGE_MAX=100

PRIMER_WT_TM_LT=1.0

PRIMER_WT_TM_GT=1.0

PRIMER_WT_SIZE_LT=1.0

PRIMER_WT_SIZE_GT=1.0

PRIMER_WT_GC_PERCENT_LT=0.0

PRIMER_WT_GC_PERCENT_GT=0.0

PRIMER_WT_COMPL_ANY=0.0

PRIMER_WT_COMPL_END=0.0

PRIMER_WT_NUM_NS=0.0

PRIMER_WT_REP_SIM=0.0

PRIMER_WT_SEQ_QUAL=0.0

PRIMER_WT_END_QUAL=0.0

PRIMER_WT_POS_PENALTY=0.0

PRIMER_WT_END_STABILITY=0.0

PRIMER_PAIR_WT_PRODUCT_SIZE_LT=0.05

PRIMER_PAIR_WT_PRODUCT_SIZE_GT=0.05

PRIMER_PAIR_WT_PRODUCT_TM_LT=0.0

PRIMER_PAIR_WT_PRODUCT_TM_GT=0.0

PRIMER_PAIR_WT_DIFF_TM=0.0

PRIMER_PAIR_WT_COMPL_ANY=0.0

PRIMER_PAIR_WT_COMPL_END=0.0

PRIMER_PAIR_WT_REP_SIM=0.0

PRIMER_PAIR_WT_PR_PENALTY=1.0

PRIMER_PAIR_WT_IO_PENALTY=0.0

PRIMER_INTERNAL_OLIGO_MIN_SIZE=18

PRIMER_INTERNAL_OLIGO_OPT_SIZE=20

PRIMER_INTERNAL_OLIGO_MAX_SIZE=35

PRIMER_INTERNAL_OLIGO_MIN_TM=71.7

PRIMER_INTERNAL_OLIGO_OPT_TM=72.7

PRIMER_INTERNAL_OLIGO_MAX_TM=73.7

PRIMER_INTERNAL_OLIGO_MIN_GC=20.0

PRIMER_INTERNAL_OLIGO_MAX_GC=80.0

PRIMER_INTERNAL_OLIGO_SELF_ANY=12.00

PRIMER_INTERNAL_OLIGO_SELF_END=12.00

PRIMER_INTERNAL_OLIGO_NUM_NS=0

PRIMER_INTERNAL_OLIGO_MAX_POLY_X=5

PRIMER_INTERNAL_OLIGO_MISHYB_LIBRARY=

PRIMER_INTERNAL_OLIGO_MAX_MISHYB=12.00

PRIMER_INTERNAL_OLIGO_MIN_QUALITY=0

PRIMER_INTERNAL_OLIGO_SALT_CONC=50.0

PRIMER_INTERNAL_OLIGO_DNA_CONC=50.0

PRIMER_IO_WT_TM_LT=1.0

PRIMER_IO_WT_TM_GT=1.0

PRIMER_IO_WT_SIZE_LT=1.0

PRIMER_IO_WT_SIZE_GT=1.0

PRIMER_IO_WT_GC_PERCENT_LT=0.0

PRIMER_IO_WT_GC_PERCENT_GT=0.0

PRIMER_IO_WT_COMPL_ANY=0.0

PRIMER_IO_WT_NUM_NS=0.0

PRIMER_IO_WT_REP_SIM=0.0

PRIMER_IO_WT_SEQ_QUAL=0.0

PRIMER_TASK=pick_pcr_primers_and_hyb_probe

PRIMER_PRODUCT_SIZE_RANGE=50–150

PRIMER_FIRST_BASE_INDEX=1

PRIMER_PICK_ANYWAY=1

=

[0912] After running PRIMER3, 100 sets of primers and probes were generated for ACTB and GUSB. From this set, nested primers were chosen based on whether both left primers could be paired with both right primers and a single Taqman probe could be used on an insert of the correct size. With more experience we have decided not use the mix and match approach to primer selection and just use several of the top pairs of predicted primers.

[0913] For ACTB this turned out to be:

[0914] Forward 75 CACAATGTGGCCGAGGACTT (SEQ ID NO: 9174),

[0915] Forward 80 TGTGGCCGAGGACTTTGATT (SEQ ID NO: 9959),

[0916] Reverse 178 TGGCTTTTAGGATGGCAAGG (SEQ ID NO: 9176), and

[0917] Reverse 168 GGGGGCTTAGTTTGCTTCCT (SEQ ID NO: 9177).

[0918] Upon testing, the F75 and R178 pair worked best.

[0919] For GUSB the following primers were chosen:

[0920] Forward 59 AAGTGCAGCGTTCCTTTTGC,

[0921] Forward 65 AGCGTTCCTTTTGCGAGAGA,

[0922] Reverse 158 CGGGCTGTTTTCCAAACATT and

[0923] Reverse 197 GAAGGGACACGCAGGTGGTA.

[0924] No combination of these GUSB pairs worked well.

[0925] In addition to the primer pairs above, Primer Express predicted the following primers for GUSB:

[0926] Forward 178 TACCACCTGCGTGTCCCTTC (SEQ ID NO: 9182) and Reverse 242 GAGGCACTTGTTCTGCTGCTG (SEQ ID NO: 9183). This pair of primers worked to amplify the GUSB mRNA.

[0927] The parameters used to predict these primers in Primer Express were:

[0928] Primer Tm: min 58, Max=60, opt 59, max difference=2 degrees

[0929] Primer GC: min=20% Max=80% no 3′ G/C clamp

[0930] Primer: Length: min=9 max=40 opt=20

[0931] Amplicon: min Tm=0 max Tm=85

• • [0932] min=50 bp max=150 bp

[0933] Probe: Tm 10 degrees>primers, do not begin with a G on 5′ end

[0934] Other: max base pair repeat=3

• • [0935] max number of ambiguous residues=0
  • [0936] secondary structure: max consec bp=4, max total bp=8
  • [0937] Uniqueness: max consec match=9
    • [0938] max % match=75
    • [0939] max 3′ consecutive match=7

[0940] Granzyme B is an important marker of CMV infection and transplant rejection. Tables 13 A and B.

[0941] For Granzyme B the following sequence (NM_—004131) was used as input for Primer3:

[0942]

GGGGACTCTGGAGGCCCTCTTGTGTGTAACAAGGTGGCCCAGGGCATTGT

CTCCTATGGACGAAACAATGGCATGCCTCCACGAGCCTGCACCAAAGTCT

CAAGCTTTGTACACTGGATAAAGAAAACCATGAAACGCTACTAACTACAG

GAAGCAAACTAAGCCCCCGCTGTAATGAAACACCTTCTCTGGAGCCAAGT

CCAGATTTACACTGGGAGAGGTGCCAGCAACTGAATAAATACCTCTCCCA

GTGTAAATCTGGAGCCAAGTCCAGATTTACACTGGGAGAGGTGCCAGCAA

CTGAATAAATACCTCTTAGCTGAGTGG (SEQ ID NO: 9184)

• • [0943] For Granzyme B the following primers were chosen for testing:
  • [0944] Forward 81 ACGAGCCTGCACCAAAGTCTCT (SEQ ID NO: 9185)
  • [0945] Forward 63 AAACAATGGCATGCCTCCAC (SEQ ID NO: 9186)
  • [0946] Reverse 178 TCATTACAGCGGGGGCTTAG (SEQ ID NO: 9187)
  • [0947] Reverse 168 GGGGGCTTAGTTTGCTTCCT (SEQ ID NO: 9188)
  • [0948] Testing demonstrated that F81 and R178 worked well.

[0949] Using this approach, multiple primers were designed for genes that were shown to have expression patterns that correlated with clinical data in examples 17 and 24. These primer pairs are shown in Tables 13B and 14B and are added to the sequence listing. Primers can be designed from any region of a target gene using this approach.

Primer Endpoint Testing

[0950] Primers were first tested to examine whether they would produce the correct size product without non-specific amplification. The standard real-time PCR protocol was used with out the Rox and Sybr green dyes. Each primer pair was tested on cDNA made from universal mononuclear leukocyte reference RNA that was produced from 50 individuals as described in Example 9 (R50).

[0951] The PCR reaction consisted of 1× RealTime PCR Buffer (Ambion, Austin, Tex.), 3 mM MgCl2 (Applied BioSystems, B02953), 0.2 mM dATP (NEB), 0.2 mM dTTP (NEB), 0.2 mM dCTP (NEB), 0.2 mM dGTP (NEB), 1.25U AmpliTaq Gold (Applied BioSystems, Foster City, Calif.), 0.3 μM of each primer to be used (Sigma Genosys, The Woodlands, Tex.), 5 μl of the R50 reverse-transcription reaction and water to a final volume of 19 μl.

[0952] Following 40 cycles of PCR, one microliter of the product was examined by agarose gel electrophoresis and on an Agilent Bioanalyzer, DNA1000 chip (Palo Alto, Calif.). Results for 2 genes are shown in FIG. 10. From the primer design and the sequence of the target gene, one can calculate the expected size of the amplified DNA product. Only primer pairs with amplification of the desired product and minimal amplification of contaminants were used for real-time PCR. Primers that produced multiple products of different sizes are likely not specific for the gene of interest and may amplify multiple genes or chromosomal loci.

Primer Optimization/Efficiency

[0953] Once primers passed the end-point PCR, the primers were tested to determine the efficiency of the reaction in a real-time PCR reaction. cDNA was synthesized from starting total RNA as described above. A set of 5 serial dilutions of the R50 reverse-transcribed cDNA (as described above) were made in water: 1:10, 1:20, 1:40, 1:80, and 1:160.

[0954] The Sybr Green real-time PCR reaction was performed using the Taqman PCR Reagent kit (Applied BioSystems, Foster City, Calif., N808-0228). A master mix was made that consisted of all reagents except the primes and template. The final concentration of all ingredients in the reaction was 1× Taqman Buffer A (Applied BioSystems), 2 mM MgCl2 (Applied BioSystems), 200 μM dATP (Applied BioSystems), 200 μM dCTP (Applied BioSystems), 200 μM dGTP (Applied BioSystems), 400 μM dUTP (Applied BioSystems), 1:400,000 diluted Sybr Green dye (Molecular Probes), 1.25U AmpliTaq Gold (Applied BioSystems). The master mix for 92 reactions was made to a final volume of 2112 μl. 1012 μl of PCR master mix was dispensed into two, light-tight tubes. Each β-Actin primer F75 and R178 (Genosys), was added to one tube of PCR master mix and Each β-GUS primer F178 and R242 (Genosys), was added to the other tube of PCR master mix to a final primer concentration of 300 nM, and a final volume of 1035 μl per reaction tube. 45 μl of the β-Actin master mix was dispensed into 23 wells, in a 96 well plate (Applied BioSystems). 45 μl of the β-GUS master mix was dispensed into 23 wells, in a 96 well plate (Applied BioSystems). 5 μl of the template dilution series was dispensed into triplicate wells for each primer. The reaction was run on an ABI 7700 Sequence Detector (Applied BioSystems).

[0955] The Sequence Detector v1.7 software was used to analyze the fluorescent signal from each well. A threshold value was selected that allowed most of the amplification curves to cross the threshold during the linear phase of amplification. The cycle number at which each amplification curve crossed the threshold (C_T) was recorded and the file transferred to MS Excel for further analysis. The C_T values for triplicate wells were averaged. The data were plotted as a function of the log₁₀ of the calculated starting concentration of RNA. The starting RNA concentration for each cDNA dilution was determined based on the original amount of RNA used in the RT reaction, the dilution of the RT reaction, and the amount used (5 μl) in the real-time PCR reaction. For each gene, a linear regression line was plotted through all of the dilutions series points. The slope of the line was used to calculate the efficiency of the reaction for each primer set using the equation:
E=10^(−1/slope)

[0956] Using this equation (Pfaffl 2001), the efficiency for these β-actin primers is 2.28 and the efficiency for these β-GUS primers is 2.14 (FIG. 11). This efficiency was used when comparing the expression levels among multiple genes and multiple samples. This same method was used to calculate reaction efficiency for primer pairs for each gene we studied.

Assay and Results

[0957] Once primers were designed and tested and efficiency analysis was completed, primers were used examine expression of a single gene among many clinical samples. The basic design was to examine expression of both the experimental gene and a reference gene in each sample and, at the same time, in a control sample. The control sample we used was the universal mononuclear leukocyte reference RNA described in example 9 (R50).

[0958] In this example, three patient samples from patients with known CMV infection were compared to three patient samples from patients with no diagnosis of CMV infection based on standard diagnostic algorithms for active CMV infection (including viral PCR assays, serologies, culture and other tests). cDNA was made from all six RNA samples and the R50 control as described above. The cDNA was diluted 1:10 in water and 5 μl of this dilution was used in the 50 μl PCR reaction. Each 96-well plate consisted of 32 reactions, each done in triplicate. There were 17 templates and 3 primer sets. The three primer sets were β-GUS, β-Actin, and Granzyme B. The β-GUS and β-Actin primers are shown above and the Granzyme B primers were:

[0959] F81: ACGAGCCTGCACCAAAGTCT (SEQ ID NO: 9185),

[0960] R178: TCATTACAGCGGGGGCTT (SEQ ID NO: 9190),

[0961] Each of the three primer sets was used to measure template levels in 8 templates: the six experimental samples, R50, and water (no-template control). The β-GUS primers were also used to measure template levels a set of 8 templates identical except for the absence of the reverse transcriptase enzyme in the cDNA synthesis reaction (−RT). The real-time PCR reactions were performed as described above in “primer optimization/efficiency”.

[0962] The β-GUS amplification with +RT and −RT cDNA synthesis reaction templates were compared to measure the amount of genomic DNA contamination of the patient RNA sample (FIG. 9A). The only source of amplifiable material in the −RT cDNA synthesis reaction is contaminating genomic DNA. Separation by at least four CT between the −RT and +RT for each sample was required to consider the sample useful for analysis of RNA levels. Since a C_T decrease of one is a two-fold increase in template, a difference of four C_T would indicate that genomic DNA contamination level in the +RT samples was 6.25% of the total signal. Since we used these reactions to measure 30% or greater differences, a 6% contamination would not change the result.

[0963] For samples with sufficiently low genomic DNA contamination the data were used to identify differences in gene expression by measuring RNA levels. C_T values from the triplicate wells for each reaction were averaged and the coefficient of variation (CV) determined. Samples with high CV (>2%) were examined and outlier reaction wells were discarded from further analysis. The average of the wells for each sample was taken as the C_T value for each sample. For each gene, the ΔC_T was the R50 control C_T minus the sample C_T. The equation below was then used to identify an expression ratio compared to a reference gene (β-Actin) and control sample (R50) for Granzyme B expression in each experimental sample (Pfaffl, M. W. 2001). E is the amplification efficiency determined above.

[0964] $\mathrm{ratio}=\frac{{\left(E_{\mathrm{target}}\right)}^{\Delta C_{\tau }\mathrm{target}\left(\mathrm{control}-\mathrm{sample}\right)}}{{\left(E_{\mathrm{ref}}\right)}^{\Delta C_{\tau }\mathrm{ref}\left(\mathrm{control}-\mathrm{sample}\right)}}$

[0965] The complete experiment was performed in duplicate and the average of the two ratios taken for each gene. When β-Actin was used as the reference gene, the data show that Granzyme B is expressed at 25-fold higher levels in mononuclear cell RNA from patients with CMV than from patients without CMV (FIG. 9B). In this graph, each circle represents a patient sample and the black bars are the average of the three samples in each category.

[0966] After generation and processing of expression data sets from microarrays as described in Example 23, a log ratio value is used for most subsequent analysis. This is the logarithm of the expression ratio for each gene between sample and universal reference. The processing algorithm assigns a number of flags to data that are of low signal to noise or are in some other way of uncertain quality. Correlation analysis can proceed with all the data (including the flagged data) or can be done on filtered data sets where the flagged data is removed from the set. Filtered data should have less variability and may result in more significant results. Flagged data contains all information available and may allow discovery of genes that are missed with the filtered data set.

[0967] In addition to expression data, clinical data are included in the analysis. Continuous variables, such as the ejection fraction of the heart measured by echocardiography or the white blood cell count can be used for correlation analysis. In some cases, it may be desirable to take the logarithm of the values before analysis. These variables can be included in an analysis along with gene expression values, in which case they are treated as another “gene”. Sets of markers can be discovered that work to diagnose a patient condition and these can include both genes and clinical parameters. Categorical variables such as male or female can also be used as variables for correlation analysis. For example, the sex of a patient may be an important splitter for a classification tree.

[0968] Clinical data are used as supervising vectors for the significance or classification analysis. In this case, clinical data associated with the samples are used to divide samples in to clinically meaningful diagnostic categories for correlation or classification analysis. For example, pathologic specimens from kidney biopsies can be used to divide lupus patients into groups with and without kidney disease. A third or more categories can also be included (for example “unknown” or “not reported”). After generation of expression data and definition of using supervising vectors, correlation, significance and classification analysis is used to determine which set of genes are most appropriate for diagnosis and classification of patients and patient samples.

Significance Analysis for Microarrays (SAM)

[0969] Significance analysis for microarrays (SAM) (Tusher 2001) is a method through which genes with a correlation between their expression values and the response vector are statistically discovered and assigned a statistical significance. The ratio of false significant to significant genes is the False Discovery Rate (FDR). This means that for each threshold there are a set of genes which are called significant, and the FDR gives a confidence level for this claim. If a gene is called differentially expressed between 2 classes by SAM, with a FDR of 5%, there is a 95% chance that the gene is actually differentially expressed between the classes. SAM takes into account the variability and large number of variables of microarrays. SAM will identify genes that are most globally differentially expressed between the classes. Thus, important genes for identifying and classifying outlier samples or patients may not be identified by SAM.

[0970] After generation of data from patient samples and definition of categories using clinical data as supervising vectors, SAM is used to detect genes that are likely to be differentially expressed between the groupings. Those genes with the highest significance can be validated by real-time PCR (Example 25) or can be used to build a classification algorithm as described here.

Classification

[0971] Supervised harvesting of expression trees (Hastie et al. 2001) identifies genes or clusters that best distinguish one class from all the others on the data set. The method is used to identify the genes/clusters that can best separate one class versus all the others for datasets that include two or more classes from each other. This algorithm can be used to identify genes that are used to create a diagnostic algorithm. Genes that are identified can be used to build a classification tree with algorithms such as CART.

[0972] CART is a decision tree classification algorithm (Breiman 1984). From gene expression and or other data, CART can develop a decision tree for the classification of samples. Each node on the decision tree involves a query about the expression level of one or more genes or variables. Samples that are above the threshold go down one branch of the decision tree and samples that are not go down the other branch. Genes from expression data sets can be selected for classification building using CART by significant differential expression in SAM analysis (or other significance test), identification by supervised tree-harvesting analysis, high fold change between sample groups, or known relevance to classification of the target diseases. In addition, clinical data can also be used as variables for CART that are of know importance to the clinical question or are found to be significant predictors by multivariate analysis or some other technique. CART identifies surrogates for each splitter (genes that are the next best substitute for a useful gene in classification). Analysis is performed in CART by weighting misclassification costs to optimize desired performance of the assay. For example, it may be most important the sensitivity of a test for a given diagnosis be near 100% while specificity is less important.

[0973] Once a set of genes and expression criteria for those genes have been established for classification, cross validation is done. There are many approaches, including a 10 fold cross validation analysis in which 10% of the training samples are left out of the analysis and the classification algorithm is built with the remaining 90%. The 10% are then used as a test set for the algorithm. The process is repeated 10 times with 10% of the samples being left out as a test set each time. Through this analysis, one can derive a cross validation error which helps estimate the robustness of the algorithm for use on prospective (test) samples. When a gene set is established for a diagnosis with a low cross validation error, this set of genes is tested using samples that were not included in the initial analysis (test samples). These samples may be taken from archives generated during the clinical study. Alternatively, a new prospective clinical study can be initiated, where samples are obtained and the gene set is used to predict patient diagnoses.

Amplification, Labeling, Hybridization and Scanning

[0974] Samples consisting of at least 0.5 to 2 μg of intact total RNA were further processed for array hybridization. When available, 2 μg of intact total RNA is used for amplification. Amplification and labeling of total RNA samples was performed in three successive enzymatic reactions. First, a single-stranded DNA copy of the RNA was made (hereinafter, “ss-cDNA”). Second, the ss-cDNA was used as a template for the complementary DNA strand, producing double-stranded cDNA (hereinafter, “ds-cDNA, or cDNA”). Third, linear amplification was performed by in vitro transcription from a bacterial T₇ promoter. During this step, fluorescent-conjugated nucleotides were incorporated into the amplified RNA (hereinafter, “aRNA”).

[0975] The first strand cDNA was produced using the Invitrogen kit (Superscript II). The first strand cDNA was produced in a reaction composed of 50 mM Tris-HCl (pH 8.3), 75 mM KCl, and 3 mM MgCl₂ (1× First Strand Buffer, Invitrogen), 0.5 mM dGTP, 0.5 mM dATP, 0.5 mM dTTP, 0.5 mM dCTP, 10 mM DTT, 200 U reverse transcriptase (Superscript II, Invitrogen, #18064014), 15 U RNase inhibitor (RNAGuard, Amersham Pharmacia, #27-0815-01), 5 μM T7T24 primer (5′-GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGGTTTTTTTTTTTT TTTTTTTTTTTT-3′), (SEQ ID NO:8831) and 0.5 to 2 μg of selected sample total RNA. Several purified, recombinant control mRNAs from the plant Arabidopsis thaliana were added to the reaction mixture: 2–20 pg of the following genes CAB, RCA, LTP4, NAC1, RCP1,XCP2, RBCL, LTP6, TIM, and PRKase (Stratagene, #252201, #252202, #252204, #252208, #252207, #252206, #252203, #252205, #252209, #252210 respectively). The control RNAs allow the estimate of copy numbers for individual mRNAs in the clinical sample because corresponding sense oligonucleotide probes for each of these plant genes are present on the microarray. The final reaction volume of 20 μl was incubated at 42° C. for 90 min.

[0976] For synthesis of the second cDNA strand, DNA polymerase and RNase were added to the previous reaction, bringing the final volume to 150 μl. The previous contents were diluted and new substrates were added to a final concentration of 20 mM Tris-HCl (pH 7.0) (Fisher Scientific, Pittsburgh, Pa. #BP1756-100), 90 mMKCl (Teknova, Half Moon Bay, Calif., #0313-500), 4.6 mM MgCl₂ (Teknova, Half Moon Bay, Calif., #0304-500), 10 mM(NH₄)₂SO₄ (Fisher Scientific #A702-500)(1× Second Strand buffer, Invitrogen), 0.266 mM dGTP, 0.266 mM dATP, 0.266 mM dTTP, 0.266 mM dCTP, 40 U E. coli DNA polymerase (Invitrogen, #18010-025), and 2 U RNaseH (Invitrogen, #18021-014). The second strand synthesis took place at 16° C. for 150 minutes.

[0977] Following second-strand synthesis, the ds-cDNA was purified from the enzymes, dNTPs, and buffers before proceeding to amplification, using phenol-chloroform extraction followed by ethanol precipitation of the cDNA in the presence of glycogen.

[0978] Alternatively, a silica-gel column is used to purify the cDNA (e.g. Qiaquick PCR cleanup from Qiagen, #28104). The volume of the column purified cDNA was reduced by ethanol precipitation in the presence of glycogen in which the cDNA was collected by centrifugation at >10,000×g for 30 minutes, the supernatant is aspirated, and 150 μl of 70% ethanol, 30% water was added to wash the DNA pellet. Following centrifugation, the supernatant was removed, and residual ethanol was evaporated at room temperature. Alternatively, the volume of the column purified cDNA is reduce in a vacuum evaporator where the supernatant is reduce to a final volume of 7.4 μl.

[0979] Linear amplification of the cDNA was performed by in vitro transcription of the cDNA. The cDNA pellet from the step described above was resuspended in 7.4 μl of water, and in vitro transcription reaction buffer was added to a final volume of 20 μl containing 7.5 mM GTP, 7.5 mM ATP, 7.5 mM TTP, 2.25 mM CTP, 1.025 mM Cy3-conjugated CTP (Perkin Elmer; Boston, Mass., #NEL-580), 1× reaction buffer (Ambion, Megascript Kit, Austin, Tex. and #1334) and 1% T₇ polymerase enzyme mix (Ambion, Megascript Kit, Austin, Tex. and #1334). This reaction was incubated at 37° C. overnight. Following in vitro transcription, the RNA was purified from the enzyme, buffers, and excess NTPs using the RNeasy kit from Qiagen (Valencia, Calif.; # 74106) as described in the vendor's protocol. A second elution step was performed and the two eluates were combined for a final volume of 60 μl. RNA is quantified using an Agilent 2100 bioanalyzer with the RNA 6000 nano LabChip.

[0980] Reference RNA was prepared as described above, except Cy5-CTP was incorporated instead of Cy3CTP. Reference RNA from five reactions, each reaction started with 2 ug total RNA, was pooled together and quantitated as described above.

Hybridization to an Array

[0981] RNA was prepared for hybridization as follows: for an 18 mm×55 mm array, 20 μg of amplified RNA (aRNA) was combined with 20 μg of reference aRNA. The combined sample and reference aRNA was concentrated by evaporating the water to 10 μl in a vacuum evaporator. The sample was fragmented by heating the sample at 95° C. for 30 minutes to fragment the RNA into 50–200 bp pieces. Alternatively, the combined sample and reference aRNA was concentrated by evaporating the water to 5 μl in a vacuum evaporator. Five μl of 20 mM zinc acetate was added to the aRNA and the mix incubated at 60° C. for 10 minutes. Following fragmentation, 40 μl of hybridization buffer was added to achieve final concentrations of 5×SSC and 0.20% SDS with 0.1 μg/ul of Cot-1 DNA (Invitrogen) as a competitor DNA. The final hybridization mix was heated to 98° C., and then reduced to 50° C. at 0.1° C. per second.

[0982] Alternatively, formamide is included in the hybridization mixture to lower the hybridization temperature.

[0983] The hybridization mixture was applied to a pre-heated 65° C. microarray, surface, covered with a glass coverslip (Corning, #2935-246), and placed on a pre-heated 65° C. hybridization chamber (Telechem, AHC-10). 15 ul of 5×SSC was placed in each of the reservoir in the hybridization chamber and the chamber was sealed and placed in a water bath at 62° C. for overnight (16–20 hrs). Following incubation, the slides were washed in 2×SSC, 0.1% SDS for five minutes at 30° C., then in 2×SSC for five minutes at 30° C., then in 2×SSC for another five minutes at 30° C., then in 0.2×SSC for two minutes at room temperature. The arrays were spun at 1000×g for 2 minutes to dry them. The dry microarrays are then scanned by methods described above.

[0984] The microarrays were imaged on the Agilent (Palo Alto, Calif.) scanner G2565AA. The scan settings using the Agilent software were as follows: for the PMT Sensitivity (100% Red and 100% Green); Scan Resolution (10 microns); red and green dye channels; used the default scan region for all slides in the carousel; using the largest scan region; scan date for Instrument ID; and barcode for Slide ID. The full image produced by the Agilent scanner was flipped, rotated, and split into two images (one for each signal channel) using TIFFSplitter (Agilent, Palo Alto, Calif.). The two channels are the output at 532 nm (Cy3-labeled sample) and 633 nm (Cy5-labeled R50). The individual images were loaded into GenePix 3.0 (Axon Instruments, Union City, Calif.) for feature extraction, each image was assigned an excitation wavelength corresponding the file opened; Red equals 633 nm and Green equals 532 nm. The setting file (gal) was opened and the grid was laid onto the image so that each spot in the grid overlapped with >50% of the feature. Then the GenePix software was used to find the features without setting minimum threshold value for a feature. For features with low signal intensity, GenePix reports “not found”. For all features, the diameter setting was adjusted to include only the feature if necessary.

[0985] The GenePix software determined the median pixel intensity for each feature (F_i) and the median pixel intensity of the local background for each feature (B_i) in both channels. The standard deviation (SDF_{i and} SDB_i) for each is also determined. Features for which GenePix could not discriminate the feature from the background were “flagged” as described below.

[0986] Following feature extraction into a “.gpr” file, the header information of the .gpr file was changed to carry accurate information into the database. An Excel macro was written to include the following information: Name of the original tif image file, SlideID, Version of the feature extraction software, GenePix Array List file, GenePix Settings file, ScanID, Name of person who scanned the slide, Green PMT setting, Red PMT setting, ExtractID (date .gpr file was created, formatted as yyyy.mm.dd-hh.mm.ss), Results file name (same as the .gpr file name), StorageCD, and Extraction comments.

Pre-Processing with Excel Templates

[0987] Following analysis of the image and extraction of the data, the data from each hybridization was pre-processed to extract data that was entered into the database and subsequently used for analysis. The complete GPR file produced by the feature extraction in GenePix was imported into an excel file pre-processing template or processed using a AWK script. Both programs used the same processing logic and produce identical results. The same excel template or AWK script was used to process each GPR file. The template performs a series of calculations on the data to differentiate poor features from others and to combine duplicate or triplicate feature data into a single data point for each probe.

[0988] The data columns used in the pre-processing were: Oligo ID, F633 Median (median value from all the pixels in the feature for the Cy5 dye), B633 Median (the median value of all the pixels in the local background of the selected feature for Cy5), B633 SD (the standard deviation of the values for the pixels in the local background of the selected feature for Cy5), F532 Median (median value from all the pixels in the feature for the Cy3 dye), B532 Median (the median value of all the pixels in the local background of the selected feature for Cy3), B532 SD (the standard deviation of the values for the pixels in the local background of the selected feature for Cy3), and Flags. The GenePix Flags column contains the flags set during feature extraction. “−75” indicates there were no features printed on the array in that position, “−50” indicates that GenePix could not differentiate the feature signal from the local background, and “−100” indicates that the user marked the feature as bad.

[0989] Once imported, the data associated with features with −75 flags was not used. Then the median of B633 SD and B532 SD were calculated over all features with a flag value of “0”. The minimum values of B633 Median and B532 Median were identified, considering only those values associated with a flag value of “0”. For each feature, the signal to noise ratio (S/N) was calculated for both dyes by taking the fluorescence signal minus the local background (BGSS) and dividing it by the standard deviation of the local background:

[0990] $S/N=\frac{F_i-B_i}{{\mathrm{SDB}}_i}$

[0991] If the S/N was less than 3, then an adjusted background-subtracted signal was calculated as the fluorescence minus the minimum local background on the slide. An adjusted S/N was then calculated as the adjusted background subtracted signal divided by the median noise over all features for that channel. If the adjusted S/N was greater than three and the original S/N were less than three, a flag of 25 was set for the Cy5 channel, a flag of 23 was set for the Cy3 channel, and if both met these criteria, then a flag of 28 was set. If both the adjusted S/N and the original S/N were less than three, then a flag of 65 was set for Cy5, 63 set for Cy3, and 68 set if both dye channels had an adjusted S/N less than three. All signal to noise calculations, adjusted background-subtracted signal, and adjusted S/N were calculated for each dye channel. If the BGSS value was greater than or equal to 64000, a flag was set to indicate saturation; 55 for Cy5, 53 for Cy3, 58 for both.

[0992] The BGSS used for further calculations was the original BGSS if the original S/N was greater than or equal to three. If the original S/N ratio was less than three and the adjusted S/N ratio was greater than or equal to three, then the adjusted BGSS was used. If the adjusted S/N ratio was less than three, then the adjusted BGSS was used, but with knowledge of the flag status.

[0993] To facilitate comparison among arrays, the Cy3 and Cy5 data were scaled. The log of the ratio of Green/Red was determined for all features. The median log ratio value for good features (Flags 0, 23, 25, 28, 63) was determined. The feature values were scaled using the following formula: Log_Scaled_Feature_Ratio=Log_Feature_Ratio−Median_Log_Ratio.

[0994] The flag setting for each feature was used to determine the expression ratio for each probe, a choice of one, two or three features. If all features had flag settings in the same category (categories=negatives, 0 to 28, 53–58, and 63–68), then the average of the three scaled, anti log feature ratios was calculated If the three features did not have flags in the same category, then the feature or features with the best quality flags were used (0>25>23>28>55>53>58>65>63>68). Features with negative flags were never used. When the best flags were two or three features in the same category, the anti log average was used. If a single feature had a better flag category than the other two then the anti log of that feature ratio was used.

[0995] Once the probe expression ratios were calculated from the one, two, or three features, the log of the scaled, averaged ratios was taken as described below and stored for use in analyzing the data. Whichever features were used to calculate the probe value, the flag from those features was carried forward and stored as the flag value for that probe. 2 different data sets can be used for analysis. Flagged data uses all values, including those with flags. Filtered data sets are created by removing flagged data from the set before analysis.

[0996] Leukocyte microarray gene expression was used to discover expression markers and diagnostic gene sets for clinical outcomes. It is desirable to validate the gene expression results for each gene using a more sensitive and quantitative technology such as real-time PCR. Further, it is possible for the diagnostic nucleotide sets to be implemented as a diagnostic test as a real-time PCR panel. Alternatively, the quantitative information provided by real-time PCR validation can be used to design a diagnostic test using any alternative quantitative or semi-quantitative gene expression technology.

[0997] To validate the results of the microarray experiments we used real-time, or kinetic, PCR. In this type of experiment the amplification product is measured during the PCR reaction. This enables the researcher to observe the amplification before any reagent becomes rate limiting for amplification. In kinetic PCR the measurement is of C_T (threshold cycle) or C_P (crossing point). This measurement (C_T=C_P) is the point at which an amplification curve crosses a threshold fluorescence value. The threshold is set to a point within the area where all of the reactions were in their linear phase of amplification. When measuring C_T, a lower C_T value is indicative of a higher amount of starting material since an earlier cycle number means the threshold was crossed more quickly.

[0998] Several fluorescence methodologies are available to measure amplification product in real-time PCR. Taqman (Applied BioSystems, Foster City, Calif.) uses fluorescence resonance energy transfer (FRET) to inhibit signal from a probe until the probe is degraded by the sequence specific binding and Taq 3′ exonuclease activity. Molecular Beacons (Stratagene, La Jolla, Calif.) also use FRET technology, whereby the fluorescence is measured when a hairpin structure is relaxed by the specific probe binding to the amplified DNA. The third commonly used chemistry is Sybr Green, a DNA-binding dye (Molecular Probes, Eugene, Oreg.). The more amplified product that is produced, the higher the signal. The Sybr Green method is sensitive to non-specific amplification products, increasing the importance of primer design and selection. Other detection chemistries can also been used, such as ethidium bromide or other DNA-binding dyes and many modifications of the fluorescent dye/quencher dye Taqman chemistry.

Sample Prep and cDNA Synthesis

[0999] The inputs for real time PCR reaction are gene-specific primers, cDNA from specific patient samples, and standard reagents. The cDNA was produced from mononuclear RNA (prepared as in example 8) or whole blood RNA by reverse transcription using Oligo dT primers (Invitrogen, 18418-012) and random hexamers (Invitrogen, 48190-011) at a final concentration of 0.5 ng/μl and 3 ng/μl respectively. For the first strand reaction mix, 0.5 μg of mononuclear total RNA or 2 μg of whole blood RNA and 1 μl of the Oligo dT/Random Hexamer Mix, were added to water to a final volume of 11.5 μl. The sample mix was then placed at 70° C. for 10 minutes. Following the 70° C. incubation, the samples were chilled on ice, spun down, and 88.5 μl of first strand buffer mix dispensed into the reaction tube. The final first strand buffer mix produced final concentrations of 1× first strand buffer (Invitrogen, Y00146, Carlsbad, Calif.), 10 mM DTT (Invitrogen, Y00147), 0.5 mM dATP (NEB, N0440S, Beverly, Mass.), 0.5 mM dGTP (NEB, N0442S), 0.5 mM dTTP (NEB, N0443S), 0.5 mM dCTP (NEB, N0441S), 200U of reverse transcriptase (Superscript II, Invitrogen, 18064-014), and 18U of RNase inhibitor (RNAGaurd Amersham Pharmacia, 27-0815-01, Piscataway, N.J.). The reaction was incubated at 42° C. for 90 minutes. After incubation the enzyme was heat inactivated at 70° C. for 15 minutes, 2 U of RNAse H added to the reaction tube, and incubated at 37° C. for 20 minutes.

Primer Design

[1000] Two methods were used to design primers. The first was to use the software, Primer Express™ and recommendations for primer design that are provided with the GeneAmp® 7700 Sequence Detection System supplied by Applied BioSystems (Foster City, Calif.).

[1001] The second method used to design primers was the PRIMER3 ver 0.9 program that is available from the Whitehead Research Institute, Cambridge, Mass. at the web site genome.wi.mit.edu/genome_software/other/primer3.html. The program can also be accessed on the World Wide Web at the web site genome.wi.mit.edu/cgi-bin/primer/primer3_www.cgi. Primers and Taqman/hybridization probes were designed as described below using both programs.

[1002] The Primer Express literature explains that primers should be designed with a melting temperature between 58 and 60 degrees C. while the Taqman probes should have a melting temperature of 68 to 70 under the salt conditions of the supplied reagents. The salt concentration is fixed in the software. Primers should be between 15 and 30 basepairs long. The primers should produce and amplicon in size between 50 and 150 base pairs, have a C–G content between 20% and 80%, have no more than 4 identical base pairs next to one another, and no more than 2 C's and G's in the last 5 bases of the 3′ end. The probe cannot have a G on the 5′ end and the strand with the fewest G's should be used for the probe.

[1003] Primer3 has a large number of parameters. The defaults were used for all except for melting temperature and the optimal size of the amplicon was set at 100 bases. One of the most critical is salt concentration as it affects the melting temperature of the probes and primers. In order to produce primers and probes with melting temperatures equivalent to Primer Express, a number of primers and probes designed by Primer Express were examined using PRIMER3. Using a salt concentration of 50 mM these primers had an average melting temperature of 3.7 degrees higher than predicted by Primer Express. In order to design primers and probes with equivalent melting temperatures as Primer Express using PRIMER3, a melting temperature of 62.7 plus/minus 1.0 degree was used in PRIMER3 for primers and 72.7 plus/minus 1.0 degrees for probes with a salt concentration of 50 mM.

[1004] The C source code for Primer3 was downloaded and complied on a Sun Enterprise 250 server using the GCC complier. The program was then used from the command line using a input file that contained the sequence for which we wanted to design primers and probes along with the input parameters as described by help files that accompany the software. Using scripting it was possible to input a number of sequences and automatically generate a number of possible probes and primers.

[1005] Primers for P-Actin (Beta Actin, Genbank Locus: NM_—001101) and β-GUS: glucuronidase, beta, (GUSB, Genbank Locus: NM_—000181), two reference genes, were designed using both methods and are shown here as examples:

[1006] The first step was to mask out repetitive sequences found in the mRNA sequences using RepeatMasker program that can be accessed at: the web site repeatmasker.genome.washington.edu/cgi-bin/RepeatMasker (Smit, A F A & Green, P “RepeatMasker” at the web site ftp.genome.washington.edu/RM/RepeatMasker.html). The last 500 basepairs on the last 3′ end of masked sequence was then submitted to PRIMER3 using the following exemplary input sequences:

[1007]

PRIMER_SEQUENCE_ID=>ACTB Beta Actin (SEQID 9956)
SEQUENCE=TTGGCTTGACTCAGGATTTAAAAACTGGAACGGTGAAGGTGACAGCAGTCGGTTGGACGAGC
ATCCCCCAAAGTTCACAATGTGGCCGAGGACTTTGATTGCACATTGTTGTTTTTTAATAGTCATTCCAAAT
ATGAGATGCATTGTTACAGGAAGTCCCTTGCCATCCTAAAAGCACCCCACTTCTCTCTAAGGAGAATGGCC
CAGTCCTCTCCCAAGTCCACACAGGGGAGGGATAGCATTGCTTTCGTGTAAATTATGTAATGCAAAATTTT
TTTAATCTTCGCCTTAATCTTTTTTATTTTGTTTTATTTTGAATGATGAGCCTTCGTGCCCCCCCTTCCCC
CTTTTTTCCCCCAACTTGAGATGTATGAAGGCTTTTGGTCTCCCTGGGAGTGGGTGGAGGCAGCCGGGCTT
ACCTGTACACTGACTTGAGACCAGTTGAATAAAAGTGCACACCTTA
PRIMER_SEQUENCE_ID=>GUSB (SEQID 9957)
SEQUENCE=GAAGAGTACCAGAAAAGTCTGCTAGAGCAGTACCATCTGGGTCTGGATCAAAAACGCAGAAA
ATATGTGGTTGGAGAGCTCATTTGGAATTTTGCCGATTTCATGACTGAACAGTCACCGACGAGAGTGCTGG
GGAATAAAAAGGGGATCTTCACTCGGCAGAGACAACCAAAAAGTGCAGCGTTCCTTTTGCGAGAGAGATAC
TGGAAGATTGCCAATGAAACCAGGTATCCCCACTCAGTAGCCAAGTCACAATGTTTGGAAAACAGCCCGTT
TACTTGAGCAAGACTGATACCACCTGCGTGTCCCTTCCTCCCCGAGTCAGGGCGACTTCCACAGCAGCAGA
ACAAGTGCCTCCTGGACTGTTCACGGCAGACCAGAACGTTTCTGGCCTGGGTTTTGTGGTCATCTATTCTA
GCAGGGAACACTAAAGGTGGAAATAAAAGATTTTCTATTATGGAAATAAAGAGTTGGCATGAAAGTCGCTA
CTG

[1008] After running PRIMER3, 100 sets of primers and probes were generated for ACTB and GUSB. From this set, nested primers were chosen based on whether both left primers could be paired with both right primers and a single Taqman probe could be used on an insert of the correct size. With more experience we have decided not use the mix and match approach to primer selection and just use several of the top pairs of predicted primers.

[1009] For ACTB this turned out to be:

[1010] Forward 75 CACAATGTGGCCGAGGACTT(SEQ ID 9456),

[1011] Forward 80 TGTGGCCGAGGACTTTGATT(SEQ ID 9959),

[1012] Reverse 178 TGGCTTTTAGGATGGCAAGG(SEQ ID 9589), and

[1013] Reverse 168 GGGGGCTTAGTTTGCTTCCT(SEQ ID 9960).

[1014] Upon testing, the F75 and R178 pair worked best.

[1015] For GUSB the following primers were chosen:

[1016] Forward 59 AAGTGCAGCGTTCCTTTTGC(SEQ ID 9961),

[1017] Forward 65 AGCGTTCCTTTTGCGAGAGA (SEQ ID 9962),

[1018] Reverse 158 CGGGCTGTTTTCCAAACATT (SEQ ID 9963), and

[1019] Reverse 197 GAAGGGACACGCAGGTGGTA (SEQ ID 9964).

[1020] No combination of these GUSB pairs worked well:

[1021] In addition to the primer pairs above, Primer Express predicted the following primers for GUSB: Forward 178 TACCACCTGCGTGTCCCTTC (SEQ ID 9446) and Reverse 242 GAGGCACTTGTTCTGCTGCTG (SEQ ID 9579). This pair of primers worked to amplify the GUSB mRNA.

[1022] The parameters used to predict these primers in Primer Express were:

[1023] Primer Tm: min 58, Max=60, opt 59, max difference=2 degrees

[1024] Primer GC: min=20% Max=80% no 3′ G/C clamp

[1025] Primer: Length: min=9 max=40 opt=20

[1026] Amplicon: min Tm=0 max Tm=85

[1027] min=50 bp max=150 bp

[1028] Probe: Tm 10 degrees>primers, do not begin with a G on 5′ end

[1029] Other: max base pair repeat=3

[1030] max number of ambiguous residues=0

[1031] secondary structure: max consecutive bp=4, max total bp=8

[1032] Uniqueness: max consecutive match=9

[1033] max % match=75

[1034] max 3′ consecutive match=7

[1035] Granzyme B is a marker of transplant rejection.

[1036] For Granzyme B the following sequence (NM_—004131) (SEQ ID 9958) was used as input for Primer3:

[1037]

GGGGACTCTGGAGGCCCTCTTGTGTGTAACAAGGTGGCCCAGGGCATTGT

CTCCTATGGACGAAACAATGGCATGCCTCCACGAGCCTGCACCAAAGTCT

CAAGCTTTGTACACTGGATAAAGAAAACCATGAAACGCTACTAACTACAG

GAAGCAAACTAAGCCCCCGCTGTAATGAAACACCTTCTCTGGAGCCAAGT

CCAGATTTACACTGGGAGAGGTGCCAGCAACTGAATAAATACCTCTCCCA

GTGTAAATCTGGAGCCAAGTCCAGATTTACACTGGGAGAGGTGCCAGCAA

CTGAATAAATACCTCTTAGCTGAGTGG

[1038] For Granzyme B the following primers were chosen for testing:

[1039] Forward 81 ACGAGCCTGCACCAAAGTCT (SEQ ID 9488)

[1040] Forward 63 AAACAATGGCATGCCTCCAC (SEQ ID 9965)

[1041] Reverse 178 TCATTACAGCGGGGGCTTAG (SEQ ID 9621)

[1042] Reverse 168 GGGGGCTTAGTTTGCTTCCT (SEQ ID 9966)

[1043] Testing demonstrated that F81 and R178 worked well.

[1044] Using this approach, primers were designed for all the genes that were shown to have expression patterns that correlated with allograft rejection. These primer pairs are shown in Table 15C and are added to the sequence listing. Primers can be designed from any region of a target gene using this approach.

Primer Endpoint Testing

[1045] Primers were first tested to examine whether they would produce the correct size product without non-specific amplification. The standard real-time PCR protocol was used without the Rox and Sybr green dyes. Each primer pair was tested on cDNA made from universal mononuclear leukocyte reference RNA that was produced from 50 individuals as described in Example 9 (R50).

[1046] The PCR reaction consisted of 1× RealTime PCR Buffer (Ambion, Austin, Tex.), 2 mM MgCl2 (Applied BioSystems, B02953), 0.2 mM dATP (NEB), 0.2 mM dTTP (NEB), 0.2 mM dCTP (NEB), 0.2 mM dGTP (NEB), 0.625U AmpliTaq Gold (Applied BioSystems, Foster City, Calif.), 0.3 μM of each primer to be used (Sigma Genosys, The Woodlands, Tex.), 5 μl of the R50 reverse-transcription reaction and water to a final volume of 19 μl.

[1047] Following 40 cycles of PCR, 10 microliters of each product was combined with Sybr green at a final dilution of 1:72,000. Melt curves for each PCR product were determined on an ABI 7900 (Applied BioSystems, Foster City, Calif.), and primer pairs yielding a product with one clean peak were chosen for further analysis. One microliter of the product from these primer pairs was examined by agarose gel electrophoresis on an Agilent Bioanalyzer, DNA1000 chip (Palo Alto, Calif.). Results for 2 genes are shown in FIG. 10. From the primer design and the sequence of the target gene, one can calculate the expected size of the amplified DNA product. Only primer pairs with amplification of the desired product and minimal amplification of contaminants were used for real-time PCR. Primers that produced multiple products of different sizes are likely not specific for the gene of interest and may amplify multiple genes or chromosomal loci.

Primer Optimization/Efficiency

[1048] Once primers passed the end-point PCR, the primers were tested to determine the efficiency of the reaction in a real-time PCR reaction. cDNA was synthesized from starting total RNA as described above. A set of 5 serial dilutions of the R50 reverse-transcribed cDNA (as described above) were made in water: 1:10, 1:20, 1:40, 1:80, and 1:160.

[1049] The Sybr Green real-time PCR reaction was performed using the Taqman PCR Reagent kit (Applied BioSystems, Foster City, Calif., N808-0228). A master mix was made that consisted of all reagents except the primes and template. The final concentration of all ingredients in the reaction was 1× Taqman Buffer A (Applied BioSystems), 2 mM MgCl2 (Applied BioSystems), 200 μM dATP (Applied BioSystems), 200 μM dCTP (Applied BioSystems), 200 μM dGTP (Applied BioSystems), 400 μM dUTP (Applied BioSystems), 1:400,000 diluted Sybr Green dye (Molecular Probes), 1.25U AmpliTaq Gold (Applied BioSystems). The PCR master mix was dispensed into two, light-tight tubes. Each β-Actin primer F75 and R178 (Sigma-Genosys, The Woodlands, Tex.), was added to one tube of PCR master mix and Each β-GUS primer F178 and R242 (Sigma-Genosys), was added to the other tube of PCR master mix to a final primer concentration of 300 nM. 45 μl of the β-Actin or β-GUS master mix was dispensed into wells, in a 96-well plate (Applied BioSystems). 5 μl of the template dilution series was dispensed into triplicate wells for each primer. The reaction was run on an ABI 7900 Sequence Detection System (Applied BioSystems) with the following conditions: 10 min. at 95° C.; 40 cycles of 95° C. for 15 sec, 60° C. for 1 min; followed by a disassociation curve starting at 50° C. and ending at 95° C.

[1050] The Sequence Detection System v2.0 software was used to analyze the fluorescent signal from each well. The high end of the baseline was adjusted to between 8 and 20 cycles to reduce the impact on any data curves, yet be as high as possible to reduce baseline drift. A threshold value was selected that allowed the majority of the amplification curves to cross the threshold during the linear phase of amplification. The disassociation curve for each well was compared to other wells for that marker. This comparison allowed identification of “bad” wells, those that did not amplify, that amplified the wrong size product, or that amplified multiple products. The cycle number at which each amplification curve crossed the threshold (C_T) was recorded and the file transferred to MS Excel for further analysis. The C_T values for triplicate wells were averaged. The data were plotted as a function of the log₁₀ of the calculated starting concentration of RNA. The starting RNA concentration for each cDNA dilution was determined based on the original amount of RNA used in the RT reaction, the dilution of the RT reaction, and the amount used (5 μl) in the real-time PCR reaction. For each gene, a linear regression line was plotted through all of the dilutions series points. The slope of the line was used to calculate the efficiency of the reaction for each primer set using the equation:
E=10^(−1/slope)−1

[1051] Using this equation (Pfaffl 2001, Applied Biosystems User Bulletin #2), the efficiency for these β-actin primers is 1.28 and the efficiency for these P-GUS primers is 1.14 (FIG. 11). This efficiency was used when comparing the expression levels among multiple genes and multiple samples. This same method was used to calculate reaction efficiency for primer pairs for each gene studied. A primer pair was considered successful if the efficiency was reproducibly determined to be between 0.7 and 2.4.

SYBR-Green Assays

[1052] Once markers passed the Primer Efficiency QPCR (as stated above), they were used in real-time PCR assays. Patient RNA samples were reverse-transcribed to cDNA (as described above) and 1:10 dilutions made in water. In addition to the patient samples, a no template control (NTC) and a pooled reference RNA (see example 9) described in were included on every plate.

[1053] The Sybr Green real-time PCR reaction was performed using the Taqman Core PCR Reagent kit (Applied BioSystems, Foster City, Calif., N808-0228). A master mix was made that consisted of all reagents except the primers and template. The final concentration of all ingredients in the reaction was 1× Taqman Buffer A (Applied BioSystems), 2 mM MgCl2 (Applied BioSystems), 200 μM dATP (Applied BioSystems), 200 μM dCTP (Applied BioSystems), 200 μM dGTP (Applied BioSystems), 400 μM dUTP (Applied BioSystems), 1:400,000 diluted Sybr Green dye (Molecular Probes), 1.25U AmpliTaq Gold (Applied BioSystems). The PCR master mix was aliquoted into eight light-tight tubes, one for each marker to be examined across a set of samples. The optimized primer pair for each marker was then added to the PCR master mix to a final primer concentration of 300 nM. 18 μl of the each marker master mix was dispensed into wells in a 384 well plate (Applied BioSystems). 2 μl of the 1:10 diluted control or patient cDNA sample was dispensed into triplicate wells for each primer pair. The reaction was run on an ABI 7900 Sequence Detection System (Applied BioSystems) using the cycling conditions described above.

[1054] The Sequence Detection System v2.0 software (Applied BioSystems) was used to analyze the fluorescent signal from each well. The high end of the baseline was adjusted to between 8 and 20 cycles to reduce the impact on any data curves, yet be as high as possible to reduce baseline drift. A threshold value was selected that allowed the majority of the amplification curves to cross the threshold during the linear phase of amplification. The disassociation curve for each well was compared to other wells for that marker. This comparison allowed identification of “bad” wells, those that did not amplify, that amplified the wrong size product, or that amplified multiple products. The cycle number at which each amplification curve crossed the threshold (C_T) was recorded and the file transferred to MS Excel for further analysis. The C_T value representing any well identified as bad by analysis of disassociation curves was deleted. The C_T values for triplicate wells were averaged. A standard deviation (Stdev) and a coefficient of variation (CV) were calculated for the triplicate wells. If the CV was greater than 2, an outlier among the three wells was identified and deleted. Then the average was re-calculated. In each plate, ΔC_T was calculated for each marker-control combination by subtracting the average C_T of the target marker from the average C_T of the control (β-Actin or β-GUS). The expression relative to the control marker was calculated by taking two to the power of the ΔC_T of the target marker. For example, expression relative to β-Actin was calculated by the equation:
ErA=2^{(cT,Actin−ct,target)}

[1055] All plates were run in duplicate and analyzed in the same manner. The percent variation was determined for each sample-marker combination (relative expression) by taking the absolute value of the value of the RE for the second plate from the RE for the first plate, and dividing that by the average. If more than 25% of the variation calculations on a plate are greater than 50%, then a third plate was run.

Taqman Protocol

[1056] Real-time PCR assays were also done using Taqman PCR chemistry.

[1057] The Taqman real-time PCR reaction was performed using the Taqman Universal PCR Master Mix (Applied BioSystems, Foster City, Calif., #4324018). The master mix was aliquoted into eight, light-tight tubes, one for each marker. The optimized primer pair for each marker was then added to the correctly labeled tube of PCR master mix. A FAM/TAMRA dual-labeled Taqman probe (Biosearch Technologies, Navoto, Calif., DLO-FT-2) was then added to the correctly labeled tube of PCR master mix. Alternatively, different combinations of fluorescent reporter dyes and quenchers can be used such that the absorption wavelength for the quencher matches the emission wavelength for the reporter, as shown in table 16. 18 μl of the each marker master mix was dispensed into a 384 well plate (Applied BioSystems). 2 μl of the template sample was dispensed into triplicate wells for each primer pair. The final concentration of each reagent was: 1× TaqMan Universal PCR Master Mix, 300 nM each primer, 0.25 nM probe, 2 μl 1:10 diluted template. The reaction was run on an ABI 7900 Sequence Detection System (Applied Biosystems) using standard conditions (95° C. for 10 min., 40 cycles of 95° C. for 15 sec, 60° C. for 1 min.).

[1058] The Sequence Detector v2.0 software (Applied BioSystems) was used to analyze the fluorescent signal from each well. The high end of the baseline was adjusted to between 8 and 20 cycles to reduce the impact on any data curves, yet be as high as possible to reduce baseline drift. A threshold value was selected that allowed most of the amplification curves to cross the threshold during the linear phase of amplification. The cycle number at which each amplification curve crossed the threshold (C_T) was recorded and the file transferred to MS Excel for further analysis. The C_T values for triplicate wells were averaged. The C_T values for triplicate wells were averaged. A standard deviation (Stdev) and a coefficient of variation (CV) were calculated for the triplicate wells. If the CV was greater than 2, an outlier among the three wells was identified and deleted. Then the average was re-calculated. In each plate, ΔC_T was calculated for each marker-control combination by subtracting the average C_T of the target marker from the average C_T of the control (β-Actin or β-GUS). The expression relative to the control marker was calculated by taking two to the power of the ΔC_T of the target marker. All plates were run in duplicate and analyzed in the same manner. The percent variation was determined for each sample-marker combination (relative expression) by taking the absolute value of the value of the RE for the second plate from the RE for the first plate, and dividing that by the average. If more than 25% of the variation calculations on a plate are greater than 50%, then a third plate was run.

Bi-Plexing

[1059] Variation of real-time PCR assays can arise from unequal amounts of RNA starting material between reactions. In some assays, to reduce variation, the control gene amplification was included in the same reaction well as the target gene. To differentiate the signal from the two genes, different fluorescent dyes were used for the control gene. β-Actin was used as the control gene and the TaqMan probe used was labeled with the fluorescent dye VIC and the quencher TAMRA (Biosearch Technologies, Navoto, Calif., DLO-FT-2). Alternatively, other combinations of fluorescent reporter dyes and quenchers (table 16) can be used as long as the emission wavelength of the reporter for the control gene is sufficiently different from the wavelength of the reporter dye used for the target. The control gene primers and probe were used at limiting concentrations in the reaction (150 nM primers and 0.125 nM probe) to ensure that there were enough reagents to amplify the target marker. The plates were run under the same protocol and the data are analyzed in the same way, but with a separate baseline and threshold for the VIC signal. Outliers were removed as above from both the FAM and VIC signal channels. The expression relative to control was calculated as above, using the VIC signal from the control gene.
ErA=2^{(cT,VIC−CT,FAM)}
Absolute Quantitation

[1060] Instead of calculating the expression relative to a reference marker, an absolute quantitation can be performed using real-time PCR. To determine the absolute quantity of each marker, a standard curve is constructed using serial dilutions from a known amount of template for each marker on the plate. The standard curve may be made using cloned genes purified from bacteria or using synthetic complimentary oligonucleotides. In either case, a dilution series that covers the expected range of expression is used as template in a series of wells in the plate. From the average C_T values for these known amounts of template a standard curve can be plotted. From this curve the C_T values for the unknowns are used to identify the starting concentration of cDNA. These absolute quantities can be compared between disease classes (i.e. rejection vs. no-rejection) or can be taken as expression relative to a control gene to correct for variation among samples in sample collection, RNA purification and quantification, cDNA synthesis, and the PCR amplification.

Cell Type Specific Expression

[1061] Some markers are expressed only in specific types of cells. These markers may be useful markers for differentiation of rejection samples from no-rejection samples or may be used to identify differential expression of other markers in a single cell type. A specific marker for cytotoxic T-lymphocytes (such as CD8) can be used to identify differences in cell proportions in the sample. Other markers that are known to be expressed in this cell type can be compared to the level of CD8 to indicate differential gene expression within CD8 T-cells.

Control Genes for PCR

[1062] As discussed above, PCR expression measurements can be made as either absolute quantification of gene expression using a standard curve or relative expression of a gene of interest compared to a control gene. In the latter case, the gene of interest and the control gene are measured in the same sample. This can be done in separate reactions or in the same reaction (biplex format, see above). In either case, the final measurement for expression of a gene is expressed as a ratio of gene expression to control gene expression. It is important for a control gene to be constitutively expressed in the target tissue of interest and have minimal variation in expression on a per cell basis between individuals or between samples derived from an individual. If the gene has this type of expression behavior, the relative expression ratio will help correct for variability in the amount of sample RNA used in an assay. In addition, an ideal control gene has a high level of expression in the sample of interest compared to the genes being assayed. This is important if the gene of interest and control gene are used in a biplex format. The assay is set up so that the control gene reaches its threshold Ct value early and its amplification is limited by primers so that it does not compete for limiting reagents with the gene of interest.

[1063] To identify an ideal control gene for an assay, a number of genes were tested for variability between samples and expression in both mononuclear RNA samples and whole blood RNA samples using the RNA procurement and preparation methods and real-time PCR assays described above. 6 whole-blood and 6 mononuclear RNA samples from transplant recipients were tested. The intensity levels and variability of each gene in duplicate experiments on both sample types are shown in FIG. 13.

[1064] Based on criteria of low variability and high expression across samples, β-actin, 18s, GAPDH, b2 microglobulin were found to be good examples of control genes for the PAX samples. A single control gene may be incorporated as an internal biplex control is assays.

[1065] In examples 30 and 32, genes were identified as useful markers of cardiac and renal allograft rejection using microarrays. Genes identified through these studies are listed in Table 15A. In order to validate these findings, obtain a more precise measurement of expression levels and develop PCR reagents for diagnostic testing, real-time PCR assays were performed on samples from allograft recipients using primers to the identified genes. Gene specific PCR primers were developed and tested for all genes in Table 15A as described in example 28. These primer are listed in Table 15C and the sequence listing These primers were used to measure expression of the genes relative to β-actin or β-gus in 69 mononuclear RNA samples obtained from cardiac allograft recipients using Sybr green real-time PCR assays as described in example 28. Each sample was associated with an ISHLT cardiac rejection biopsy grade. The samples were tested in 2 phases. In phase I, 14 Grade 0, 1 Grade 1A, 3 Grade 2 and 9 Grade 3A samples were tested. In phase II, 19 Grade 2, 4 Grade 1B, 4 Grade 2 and 15 Grade 3A samples were tested. Data was analyzed for each phase individually and for the combined phase I+II sample set. These data are summarized in Table 17.

[1066] The average fold change in expression between rejection (3A) and no rejection (0) samples was calculated. A t-test was done to determine the significance with which each gene was differentially expressed between rejection and no rejection and a p-value was calculated. Genes with high average fold changes and low p-values are considered best candidates for further development as rejection markers. However, it is important to note that a gene with a low average fold change and a high p-value may still be a useful marker for rejection in some patients and may work as part of a gene expression panel to diagnose rejection. These same PCR data were used to create PCR gene expression panels for diagnosis of acute rejection as discussed in example 33.

[1067] Genes were identified which have expression patterns useful for the diagnosis and monitoring of acute cardiac allograft rejection. Further, sets of genes that work together in a diagnostic algorithm for allograft rejection were identified. Acute allograft rejection is a process that occurs in all solid organ transplantation including, heart, lung, liver, kidney, pancreas, pancreatic islet cell, intestine and others. Gene expression markers of acute cardiac rejection may be useful for diagnosis and monitoring of all allograft recipients. Patients, patient clinical data and patient samples used in the discovery of markers below were derived from a clinical study described in example 11.

[1068] The collected clinical data was used to define patient or sample groups for correlation of expression data. Patient groups were identified for comparison. For example, a patient group that possesses a useful or interesting clinical distinction, verses a patient group that does not possess the distinction. Measures of cardiac allograft rejection were derived from the clinical data to divide patients (and patient samples) into groups with higher and lower rejection activity over some period of time or at any one point in time. Such data were rejection grades as determined from histological reading of the cardiac biopsy specimens by a pathologist and data measuring progression of end-organ damage, including depressed left ventricular dysfunction (decreased cardiac output, decreased ejection fraction, clinical signs of low cardiac output) and usage of inotropic agents (Kobashigawa 1998).

[1069] Mononuclear RNA samples were collected and prepared from patients who had recently undergone a cardiac allograft transplantation using the protocol described in example 8. The allograft rejection status at the time of sample collection was determined by examination of cardiac biopsies as described in example 11 and as summarized here. 300 patient samples were included in the analysis. Each patient sample was associated with a biopsy and other clinical data collected at the time of the sample. The cardiac biopsies were graded by a pathologist at the local center and by three centralized pathologists who read the biopsy slides from all four local centers in a blinded manner.

[1070] Biopsy grades included 0, 1A, 1B, 2, 3A, and 3B. No grade 4 rejection was identified. Dependent variables were developed based on these grades using the local center pathology reading, the reading of a centralized and blinded pathologist, the highest of the readings, local or centralized and a consensus grade derived from all pathological readings. Samples were classified as no rejection or rejection in the following ways: Grade 0 vs. Grades 1–4, Grades 0 and 1A vs. Grades 1B–4, Grade 0 vs. Grade 3A, Grade 0 vs. Grades 1B–4, and Grade 0 vs. Grades 1B and 3A–4. Grade 0 samples were selected such that they were not immediately followed by an episode of acute rejection in the same patient. Comparing Grade 0 samples to Grade 3A samples gives the greatest difference between the rejection and no rejection groups on average.

[1071] Taking the highest of all pathologist readings has the effect of removing any sample from the no rejection class that was not a unanimous Grade 0. It also results in an increase in the number of rejection samples used in an analysis with the assumption that if a pathologist saw features of rejection, the call was likely correct and the other pathologists may have missed the finding. Many leading cardiac pathologists and clinicians believe that ISHLT grade 2 rejection does not represent significant acute rejection. Thus, for correlation analysis, exclusion of Grade 2 samples may be warranted.

[1072] Clinical data were also used to determine criteria for including samples in the analysis. For example, a patient with an active infection or in the early post-transplant period (ongoing surgical inflammation) might have immune activation unrelated to rejection and thus be difficult to identify as patients without rejection. The strictest inclusion criteria required that samples be from patients who did not have a bacterial or viral infection, were at least two weeks post cardiac transplant, were asymptomatic and were not currently admitted to the hospital.

[1073] After preparation of RNA (example 8), amplification, labeling, hybridization, scanning, feature extraction and data processing were done as described in Example 27, using the oligonucleotide microarrays described in Examples 20–22. The resulting log ratio of expression of Cy3 (patient sample)/Cy5 (R50 reference RNA) was used for analysis. Significance analysis for microarrays (SAM, Tusher 2001, Example 31) was used to discover genes that were differentially expressed between the rejection and no-rejection groups. Many different combinations of dependent variables, inclusion criteria, static/referenced, and data subsets were used in SAM analysis to develop the primary lists of genes significantly differentially expressed between rejection and no-rejection. As described in example 31, SAM assigns a false detection rate to each gene identified as differentially expressed. The most significant of these genes were identified.

[1074] An exemplary analysis was the comparison of Grade 0 samples to Grade 3A–4 samples using SAM. Data from the all the pathological readings was used to identify consensus Grade 0 samples and samples with at least one reading of Grade 3A or above. Using this definition of rejection and no rejection, expression profiles from rejection samples were compared to no rejection samples using SAM. The analysis identified 7 genes with a FDR of 1%, 15 genes @ 1.4%, 35 genes @ 3.9%. Many more genes were identified at higher FDR levels.

[1075] In table 18, a number of SAM analyses are summarized. In each case the highest grade from the 3 pathologists was taken for analysis. No rejection and rejection classes are defined. Samples are either used regardless of redundancy with respect to patients or a requirement is made that only one sample is used per patient or per patient per class. The number of samples used in the analysis is given and the lowest FDR achieved is noted. Those genes identified by SAM as candidate rejection markers are noted in Table 15A and B. SAM chooses genes as significantly different based on the magnitude of the difference between the groups and the variation among the samples within each group. It is important to note that a gene which is not identified by SAM as differentially expressed between rejection and no rejection may still be a useful rejection marker because: 1. The microarray technology is not adequately sensitive to detect all genes expressed at low levels. 2. A gene might be a useful member of a gene expression panel in that it is a useful rejection marker only in a subset of patients. This gene may not be significantly differentially expressed between all rejection and no rejection samples.

[1076] For the purposes of cross-validation of the results, the datasets were also divided into subsets to compare analysis between two subsets of roughly half of the data. The types of subsets constructed were as follows. First half/second half subsets were the first half of the samples and the second half of the samples from a dataset ordered by sample number. Odd/even subsets used the same source, a dataset ordered by sample number, but the odd subset consisted of every 2^nd sample starting with the first and the even subset consisted of every 2^nd sample starting with the second sample, Center 14/other subsets were the same datasets, divided by transplant hospital. The center 14 subset consisted of all samples from patients at center 14, while the other subset consisted of all samples from the other three centers (12,13, and 15). When a gene was found to be significantly differentially expressed in both sets of data, a higher priority was put on that gene for development of a diagnostic test. This was reflected in a “Array Score” value (Table 15B) that also considered the false detection rate for the gene and the importance of the gene in classification models (see example 33).

[1077] Alternatively one can divide samples into 10 equal parts and do 10-fold cross validation of the results of SAM.

[1078] Microarray data was also used to generate classification models for diagnosis of rejection as described in example 33. Genes identified through classification models as useful in the diagnosis of rejection are noted in Table 15B in the column “models”.

[1079] As genes were identified as useful rejection markers by microarray significance analysis, classification models, PCR analysis, or through searching the prior art, a variety of approaches were employed to discover genes that had similar expression behavior (coexpression) to the gene of interest. If a gene is a useful rejection marker, then a gene that is identified as having similar expression behavior is also likely to be a useful rejection marker. Hierarchical clustering (Eisen et al. 1998, see example 31) was used to identify co-expressed genes for established rejection markers. Genes were identified from the nearest branches of the clustering dendrogram. Gene expression profiles generated from 240 samples derived from transplant recipients were generated as described above. Hierarchical clustering was performed and co-expressed genes of rejection markers were identified. An example is shown in FIG. 14. SEQ ID NO:4460 was shown to be significantly differentially expressed between rejection and no rejection using both microarrays and PCR. Gene SEQ ID NO:6514 was identified by hierarchical clustering as closely co-expressed with SEQ ID NO:4460. In table 15B, genes identified as co-expressed with established markers are identified as such by listing the SEQ ID that they are co-expressed with in the column labeled “clusters”.

[1080] Primers for real-time PCR validation were designed for each of the marker genes as described in Example 28 and are listed in Table 15C and the sequence listing. PCR expression measurements using these primers were used to validate array findings, more accurately measure differential gene expression and create PCR gene expression panels for diagnosis of rejection as described in example 33.

[1081] After generation and processing of expression data sets from microarrays as described in Example 23, a log ratio value is used for most subsequent analysis. This is the logarithm of the expression ratio for each gene between sample and universal reference. The processing algorithm assigns a number of flags to data that are of low signal to noise, saturated signal or are in some other way of low or uncertain quality. Correlation analysis can proceed with all the data (including the flagged data) or can be done on filtered data sets where the flagged data is removed from the set. Filtered data should have less variability and noise and may result in more significant or predictive results. Flagged data contains all information available and may allow discovery of genes that are missed with the filtered data set.

[1082] After filtering the data for quality as described above and in example 23, missing data are common in microarray data sets. Some algorithms don't require complete data sets and can thus tolerate missing values. Other algorithms are optimal with or require imputed values for missing data. Analysis of data sets with missing values can proceed by filtering all genes from the analysis that have more than 5%, 10%, 20%, 40%, 50%, 60% or other % of values missing across all samples in the analysis. Imputation of data for missing values can be done by a variety of methods such as using the row mean, the column mean, the nearest neighbor or some other calculated number. Except when noted, default settings for filtering and imputation were used to prepare the data for all analytical software packages.

[1083] In addition to expression data, clinical data are included in the analysis. Continuous variables, such as the ejection fraction of the heart measured by echocardiography or the white blood cell count can be used for correlation analysis. Any piece of clinical data collected on study subjects can be used in a correlation or classification analysis. In some cases, it may be desirable to take the logarithm of the values before analysis. These variables can be included in an analysis along with gene expression values, in which case they are treated as another “gene”. Sets of markers can be discovered that work to diagnose a patient condition and these can include both genes and clinical parameters. Categorical variables such as male or female can also be used as variables for correlation analysis. For example, the sex of a patient may be an important splitter for a classification tree.

[1084] Clinical data are used as supervising vectors (dependent variables) for the significance or classification analysis of expression data. In this case, clinical data associated with the samples are used to divide samples in to clinically meaningful diagnostic categories for correlation or classification analysis. For example, pathologic specimens from kidney biopsies can be used to divide lupus patients into groups with and without kidney disease. A third or more categories can also be included (for example “unknown” or “not reported”). After generation of expression data and definition of supervising vectors, correlation, significance and classification analysis are used to determine which set of genes and set of genes are most appropriate for diagnosis and classification of patients and patient samples.

[1085] Two main types of expression data analyses are commonly performed on the expression data with differing results and purposes. The first is significance analyses or analyses of difference. In this case, the goal of the analysis is to identify genes that are differentially expressed between sample groups and to assign a statistical confidence to those genes that are identified. These genes may be markers of the disease process in question and are further studied and developed as diagnostic tools for the indication.

[1086] The second major type of analysis is classification analysis. While significance analysis identifies individual genes that are differentially expressed between sample groups, classification analysis identifies gene sets and an algorithm for their gene expression values that best distinguish sample (patient) groups. The resulting gene expression panel and algorithm can be used to create and implement a diagnostic test. The set of genes and the algorithm for their use as a diagnostic tool are often referred to herein as a “model”. Individual markers can also be used to create a gene expression diagnostic model. However, multiple genes (or gene sets) are often more useful and accurate diagnostic tools.

Significance Analysis for Microarrays (SAM)

[1087] Significance analysis for microarrays (SAM) (Tusher 2001) is a method through which genes with a correlation between their expression values and the response vector are statistically discovered and assigned a statistical significance. The ratio of false significant to significant genes is the False Discovery Rate (FDR). This means that for each threshold there are some number of genes that are called significant, and the FDR gives a confidence level for this claim. If a gene is called differentially expressed between two classes by SAM, with a FDR of 5%, there is a 95% chance that the gene is actually differentially expressed between the classes. SAM will identify genes that are differentially expressed between the classes. The algorithm selects genes with low variance within a class and large variance between classes. The algorithm may not identify genes that are useful in classification, but are not differentially expressed in many of the samples. For example, a gene that is a useful marker for disease in women and not men, may not be a highly significant marker in a SAM analysis, but may be useful as part of a gene set for diagnosis of a multi-gene algorithm.

[1088] After generation of data from patient samples and definition of categories using clinical data as supervising vectors, SAM is used to detect genes that are likely to be differentially expressed between the groupings. Those genes with the highest significance can be validated by real-time PCR (Example 29) or can be used to build a classification algorithm as described here.

Classification

[1089] Classification algorithms are used to identify sets of genes and formulas for the expression levels of those genes that can be applied as diagnostic and disease monitoring tests. The same classification algorithms can be applied to all types of expression and proteomic data, including microarray and PCR based expression data. Examples of classification models are given in example 33. The discussion below describes the algorithms that were used and how they were used.

[1090] Classification and Regression Trees (CART) is a decision tree classification algorithm (Breiman 1984). From gene expression and or other data, CART can develop a decision tree for the classification of samples. Each node on the decision tree involves a query about the expression level of one or more genes or variables. Samples that are above the threshold go down one branch of the decision tree and samples that are not go down the other branch. Genes from expression data sets can be selected for classification building with CART by significant differential expression in SAM analysis (or other significance test), identification by supervised tree-harvesting analysis, high fold change between sample groups, or known relevance to classification of the target diseases. In addition, clinical data can be used as independent variables for CART that are of known importance to the clinical question or are found to be significant predictors by multivariate analysis or some other technique. CART identifies predictive variables and their associated decision rules for classification (diagnosis). CART also identifies surrogates for each splitter (genes that are the next best substitute for a useful gene in classification). Analysis is performed in CART by weighting misclassification costs to optimize desired performance of the assay. For example, it may be most important that the sensitivity of a test for a given diagnosis be >90%. CART models can be built and tested using 10 fold cross-validation or v-fold cross validation (see below). CART works best with a smaller number of variables (5–50).

[1091] Multiple Additive Regression Trees (Friedman, J H 1999, MART) is similar to CART in that it is a classification algorithm that builds decision trees to distinguish groups.

[1092] MART builds numerous trees for any classification problem and the resulting model involves a combination of the multiple trees. MART can select variables as it build models and thus can be used on large data sets, such as those derived from an 8000 gene microarray. Because MART uses a combination of many trees and does not take too much information from any one tree, it resists over training. MART identifies a set of genes and an algorithm for their use as a classifier.

[1093] A Nearest Shrunken Centroids Classifier can be applied to microarray or other data sets by the methods described by Tibshirani et al. 2002. This algorithms also identified gene sets for classification and determines their 10 fold cross validation error rates for each class of samples. The algorithm determines the error rates for models of any size, from one gene to all genes in the set. The error rates for either or both sample classes can are minimized when a particular number of genes are used. When this gene number is determined, the algorithm associated with the selected genes can be identified and employed as a classifier on prospective sample.

[1094] For each classification algorithm and for significance analysis, gene sets and diagnostic algorithms that are built are tested by cross validation and prospective validation. Validation of the algorithm by these means yields an estimate of the predictive value of the algorithm on the target population. There are many approaches, including a 10 fold cross validation analysis in which 10% of the training samples are left out of the analysis and the classification algorithm is built with the remaining 90%. The 10% are then used as a test set for the algorithm. The process is repeated 10 times with 10% of the samples being left out as a test set each time. Through this analysis, one can derive a cross validation error which helps estimate the robustness of the algorithm for use on prospective (test) samples. Any % of the samples can be left out for cross validation (v-fold cross validation, LOOCV). When a gene set is established for a diagnosis with an acceptable cross validation error, this set of genes is tested using samples that were not included in the initial analysis (test samples). These samples may be taken from archives generated during the clinical study. Alternatively, a new prospective clinical study can be initiated, where samples are obtained and the gene set is used to predict patient diagnoses.

[1095] Acute allograft rejection involves activation of recipient leukocytes and infiltration into the rejecting organ. For example, CD8 T-cells are activated by CD4 T-cells and enter the allograft where they destroy graft tissue. These activated, graft-associated leukocytes may reside in the graft, die or exit the graft. Upon exiting, the cells can find their way into the urine or blood (in the case of renal allografts), bile or blood (liver allografts) or blood (cardiac allografts). These activated cells have specific gene expression patterns that can be measured using microarrays, PCR or other methods. These gene expression patterns can be measured in the graft tissue (graft associated leukocytes), blood leukocytes, urine leukocytes or stool/biliary leukocytes. Thus graft associated leukocyte gene expression patterns are used to discover markers of activated leukocytes that can be measured outside the graft for diagnostic testing.

[1096] Renal biopsy and cardiac biopsy tissue specimens were obtained for gene expression profiling. The specimens were obtained at the time of allograft biopsy and were preserved by flash freezing in liquid nitrogen using standard approaches or immersion in an RNA stabilization reagent as per the manufacturers recommendation (RNAlater, Qiagen, Valencia, Calif.). Biopsy allograft pathological evaluation was also obtained and samples were classified as having a particular ISHLT rejection grade (for cardiac) or acute rejection, chronic rejection, acute tubular necrosis or no disease (for renal).

[1097] 28 renal biopsy tissue samples were transferred to RLT buffer, homogenized and RNA was prepared using RNeasy preparation kits (Qiagen, Valencia, Calif.). Average total RNA yield was 1.3 ug. Samples were subjected to on column DNAse digestion. 18 samples were derived from patients with ongoing acute allograft rejection and 10 were from controls with chronic rejection or acute renal failure.

[1098] RNA from the samples was used for amplification, labeling and hybridization to leukocyte arrays (example 27). Significance analysis for microarrays (SAM, Tusher 2001, Example 31) was used to identify genes that were differentially expressed between the acute rejection samples and controls. Leukocyte markers of acute rejection that are associated with the graft should be genes that are expressed at some level in activated leukocytes. Since leukocytes appear in graft tissue with some frequency with acute rejection, leukocyte genes associate with rejection are identified by SAM as upregulated in acute rejection in this experiment. 35 genes were identified as upregulated in acute rejection by SAM with less than a 5% false detection rate and 139 were detected with <10.0% FDR. Results of this analysis are shown in Table 19.

[1099] For each of these genes, to 50mer oligonucleotide sequence was used to search NCBI databases including Unigene and OMIM. Genes were identified by sequence analysis to be either known leukocyte specific markers, known leukocyte expressed markers, known not to be leukocyte expressed or expression unknown. This information helped selected candidate leukocyte markers from all upregulated genes. This is necessary because some of the upregulated genes may have been expressed by renal tissue. Those genes that are leukocyte specific or leukocyte expressed were selected for evaluation by PCR in urine and blood samples from patients with and without acute allograft rejection (cardiac and renal). These genes are useful expression markers of acute rejection in allograft tissue specimens and may also be useful gene expression markers for the process in circulating leukocytes, or urine leukocytes. Genes with known leukocyte expression are noted in Table 19. In addition, some of the leukocyte expressed genes from this analysis were selected for PCR validation and development for diagnosis of acute cardiac rejection and are noted in Table 15B in the column “KTx”.

[1100] Five cardiac rejection markers in the peripheral blood were assayed using real-time PCR in renal biopsy specimens. The average fold change for these genes between acute rejection (n=6) and controls (n=6) is given below. Work is ongoing to increase the number of samples tested and the significance of the results.

[1101] PCR assays of cardiac rejection peripheral blood markers in renal allograft tissue. R=rejection, NR=No rejection.

[1102]

	Gene	Fold change (R/NR)
	Granzyme B	2.16
	CD20	1.42
	NK cell receptor	1.72
	T-box 21	1.74
	IL4	1.3

[1103] Markers of renal rejection that are secreted from cells may be measured in the urine or serum of patients as a diagnostic or screening assay for rejection. Genes with lower molecular weight are most likely to be filtered into the urine to be measured in this way. Standard immunoassays may be used to measure these proteins. In table 19, genes that are known to be secreted are noted.

Array Panels/Classification Models

[1104] Using the methods of the invention, gene expression panels were discovered for screening and diagnosis of acute allograft rejection. Gene expression panels can be implemented for diagnostic testing using any one of a variety of technologies, including, but not limited to, microarrays and real-time PCR.

[1105] Using peripheral blood mononuclear cell RNA that was collected and prepared from cardiac allograft recipients as described in examples 8 and 11, leukocyte gene expression profiles were generated and analyzed using microarrays as described in examples 23, 24 and 26. 300 samples were analyzed. ISHLT rejection grades were used to divide patients into classes of rejection and no rejection. Multiple Additive Regression Trees (MART, Friedman, J H 1999, example 31) was used to build a gene expression panel and algorithm for the diagnosis of rejection with high sensitivity. Default settings for the implementation of MART called TreeNet 1.0 (Salford Systems, San Diego, Calif.) were used except where noted.

[1106] 82 Grade 0 (rejection) samples and 76 Grade 1B–4 (no rejection) samples were divided into training (80% of each class) and testing (20% of each class) sets. A MART algorithm was then developed on the training set to distinguish rejection from no rejection samples using a cost of 1.02:1 for misclassification of rejection as no rejection. The resulting algorithm was then used to classify the test samples. The algorithm correctly classified 51 of 66 (77%) no rejection samples in the training set and 9 of 16 (56%) no rejection samples in the test set. For rejection samples 64 of 64 (100%) were correctly classified in the training set and 12 of 12 were correctly classified in the test set. The algorithm used 37 genes. MART ranks genes by order of importance to the model. In order, the 37 genes were: SEQ IDs: 4759, 2516, 591, 6514, 1073, 2427, 1617, 3755, 2626, 3382, 1721, 4494, 2080, 3393, 459, 2033, 640, 2011, 7230, 3241, 4085, 6936, 3347, 2416, 8112, 4972, 290, 5187, 6341, 7508, 7803, 6882, 6144, 8108, 7413.

[1107] Another MART model was built by excluding samples derived from patients in the first month post transplant and from patients with known CMV infection. 20 Grade 0 (rejection) samples and 25 Grade 1B–4 (no rejection) samples were divided into training (80% of each class) and testing (20% of each class) sets. A MART algorithm was then developed on the training set to distinguish rejection from no rejection samples using default settings. The resulting algorithm was then used to classify the test samples. The algorithm correctly classified 100% of samples of both classes in the training and testing sets. However, this model required 169 genes. The sample analysis was done a second time with the only difference being requirement that all decision trees in the algorithm be composed of two nodes (single decision, “stump model”). In this case 15/16 no rejection samples were correctly identified in the training set and 4/4 no rejection samples were correctly identified in the test set. For the rejection samples, 17/19 were correctly identified in the training set and 5/6 were correctly classified in the test set. This model required 23 genes. In order of importance, they were: SEQ IDs: 2715, 758, 2168, 7652, 1991, 7388, 4517, 7264, 3501, 5893, 3399, 4582, 8018, 2271, 3508, 2961, 2907, 1430, 7554, 838, 8049, 8093, 8055.

Real-Time PCR Panels/Classification Models

[1108] PCR primers were developed for top rejection markers and used in real-time PCR assays on transplant patient samples as described in examples 28 and 29. This data was used to build PCR gene expression panels for diagnosis of rejection. Using MART (example 31) a 10-fold cross validated model was created to diagnose rejection using 12 no rejection samples (grade 0) and 10 rejection samples (grade 3A). Default settings were used with the exception of assigning a 1.02:1 cost for misclassification of rejection as no rejection and requirement that all decision trees be limited to 2 nodes (“stump model”). 20 genes were used in the model, including: SEQ IDs 3305, 3096, 1956, 5280, 7298, 99, 4505, 259, 2855, 1984, 1975, 90, 6112, 2947, 2287, 3707, 281, 6091, 3709, 6514. The 10-fold cross-validated sensitivity for rejection was 100% and the specificity was 85%. PCR primers for the genes are listed in Table 15C and the sequence listing.

[1109] A different analysis of the PCR data was performed using the nearest shrunken centroids classifier (Tibshirani et al. 2002; PAM version 1.01, see example 31). A 10-fold cross validated model was created to diagnose rejection using 13 no rejection samples (grade 0) and 10 rejection samples (grade 3A). Default settings were used with the exception of using a prior probability setting of (0.5, 0.5). The algorithm derives algorithms using any number of the genes. A 3-gene model was highly accurate with a 10 fold cross-validated sensitivity for rejection of 90%, and a specificity of 85%.

[1110] The 3 genes used in this model were: SEQ IDs 4761, 90, and 7274. The PCR primers used are given in Table 15C and the sequence listing. An ROC curve was plotted for the 3-gene model and is shown in FIG. 16.

[1111] In order to show that XDx's leukocyte-specific markers can be detected in whole blood, we collected whole blood RNA using the PAXgene whole blood collection, stabilization, and RNA isolation kit (PreAnalytix). Varying amounts of the whole blood RNA were used in the initial RT reaction (1, 2, 4, and 8 ug), and varying dilutions of the different RT reactions were tested (1:5, 1:10, 1:20, 1:40, 1:80, 1:160). We did real-time PCR assays with primers specific to XDx's markers and showed that we can reliably detect these markers in whole blood.

[1112] Total RNA was prepared from 14 mononuclear samples (CPT, BD) paired with 14 whole blood samples (PAXgene, PreAnalytix) from transplant recipients. cDNA was prepared from each sample using 2 ug total RNA as starting material. Resulting cDNA was diluted 1:10 and Sybr green real-time PCR assays were performed.

[1113] For real-time PCR assays, Ct values of 15–30 are desired for each gene. If a gene's Ct value is much above 30, the result may be variable and non-linear. For PAX sample, target RNA will be more dilute than in CPT samples. cDNA dilutions must be appropriate to bring Ct values to less than 30.

[1114] Ct values for the first 5 genes tested in this way are shown in the table below for both whole blood RNA (PAX) and mononuclear RNA (CPT).

[1115] With one exception, the

[1116]

	Gene	Ct PAX	Ct CPT
	CD20	27.41512	26.70474
	4761	28.45656	26.52635
	3096	29.09821	27.83281
	GranzymeB	31.18779	30.56954
	IL4	33.11774	34.8002
	Actin	19.17622	18.32966
	B-GUS	26.89142	26.92735

genes have higher Ct values in whole blood. Using this protocol, all genes can be detected with Cts<35. For genes found to have Ct values above 30 in target samples, less diluted cDNA may be needed.

[1117] Gene products that are secreted from cells or expressed as surface proteins have special diagnostic utility in that an assay may be developed to detect relative quantities of proteins in blood plasma or serum. Secreted proteins may also be detectable in urine, which may be a useful sample for the detection of rejection in renal allograft recipients. Cell surface markers may be detected using antigen specific antibodies in ELISA assays or using flow string techniques such as FACS.

[1118] Each gene that is found to be differentially regulated in one population of patients has several potential applications. It may be a target for new pharmaceuticals, a diagnostic marker for a condition, a benchmark for titrating drug delivery and clearance, or used in screening small molecules for new therapeutics. Any of these applications may be improved by an understanding of the physiologic function and localization of the gene product in vivo and by relating those functions to known diseases and disorders. Identifying the basic function of each candidate gene helps identify the signaling or metabolic pathways the gene is a part of, leading us to investigate other members of those pathways as potential diagnostic markers or targets of interest to drug developers.

[1119] For each of the markers in table 15A, we attempted to identify the basic function and subcellular localization of the gene. These results are summarized in Table 20. In addition to the steps outlined in examples 3–6, information was obtained from the following public resources: Online Mendelian Inheritance in Man at the NCBI, LocusLink at the NCBI, the SWISS-PROT database (www.expasy.org/sprot/), and Protein Reviews on the Web (www.ncbi.nlm.nih.gov/PROW/). For each marker represented by a curated reference mRNA from the RefSeq project, the corresponding reference protein accession number is listed. Curated sequences are those that have been manually processed by NCBI staff to represent the best estimate of the mRNA sequence as it is transcribed, based on alignments of draft DNA sequence, predicted initiation, termination and splice sites, and submissions of EST and full-length mRNA sequences from the scientific community.

[1120]

TABLE 1
Disease Classification	Disease/Patient Group
Cardiovascular Disease	Atherosclerosis
	Unstable angina
	Myocardial Infarction
	Restenosis after angioplasty
	Congestive Heart Failure
	Myocarditis
	Endocarditis
	Endothelial Dysfunction
	Cardiomyopathy
	Cardiovascular drug use
Endocrine Disease	Diabetes Mellitus I and II
	Thyroiditis
	Addisson's Disease
Infectious Disease	Hepatitis A, B, C, D, E, G
	Malaria
	Tuberculosis
	HIV
	Pneumocystis Carinii
	Giardia
	Toxoplasmosis
	Lyme Disease
	Rocky Mountain Spotted Fever
	Cytomegalovirus
	Epstein Barr Virus
	Herpes Simplex Virus
	Clostridium Dificile Colitis
	Meningitis (all organisms)
	Pneumonia (all organisms)
	Urinary Tract Infection (all organisms)
	Infectious Diarrhea (all organisms)
	Anti-infectious drug use
Angiogenesis	Pathologic angiogenesis
	Physiologic angiogenesis
	Treatment induced angiogenesis
	Pro or anti-angiogenic drug use
Inflammatory/Rheumatic	Rheumatoid Arthritis
	Systemic Lupus Erythematosis
	Sjogrens Disease
	CREST syndrome
	Scleroderma
	Ankylosing Spondylitis
	Crohn's
	Ulcerative Colitis
	Primary Sclerosing Cholangitis
Inflammatory/Rheumatic	Appendicitis
	Diverticulitis
	Primary Biliary Sclerosis
	Wegener's Granulomatosis
	Polyarteritis nodosa
	Whipple's Disease
	Psoriasis
	Microscopic Polyanngiitis
	Takayasu's Disease
	Kawasaki's Disease
	Autoimmune hepatitis
	Asthma
	Churg-Strauss Disease
	Beurger's Disease
	Raynaud's Disease
	Cholecystitis
	Sarcoidosis
	Asbestosis
	Pneumoconioses
	Antinflammatory drug use
Transplant Rejection	Heart
	Lung
	Liver
	Pancreas
	Bowel
	Bone Marrow
	Stem Cell
	Graft versus host disease
	Transplant vasculopathy
	Skin
	Cornea
	Immunosupressive drug use
Malignant Disorders	Leukemia
	Lymphoma
	Carcinoma
	Sarcoma
Neurological Disease	Alzheimer's Dementia
	Pick's Disease
	Multiple Sclerosis
	Guillain Barre Syndrome
	Peripheral Neuropathy

[1121]

TABLE 2
Candidate genes, Database mining
Unigene clusters are listed.
Cluster numbers are defined as in Unigene build #133 uploaded on: Fri Apr. 20, 2001
CD50	Hs.99995	Homo sapiens cAMP responsive element	Hs.79194
		binding protein 1 (CREB1) mRNA.
CD70 = CD27L	Hs.99899	Nucleolin (NCL)	Hs.79110
MDC	Hs.97203	MAPK14	Hs.79107
CD3z	Hs.97087	CD100	Hs.79089
CD19	Hs.96023	OX-2	Hs.79015
	Hs.95388	PCNA	Hs.78996
CD3d	Hs.95327		Hs.78909
	Hs.9456	GRO-a	Hs.789
interleukin 6	Hs.93913	CDw32A	Hs.78864
phospholipaseA2	Hs.93304	H. sapiens mRNA for herpesvirus associated	Hs.78683
		ubiquitin-specific protease (HAUSP).
Human mRNA for KIAA0128 gene, partial	Hs.90998	CD41b = LIBS1	Hs.785
cds.
CD48	Hs.901	ANXA1 (LPC1)	Hs.78225
heat shock 70 kD protein 1A	Hs.8997	CD31	Hs.78146
TxA2 receptor	Hs.89887	Homo sapiens TERF1 (TRF1)-interacting	Hs.7797
		nuclear factor 2 (TINF2), mRNA.
fragile X mental retardation protein (FMR-	Hs.89764	major histocompatibility complex, class I, B	Hs.77961
1)
CD20	Hs.89751	LOX1	Hs.77729
ENA-78	Hs.89714	major histocompatibility complex, class II,	Hs.77522
		DM alpha
IL-2	Hs.89679	CD64	Hs.77424
CD79b	Hs.89575	CD71	Hs.77356
CD2	Hs.89476		Hs.77054
SDF-1 = CXCR4	Hs.89414	HLA-DRA	Hs.76807
CD61	Hs.87149	CD105	Hs.76753
IFN-g	Hs.856		Hs.76691
CD34	Hs.85289	TNF-alpha	Hs.76507
CD104	Hs.85266	LCP1	Hs.76506
CD8	Hs.85258	TMSB4X	Hs.75968
IGF-1	Hs.85112	PAI2	Hs.75716
CD103	Hs.851	MIP-1b	Hs.75703
IL-13	Hs.845	CD58	Hs.75626
RPA1	Hs.84318	CD36	Hs.75613
CD74	Hs.84298	hnRNP A2/hnRNP B1	Hs.75598
CD132	Hs.84	CD124	Hs.75545
CD18	Hs.83968	MIP-3a	Hs.75498
Cathepsin K	Hs.83942	beta-2-microglobulin	Hs.75415
CD80	Hs.838	FPR1	Hs.753
CD46	Hs.83532	Topo2B	Hs.75248
NFKB1	Hs.83428	interleukin enhancer binding factor 2, 45 kD	Hs.75117
IL-18	Hs.83077	chloride intracellular channel 1	Hs.74276
interleukin 14	Hs.83004	EGR3	Hs.74088
L-selectin = CD62L	Hs.82848	MIP-1a	Hs.73817
CD107b	Hs.8262	CD62P = p-selectin	Hs.73800
CD69	Hs.82401	CD21	Hs.73792
CD95	Hs.82359	APE	Hs.73722
CD53	Hs.82212	IL12Rb2	Hs.73165
Human lymphocyte specific interferon	Hs.82132	NFKB2	Hs.73090
regulatory factor/interferon regulatory factor
4 (LSIRF/IRF4) mRNA, complete cds.
IL-16	Hs.82127	I-309	Hs.72918
DUT	Hs.82113	immunoglobulin superfamily, member 4	Hs.70337
CDw121a	Hs.82112	IL-3	Hs.694
PAI-1	Hs.82085		Hs.6895
TGF-bR2	Hs.82028	NTH1	Hs.66196
CD117	Hs.81665	CD40L	Hs.652
HLA-DPB1	Hs.814	IL-11R	Hs.64310
NFKBIA	Hs.81328	Homo sapiens toll-like receptor 2 (TLR2)	Hs.63668
		mRNA.
CD6	Hs.81226	ferritin H chain	Hs.62954
IL-1 RA	Hs.81134	IL8	Hs.624
UBE2B (RAD6B)	Hs.811	Tissue Factor	Hs.62192
Lyn	Hs.80887	F-box only protein 7	Hs.5912
STAT4	Hs.80642	CD5	Hs.58685
UBE2A (RAD6A)	Hs.80612	guanine nucleotide binding protein (G	Hs.5662
		protein), beta polypeptide 2-like 1
Fractalkine	Hs.80420	SCYA11	Hs.54460
IK cytokine, down-regulator of HLA II	Hs.8024	IK1	Hs.54452
	Hs.79933	CCR1	Hs.516
CD79a	Hs.79630	Homo sapiens TRAIL receptor 2 mRNA,	Hs.51233
		complete cds.
	Hs.7942	CD11c	Hs.51077
nuclear factor, interleukin 3 regulated	Hs.79334	CD66a	Hs.50964
CD83	Hs.79197	JAK1	Hs.50651
DC-CK1	Hs.16530	Homo sapiens programmed cell death 4	Hs.100407
		(PDCD4), mRNA.
CCR7	Hs.1652	SCYB13 (CXCL13)	Hs.100431
TLR4	Hs.159239	SMAD7	Hs.100602
EST	Hs.158975	RAD51L1 (RAD51B)	Hs.100669
EST	Hs.158966	PPARG	Hs.100724
EST	Hs.158965	transcription factor 3 (E2A immunoglobulin	Hs.101047
		enhancer binding factors E12/E47)
EST	Hs.158943	major histocompatibility complex, class I-	Hs.101840
		like sequence
EST	Hs.158894	immunoglobulin superfamily containing	Hs.102171
		leucine-rich repeat
EST	Hs.158877	CD166	Hs.10247
EST	Hs.157815	fibroblast tropomyosin TM30 (pl)	Hs.102824
EST	Hs.157813	interleukin 1 receptor-like 2	Hs.102865
ESTs	Hs.157569	GTF2H4	Hs.102910
immunoglobulin kappa constant	Hs.156110		Hs.10326
INPP5D	Hs.155939	Human ITAC (IBICK)	Hs.103982
C3AR1	Hs.155935	novel protein with MAM domain	Hs.104311
PRKDC	Hs.155637	ESTs, Weakly similar to interleukin	Hs.105125
		enhancer binding factor 2 [H. sapiens]
MHC class II HLA-DRw53-associated	Hs.155122	Homo sapiens clone 24686 mRNA	Hs.105509
glycoprotein		sequence.
CD73	Hs.153952		Hs.105532
CD37	Hs.153053	Homo sapiens granulysin (GNLY),	Hs.105806
		transcript variant 519, mRNA.
IFNAR1	Hs.1513	CD77	Hs.105956
Homo sapiens solute carrier family 21	Hs.14805	RD RNA-binding protein	Hs.106061
(organic anion transporter), member 11
(SLC21A11), mRNA.
EST	Hs.146627		Hs.106673
SET translocation (myeloid leukemia-	Hs.145279		Hs.10669
associated)
EST	Hs.144119	Homo sapiens clone 24818 mRNA	Hs.106823
		sequence.
ESTs	Hs.143534		Hs.106826
STAT3	Hs.142258		Hs.10712
CD96	Hs.142023		Hs.107149
CD23	Hs.1416	hypothetical protein	Hs.10729
EGR2	Hs.1395	Tachykinin Receptor 1	Hs.1080
CDw84	Hs.137548	glycophorin A	Hs.108694
CD55	Hs.1369	Histone H1x	Hs.109804
EST	Hs.135339	CD66d	Hs.11
GM-CSF	Hs.1349	interleukin 17	Hs.110040
EST	Hs.133175		Hs.110131
CD1a	Hs.1309	major histocompatibility complex, class I, F	Hs.110309
CD10	Hs.1298	REV1	Hs.110347
HVEM	Hs.129708	HCR	Hs.110746
C9	Hs.1290	VWF	Hs.110802
C6	Hs.1282	high affinity immunoglobulin epsilon	Hs.11090
		receptor beta subunit
C1R	Hs.1279	interleukin 22 receptor	Hs.110915
IL-1b	Hs.126256		Hs.110978
CD9	Hs.1244	Homo sapiens ubiquitin specific protease 6	Hs.111065
		(Tre-2 oncogene) (USP6), mRNA.
	Hs.12305		Hs.111128
Homo sapiens Vanin 2 (VNN2) mRNA.	Hs.121102	MMP2	Hs.111301
Hsp10	Hs.1197	major histocompatibility complex, class II,	Hs.11135
		DN alpha
CD59	Hs.119663	LTBR	Hs.1116
CD51	Hs.118512	ESTs, Weakly similar to A41285	Hs.111941
		interleukin enhancer-binding factor ILF-1
		[H. sapiens]
CD49a	Hs.116774	Homo sapiens STRIN protein (STRIN),	Hs.112144
		mRNA.
CD72	Hs.116481	MSH5	Hs.112193
HLA-DMB	Hs.1162	TCRg	Hs.112259
MCP-4	Hs.11383		Hs.11307
	Hs.111554	CMKRL2	Hs.113207
ferritin L chain	Hs.111334	CCR8	Hs.113222
TGF-b	Hs.1103	LILRA3	Hs.113277
Homo sapiens ras homolog gene family,	Hs.109918	Human CXCR-5 (BLR-1)	Hs.113916
member H (ARHH), mRNA.
lysosomal alpha-mannosidase (MANB)	Hs.108969	RAD51C	Hs.11393
	Hs.108327	myosin, heavy polypeptide 8, skeletal	Hs.113973
		muscle, perinatal
granzyme B	Hs.1051	CD42a	Hs.1144
HCC-4	Hs.10458	TNFRSF11A	Hs.114676
	Hs.10362		Hs.114931
	Hs.102630	MSH4	Hs.115246
	Hs.101382	Homo sapiens dendritic cell	Hs.115515
		immunoreceptor (DCIR), mRNA.
C4BPA	Hs.1012	REV3L (POLZ)	Hs.115521
CD125	Hs.100001	JAK2	Hs.115541
TERF2	Hs.100030	OPG ligand	Hs.115770
LIG3	Hs.100299	PCDH12	Hs.115897
	Hs.157489		Hs.166235
EST	Hs.157560	POLE1	Hs.166846
EST	Hs.157808	regulatory factor X, 5 (influences HLA class	Hs.166891
		II expression)
EST	Hs.157811	PIG-F (phosphatidyl-inositol-glycan class	Hs.166982
		F)
	Hs.158127	ESTs, Moderately similar to	Hs.167154
		ILF1_HUMAN INTERLEUKIN
		ENHANCER-BINDING FACTOR 1
		[H. sapiens]
interleukin 18 receptor accessory protein	Hs.158315	HLA-DRB6	Hs.167385
CCR3	Hs.158324	ret finger protein-like 3	Hs.167751
Human DNA sequence from clone CTA-	Hs.158352	CD56	Hs.167988
390C10 on chromosome 22q11.21–12.1
Contains an Immunoglobulin-like gene and
a pseudogene similar to Beta Crystallin,
ESTs, STSs, GSSs and taga and tat repeat
polymorphisms
ESTs	Hs.158576	RBT1	Hs.169138
	Hs.158874	APOE	Hs.169401
EST	Hs.158875		Hs.16944
EST	Hs.158876		Hs.169470
EST	Hs.158878	MMP12	Hs.1695
EST	Hs.158956	CD161	Hs.169824
EST	Hs.158967	tenascin XB	Hs.169886
EST	Hs.158969		Hs.170027
EST	Hs.158971		Hs.170150
EST	Hs.158988	C4A	Hs.170250
CD120a = TNFR-1	Hs.159	TP53BP1	Hs.170263
EST	Hs.159000	ESTs	Hs.170274
	Hs.159013	ESTs, Weakly similar to ALU1_HUMAN	Hs.170338
		ALU SUBFAMILY J SEQUENCE
		CONTAMINATION WARNING
		ENTRY□ [H. sapiens]
EST	Hs.159025	ESTs	Hs.170578
EST	Hs.159059	EST	Hs.170579
IL18R1	Hs.159301	ESTs	Hs.170580
ftp-3	Hs.159494	EST	Hs.170581
CASP8	Hs.159651	ESTs	Hs.170583
EST	Hs.159655	EST	Hs.170586
EST	Hs.159660	EST	Hs.170588
EST	Hs.159678	EST	Hs.170589
kallikrein 12 (KLK12)	Hs.159679		Hs.170772
EST	Hs.159682	ESTs	Hs.170786
EST	Hs.159683	EST	Hs.170909
EST	Hs.159693	EST	Hs.170912
EST	Hs.159706	EST	Hs.170933
EST	Hs.159718	ESTs	Hs.171004
SPO11	Hs.159737	EST	Hs.171095
EST	Hs.159754	EST	Hs.171098
EST	Hs.160401	ESTs	Hs.171101
EST	Hs.160405	EST	Hs.171108
EST	Hs.160408	ESTs	Hs.171110
EST	Hs.160410	ESTs	Hs.171113
EST	Hs.160423	ESTs	Hs.171117
RPA3	Hs.1608	EST	Hs.171119
ESTs	Hs.160946	ESTs	Hs.171120
EST	Hs.160956	EST	Hs.171122
ESTs	Hs.160978	EST	Hs.171123
EST	Hs.160980	EST	Hs.171124
EST	Hs.160981	EST	Hs.171140
EST	Hs.160982	EST	Hs.171216
EST	Hs.160983	EST	Hs.171260
Tachykinin Receptor 2	Hs.161305	ESTs	Hs.171264
RAD17 (RAD24)	Hs.16184	RIP	Hs.171545
Human phosphatidylinositol 3-kinase	Hs.162808	ESTs, Weakly similar to immunoglobulin	Hs.171697
catalytic subunit p110delta mRNA,		superfamily member [D. melanogaster]
complete cds.
Human alpha-1 Ig germline C-region	Hs.163271	CD22	Hs.171763
membrane-coding region, 3′ end
GCP-2	Hs.164021		Hs.171776
	Hs.164284	sema domain, immunoglobulin domain (Ig),	Hs.171921
		short basic domain, secreted, (semaphorin)
		3C
EST	Hs.164331	interleukin 11	Hs.1721
	Hs.164427	CD11b	Hs.172631
	Hs.165568	EST, Highly similar to APS [H. sapiens]	Hs.172656
ER	Hs.1657	ALK1	Hs.172670
EST, Highly similar to JM26 [H. sapiens]	Hs.165701		Hs.172674
EST	Hs.165702	CD123	Hs.172689
EST	Hs.165704	ESTs	Hs.172822
EST	Hs.165732	CollaI	Hs.172928
regulatory factor X, 3 (influences HLA class	Hs.166019		Hs.172998
II expression)
LIG4	Hs.166091		Hs.173081
TNFSF18	Hs.248197	myosin, heavy polypeptide 3, skeletal	Hs.173084
		muscle, embryonic
EST	Hs.248228		Hs.173201
H. sapiens rearranged gene for kappa	Hs.248756	Mediterranean fever (MEFV)	Hs.173730
immunoglobulin subgroup V kappa IV
caspase 1, apoptosis-related cysteine	Hs.2490		Hs.173749
protease (interleukin 1, beta, convertase)
EST	Hs.249031	interleukin 1 receptor accessory protein	Hs.173880
TNFRSF10A	Hs.249190	EST, Weakly similar to RL13_HUMAN	Hs.174231
		60S RIBOSOMAL PROTEIN L13
		[H. sapiens]
immunoglobulin lambda variable 3-10	Hs.249208	EST	Hs.174242
Homo sapiens mRNA for single-chain	Hs.249245	EST	Hs.174300
antibody, complete cds
EST	Hs.250473	EST	Hs.174634
ESTs	Hs.250591	EST	Hs.174635
ESTs	Hs.250605	EST	Hs.174650
	Hs.25063	EST	Hs.174673
Human DNA sequence from clone RP1-	Hs.250675	EST	Hs.174716
149A16 on chromosome 22 Contains an
IGLC (Immunoglobulin Lambda Chain C)
pseudogene, the RFPL3 gene for Ret finger
protein-like 3, the RFPL3S gene for Ret
finger protein-like 3 antisense, the gene for
a novel Immunoglobulin Lambda Chain V
family protein, the gene for a novel protein
similar to mouse RGDS (RALGDS,
RALGEF, Guanine Nucleotide Dissociation
Stimulator A) and rabbit oncogene RSC, the
gene for a novel protein (ortholog of worm
F16A11.2 and bacterial and archea-bacterial
predicted proteins), the gene for a novel
protein similar to BPI (Bacterial
Permeability-Increasing Protein) and rabbit
LBP (Liposaccharide-Binding Protein) and
the 5′ part of a novel gene. Contains ESTs,
STSs, GSSs and three putative CpG islands
ACE	Hs.250711	EST	Hs.174740
TREX2	Hs.251398	EST	Hs.174778
Human DNA sequence from clone 1170K4	Hs.251417	EST	Hs.174779
on chromosome 22q12.2–13.1. Contains
three novel genes, one of which codes for a
Trypsin family protein with class A LDL
receptor domains, and the IL2RB gene for
Interleukin 2 Receptor, Beta (IL-2 Receptor,
CD122 antigen). Contains a putative CpG
island, ESTs, and GSSs
EST	Hs.251539	EST, Weakly similar to RL13_HUMAN	Hs.174780
		60S RIBOSOMAL PROTEIN L13
		[H. sapiens]
EST	Hs.251540	(KIAA0033) for ORF, partial cds.	Hs.174905
C3	Hs.251972		Hs.175270
EST	Hs.252273	EST	Hs.175281
EST	Hs.252359	EST	Hs.175300
ESTs, Moderately similar to	Hs.252867	EST	Hs.175336
T2DT_HUMAN TRANSCRIPTION
INITIATION FACTOR TFIID 105 KDA
SUBUNIT [H. sapiens]
EST, Moderately similar to RS2_HUMAN	Hs.253150	EST	Hs.175388
40S RIBOSOMAL PROTEIN S2
[H. sapiens]
EST	Hs.253151		Hs.175437
EST	Hs.253154	EST, Weakly similar to salivary proline-rich	Hs.175777
		protein precursor [H. sapiens]
EST	Hs.253165	EST	Hs.175803
EST	Hs.253166	ESTs	Hs.176337
EST	Hs.253167	EST	Hs.176374
EST	Hs.253168	EST	Hs.176380
EST	Hs.253169	EST	Hs.176404
interleukin 1 receptor, type II	Hs.25333	EST	Hs.176406
	Hs.25361	LCK	Hs.1765
EST	Hs.253742	LIG1	Hs.1770
EST	Hs.253743	EST	Hs.177012
EST, Weakly similar to AF161429_1	Hs.253744	PERB11 family member in MHC class I	Hs.17704
HSPC311 [H. sapiens]		region
EST	Hs.253747	EST	Hs.177146
EST	Hs.253748	EST	Hs.177209
EST	Hs.253753		Hs.177376
EST, Moderately similar to	Hs.254108		Hs.177461
ALU5_HUMAN ALU SUBFAMILY SC
SEQUENCE CONTAMINATION
WARNING ENTRY□ [H. sapiens]
ESTs	Hs.254948	CD99	Hs.177543
ESTs	Hs.255011	PMS2	Hs.177548
EST	Hs.255118	human calmodulin	Hs.177656
EST	Hs.255119		Hs.177712
EST	Hs.255123	Homo sapiens immunoglobulin lambda	Hs.178665
		gene locus DNA, clone:288A10
EST	Hs.255129		Hs.178743
EST	Hs.255134	EST	Hs.179008
EST	Hs.255135	EST	Hs.179070
EST	Hs.255139	EST	Hs.179130
EST	Hs.255140	EST	Hs.179132
ESTs	Hs.255142		Hs.179149
EST	Hs.255150	EST	Hs.179490
EST	Hs.255152	EST	Hs.179492
ESTs	Hs.255153	promyelocytic leukemia cell mRNA, clones	Hs.179735
		pHH58 and pHH81.
ESTs	Hs.255157		Hs.179817
ESTs	Hs.255171	major histocompatibility complex, class II,	Hs.1802
		DO beta
EST	Hs.255172	HLA-DRB1	Hs.180255
EST, Moderately similar to	Hs.255174	TNFRSF12	Hs.180338
PGTA_HUMAN RAB
GERANYLGERANYLTRANSFERASE
ALPHA SUBUNIT [H. sapiens]
EST	Hs.255177	RAD23A (HR23A)	Hs.180455
EST	Hs.255178	MKK3	Hs.180533
EST	Hs.255245	EST	Hs.180637
EST	Hs.255246	CD27	Hs.180841
EST	Hs.255249	STAT6	Hs.181015
EST	Hs.255251	TNFSF4	Hs.181097
EST	Hs.255253	immunoglobulin lambda locus	Hs.181125
EST	Hs.255254		Hs.181368
EST	Hs.255255	CD3	Hs.181392
ESTs	Hs.255256	EST	Hs.255745
EST	Hs.255330	EST	Hs.255746
EST, Weakly similar to putative G protein-	Hs.255333	EST	Hs.255747
coupled Receptor [H. sapiens]
EST	Hs.255336	EST	Hs.255749
EST	Hs.255337	EST	Hs.255754
EST	Hs.255339	ESTs, Moderately similar to KIAA1271	Hs.255759
		protein [H. sapiens]
EST	Hs.255340	EST	Hs.255762
EST	Hs.255341	EST	Hs.255763
ESTs	Hs.255343	EST	Hs.255764
EST	Hs.255347	EST	Hs.255766
EST	Hs.255349	EST	Hs.255767
EST	Hs.255350	EST	Hs.255768
EST	Hs.255354	EST	Hs.255769
ESTs	Hs.255359	EST	Hs.255770
ESTs	Hs.255387	EST	Hs.255772
EST	Hs.255388	EST	Hs.255777
EST	Hs.255389	EST	Hs.255778
ESTs	Hs.255390	EST	Hs.255779
EST	Hs.255392	EST	Hs.255782
EST	Hs.255444	EST	Hs.255783
EST	Hs.255446	EST	Hs.255784
EST	Hs.255448	EST	Hs.255785
ESTs	Hs.255449	EST, Weakly similar to Con1 [H. sapiens]	Hs.255788
EST	Hs.255454	EST	Hs.255791
EST	Hs.255455	EST	Hs.255794
EST	Hs.255457	EST	Hs.255796
EST	Hs.255459	EST	Hs.255797
EST	Hs.255462	EST	Hs.255799
EST	Hs.255464	ESTs	Hs.255877
EST	Hs.255492	EST	Hs.255880
EST	Hs.255494	EST	Hs.255920
EST	Hs.255495	EST	Hs.255927
EST	Hs.255497	CD40	Hs.25648
EST	Hs.255498	interleukin enhancer binding factor 3, 90 kD	Hs.256583
EST	Hs.255499	ESTs	Hs.256810
EST	Hs.255501	EST	Hs.256956
EST	Hs.255502	EST	Hs.256957
EST	Hs.255505	EST	Hs.256959
EST	Hs.255541	EST	Hs.256961
EST	Hs.255543	EST	Hs.256970
ESTs	Hs.255544	EST	Hs.256971
EST	Hs.255546	ESTs	Hs.256979
EST	Hs.255549	ESTs	Hs.257572
EST	Hs.255552	EST	Hs.257579
EST	Hs.255554	EST	Hs.257581
EST	Hs.255556	EST	Hs.257582
EST	Hs.255558	EST	Hs.257630
EST	Hs.255559	EST	Hs.257632
EST	Hs.255560	EST	Hs.257633
EST	Hs.255561	EST	Hs.257636
EST	Hs.255569	EST	Hs.257640
EST	Hs.255572	ESTs	Hs.257641
EST	Hs.255573	EST	Hs.257644
EST	Hs.255575	EST	Hs.257645
EST	Hs.255577	EST	Hs.257646
EST	Hs.255578	EST	Hs.257647
EST	Hs.255579	EST	Hs.257667
EST	Hs.255580	EST	Hs.257668
EST	Hs.255590	EST	Hs.257677
EST	Hs.255591	EST	Hs.257679
EST	Hs.255598	EST	Hs.257680
TNFRSF17	Hs.2556	ESTs	Hs.257682
EST	Hs.255600	ESTs	Hs.257684
EST	Hs.255601	EST	Hs.257687
ESTs, Highly similar to KIAA1039 protein	Hs.255603	EST	Hs.257688
[H. sapiens]
EST	Hs.255614	EST	Hs.257690
EST	Hs.255615	EST	Hs.257695
ESTs	Hs.255617	EST	Hs.257697
EST	Hs.255618	EST	Hs.257705
EST	Hs.255621	EST	Hs.257706
EST	Hs.255622	EST	Hs.257709
ESTs	Hs.255625	ESTs, Moderately similar to	Hs.257711
		ALU8_HUMAN ALU SUBFAMILY SX
		SEQUENCE CONTAMINATION
		WARNING ENTRY□ [H. sapiens]
EST	Hs.255626	EST	Hs.257713
ESTs	Hs.255627	EST	Hs.257716
ESTs	Hs.255630	EST	Hs.257719
EST	Hs.255632	EST	Hs.257720
EST	Hs.255633	EST	Hs.257727
EST	Hs.255634	EST	Hs.257730
EST	Hs.255635	EST	Hs.257738
EST	Hs.255637	EST	Hs.257743
ESTs	Hs.255639	ESTs	Hs.258513
EST	Hs.255641	EST	Hs.258820
EST	Hs.255644	EST	Hs.258864
EST	Hs.255645	sema domain, immunoglobulin domain (Ig),	Hs.25887
		transmembrane domain (TM) and short
		cytoplasmic domain, (semaphorin) 4F
EST	Hs.255646	EST	Hs.258898
EST	Hs.255647	EST	Hs.258933
EST	Hs.255648	interleukin 13 receptor, alpha 2	Hs.25954
EST	Hs.255649	Homo sapiens HSPC101 mRNA, partial cds	Hs.259683
EST	Hs.255650	EST	Hs.263695
EST	Hs.255653	ESTs	Hs.263784
EST	Hs.255657	TNFSF12	Hs.26401
EST	Hs.255661	EST	Hs.264154
ESTs	Hs.255664	EST	Hs.264654
EST	Hs.255665	CDw116b	Hs.265262
EST	Hs.255666	MHC binding factor, beta	Hs.2654
EST	Hs.255668	EST	Hs.265634
EST	Hs.255671	EST	Hs.266387
EST	Hs.255672	ESTs	Hs.268027
EST	Hs.255673	ATHS (LDLR?)	Hs.268571
EST	Hs.255674	ESTs, Highly similar to AAD18086 BAT2	Hs.270193
		[H. sapiens]
EST	Hs.255675	ESTs	Hs.270198
EST	Hs.255677	ESTs	Hs.270294
EST	Hs.255679	ESTs, Weakly similar to alternatively	Hs.270542
		spliced product using exon 13A [H. sapiens]
EST	Hs.255681	ESTs, Moderately similar to	Hs.270561
		ALU2_HUMAN ALU SUBFAMILY SB
		SEQUENCE CONTAMINATION
		WARNING ENTRY□ [H. sapiens]
EST	Hs.255682	ESTs, Weakly similar to pro alpha 1(I)	Hs.270564
		collagen [H. sapiens]
EST	Hs.255686	ESTs, Weakly similar to ALU1_HUMAN	Hs.270578
		ALU SUBFAMILY J SEQUENCE
		CONTAMINATION WARNING
		ENTRY□ [H. sapiens]
ESTs	Hs.255687	ESTs, Moderately similar to brain-derived	Hs.270588
		immunoglobulin superfamily molecule
		[M. musculus]
EST	Hs.255688	TALL1	Hs.270737
ESTs	Hs.255689	ESTs	Hs.271206
EST	Hs.255691	MYH	Hs.271353
EST	Hs.255692	POLI (RAD30B)	Hs.271699
ESTs	Hs.255693	ADPRTL3	Hs.271742
EST	Hs.255695	ESTs, Moderately similar to	Hs.272075
		ALU8_HUMAN ALU SUBFAMILY SX
		SEQUENCE CONTAMINATION
		WARNING ENTRY□ [H. sapiens]
EST, Highly similar to transmembrane	Hs.255697	Human DNA sequence from clone RP5-	Hs.272271
chloride conductor protein [H. sapiens]		1170K4 on chromosome 22q12.2–13.1
		Contains three novel genes, one of which
		codes for a Trypsin family protein with
		class A LDL receptor domains, and the
		IL2RB gene for Interleukin 2 Receptor,
		Beta (IL-2 Receptor, CD122 antigen), a
		putative CpG island ESTs and GSSs
EST	Hs.255698	interleukin 1 receptor accessory protein-like	Hs.272354
		2
EST	Hs.255699	Homo sapiens partial IGVH3 V3-20 gene	Hs.272355
		for immunoglobulin heavy chain V region,
		case 1, clone 2
EST	Hs.255705	Homo sapiens partial IGVH3 gene for	Hs.272356
		immunoglobulin heavy chain V region, case
		1, clone 16
EST	Hs.255706	Homo sapiens partial IGVH3 gene for	Hs.272357
		immunoglobulin heavy chain V region, case
		1, clone 19
EST	Hs.255708	Homo sapiens partial IGVH3 gene for	Hs.272358
		immunoglobulin heavy chain V region, case
		1, cell Mo IV 72
EST	Hs.255710	Homo sapiens partial IGVH1 gene for	Hs.272359
		immunoglobulin heavy chain V region, case
		1, cell Mo V 94
EST	Hs.255713	Homo sapiens partial IGVL2 gene for	Hs.272360
		immunoglobulin lambda light chain V
		region, case 1, cell Mo V 94
EST	Hs.255717	Homo sapiens partial IGVH3 gene for	Hs.272361
		immunoglobulin heavy chain V region, case
		1, cell Mo VI 7
EST	Hs.255718	Homo sapiens partial IGVL1 gene for	Hs.272362
		immunoglobulin lambda light chain V
		region, case 1, cell Mo VI 65
EST	Hs.255721	Homo sapiens partial IGVH3 gene for	Hs.272363
		immunoglobulin heavy chain V region, case
		1, cell Mo VI 162
ESTs	Hs.255723	Homo sapiens partial IGVH3 DP29 gene	Hs.272364
		for immunoglobulin heavy chain V region,
		case 1, cell Mo VII 116
EST	Hs.255725	Homo sapiens partial IGVH4 gene for	Hs.272365
		immunoglobulin heavy chain V region, case
		2, cell D 56
EST	Hs.255726	Homo sapiens partial IGVH3 gene for	Hs.272366
		immunoglobulin heavy chain V region, case
		2, cell E 172
EST	Hs.255727	interleukin 20	Hs.272373
EST	Hs.255736	Human DNA sequence from clone RP1-	Hs.272521
		149A16 on chromosome 22 Contains an
		IGLC (Immunoglobulin Lambda Chain C)
		pseudogene, the RFPL3 gene for Ret finger
		protein-like 3, the RFPL3S gene for Ret
		finger protein-like 3 antisense, the gene for
		a novel Immunoglobulin Lambda Chain V
		family protein, the gene for a novel protein
		similar to mouse RGDS (RALGDS,
		RALGEF, Guanine Nucleotide Dissociation
		Stimulator A) and rabbit oncogene RSC, the
		gene for a novel protein (ortholog of worm
		F16A11.2 and bacterial and archea-bacterial
		predicted proteins), the gene for a novel
		protein similar to BPI (Bacterial
		Permeability-Increasing Protein) and rabbit
		LBP (Liposaccharide-Binding Protein) and
		the 5′ part of a novel gene. Contains ESTs,
		STSs, GSSs and three putative CpG islands
EST	Hs.255740	TdT	Hs.272537
EST	Hs.255742	ret finger protein-like 3 antisense	Hs.274285
EST	Hs.255743	PRKR	Hs.274382
EST	Hs.7569	H. sapiens immunoglobulin epsilon chain	Hs.274600
SMAD4	Hs.75862	EST, Weakly similar to HLA-DQ alpha	Hs.275720
		chain [H. sapiens]
Homo sapiens splicing factor,	Hs.76122	EST, Weakly similar to RL13_HUMAN	Hs.276279
arginine/serine-rich 4 (SFRS4) mRNA.		60S RIBOSOMAL PROTEIN L13
		[H. sapiens]
thymosin beta-10	Hs.76293	EST	Hs.276341
CD63	Hs.76294	EST	Hs.276342
AIF1	Hs.76364	EST, Weakly similar to RL13_HUMAN	Hs.276353
		60S RIBOSOMAL PROTEIN L13
		[H. sapiens]
phospholipase A2, group IIA (platelets,	Hs.76422	EST	Hs.276774
synovial fluid),
CES1	Hs.76688	EST	Hs.276819
ubiquitin conjugating enzyme	Hs.76932	EST	Hs.276871
Homo sapiens KIAA0963 protein	Hs.7724	EST, Weakly similar to FBRL_HUMAN	Hs.276872
(KIAA0963), mRNA.		FIBRILLARIN [H. sapiens]
Homo sapiens fragile histidine triad gene	Hs.77252	EST	Hs.276887
(FHIT) mRNA.
PAF-AH	Hs.77318	EST	Hs.276902
Mig	Hs.77367	EST	Hs.276917
DDB2	Hs.77602	EST	Hs.276918
ATR	Hs.77613	EST, Weakly similar to RL13_HUMAN	Hs.276938
		60S RIBOSOMAL PROTEIN L13
		[H. sapiens]
XPB (ERCC3)	Hs.77929	EST	Hs.277051
PNKP	Hs.78016	EST	Hs.277052
C7	Hs.78065	EST, Moderately similar to RL13_HUMAN	Hs.277236
		60S RIBOSOMAL PROTEIN L13
		[H. sapiens]
Homo sapiens small nuclear RNA activating	Hs.78403	EST, Moderately similar to DEAD Box	Hs.277237
complex, polypeptide 2, 45 kD (SNAPC2)		Protein 5 [H. sapiens]
mRNA.
	Hs.78465	EST	Hs.277238
sphingolipid activator protein/cerebroside	Hs.78575	EST	Hs.277286
sulfate activator protein
Homo sapiens aminolevulinate, delta-,	Hs.78712	major histocompatibility complex, class I, C	Hs.277477
synthase 1 (ALAS1), nuclear gene encoding
mitochondrial protein, mRNA.
tyrosine kinase with immunoglobulin and	Hs.78824	EST, Weakly similar to AF150959 1	Hs.277591
epidermal growth factor homology domains		immunoglobulin G1 Fc fragment
		[H. sapiens]
Hsp72	Hs.78846	EST	Hs.277714
UNG	Hs.78853	EST	Hs.277715
CX3CR1	Hs.78913	EST	Hs.277716
MSH2	Hs.78934	EST	Hs.277717
CRHR1	Hs.79117	EST	Hs.277718
BCL2	Hs.79241	EST, Weakly similar to BAT3_HUMAN	Hs.277774
		LARGE PROLINE-RICH PROTEIN BAT3
		[H. sapiens]
P-selectin	Hs.79283	EST	Hs.277975
UBE2VE (MMS2)	Hs.79300	EST	Hs.278060
retinoid X receptor, beta	Hs.79372	cytochrome P450, subfamily XXIA (steroid	Hs.278430
		21-hydroxylase, congenital adrenal
		hyperplasia), polypeptide 2
MPG	Hs.79396	KIAA0015 gene product	Hs.278441
RPA2	Hs.79411	CD32B	Hs.278443
heat shock 70 kD protein-like 1	Hs.80288	KIR2DL1	Hs.278453
FANCG (XRCC9)	Hs.8047	CD158a	Hs.278455
CD43	Hs.80738	CD24	Hs.278667
POLG	Hs.80961	HLA class II region expressed gene KE4	Hs.278721
Human CB-4 transcript of unrearranged	Hs.81220	IL-17C	Hs.278911
immunoglobulin V(H)5 gene
Human L2-9 transcript of unrearranged	Hs.81221	HSPC048 protein (HSPC048)	Hs.278944
immunoglobulin V(H)5 pseudogene
immunoglobulin superfamily, member 3	Hs.81234	HSPC054 protein (HSPC054)	Hs.278946
UBL1	Hs.81424	HSPC073 protein (HSPC073)	Hs.278948
PF4	Hs.81564	ESTs	Hs.279066
palmitoyl-protein thioesterase 2	Hs.81737	ESTs	Hs.279067
natural killer cell receptor, immunoglobulin	Hs.81743	ESTs	Hs.279068
superfamily member
TNFRSF11B	Hs.81791	ESTs	Hs.279069
interleukin 6 signal transducer (gp130,	Hs.82065	ESTs	Hs.279070
oncostatin M receptor)
CD138	Hs.82109	ESTs	Hs.279071
Human monocytic leukaemia zinc finger	Hs.82210	ESTs	Hs.279072
protein (MOZ) mRNA, complete cds.
sema domain, immunoglobulin domain (Ig),	Hs.82222	ESTs, Weakly similar to KIAA0052 protein	Hs.279073
short basic domain, secreted, (semaphorin)		[H. sapiens]
3B
HPRT	Hs.82314	ESTs	Hs.279074
Human RNA binding protein Etr-3 mRNA,	Hs.82321	ESTs	Hs.279075
complete cds.
MNAT1	Hs.82380	ESTs	Hs.279076
SMAD2	Hs.82483	ESTs	Hs.279077
CD47	Hs.82685	EST	Hs.279078
CETN2	Hs.82794	EST	Hs.279079
protein phosphatase 1, regulatory (inhibitor)	Hs.82887	ESTs	Hs.279080
subunit 11
MMP1	Hs.83169	EST	Hs.279081
D3-type cyclin (CCND3)	Hs.83173	ESTs	Hs.279082
MMP3	Hs.83326	ESTs	Hs.279083
TNFSF10	Hs.83429	ESTs	Hs.279084
CD33	Hs.83731	ESTs	Hs.279085
CD102	Hs.83733	ESTs	Hs.279086
	Hs.84153	ESTs, Weakly similar to AF201422_1	Hs.279087
		splicing coactivator subunit SRm300
		[H. sapiens]
interleukin 8 receptor, beta	Hs.846	ESTs	Hs.279088
titin immunoglobulin domain protein	Hs.84665	ESTs	Hs.279089
(myotilin)
KU80 (XRCC5)	Hs.84981		Hs.86437
Raf-1	Hs.85181		Hs.86761
major histocompatibility complex, class I, J	Hs.85242	CD118 = IFNAR-2	Hs.86958
(pseudogene)
RELB	Hs.858		Hs.87113
	Hs.85923	PGHS-1	Hs.88474
ERK1	Hs.861		Hs.8882
FADD	Hs.86131	LT-b	Hs.890
MHC class I polypeptide-related sequence	Hs.90598	EST	Hs.92440
A
TNF receptor-associated factor 6	Hs.90957		Hs.92460
Topo3A	Hs.91175	myosin-binding protein H	Hs.927
PARG	Hs.91390	IFN-b	Hs.93177
HLA-DPA1	Hs.914	C8A	Hs.93210
SEEK1	Hs.91600	pre-B-cell leukemia transcription factor 2	Hs.93728
POLD1	Hs.99890	Tachykinin Receptor 3	Hs.942
ALK4	Hs.99954	Homo sapiens cDNA FLJ12242 fis, clone	Hs.94810
		MAMMA1001292
XPD (ERCC2)	Hs.99987	CD29	Hs287797
SCYA25 (CCL25)	Hs.50404	LIF	Hs.2250
SCYA19 (CCL19)	Hs.50002	Human IP-10	Hs.2248
TCIRG1	Hs.46465	IL-5	Hs.2247
PAF-Receptor	Hs.46	G-CSF	Hs.2233
CD26	Hs.44926	TGF-bR	Hs.220
	Hs.44865	G-CSFR	Hs.2175
REL	Hs.44313	CD15	Hs.2173
IL-17	Hs.41724	STAT1	Hs.21486
CD49d	Hs.40034	CD85	Hs.204040
CCR2	Hs.395	HCC-1	Hs.20144
	Hs.3688	Fas ligand	Hs.2007
TNF-b	Hs.36	CD28	Hs.1987
lactoferrin	Hs.347	HLA-DQA1	Hs.198253
MCP-1	Hs.340	Ku70 (G22P1)	Hs.197345
CD150	Hs.32970	PGHS-2	Hs.196384
IL-10Ra	Hs.327	CDw128	Hs.194778
EGR1	Hs.326035	IL-10	Hs.193717
SCYC1 (XCL1)	Hs.3195	CD126	Hs.193400
HLA-DR	Hs.318720		Hs.1880
Topo I (TOP1)	Hs.317	CD98	Hs.184601
SCYA2 (MCP1)	Hs.303649		Hs.184542
HuRNPD	Hs.303627	MHC class I region ORF	Hs.1845
Human C mu gene for IgM heavy chain	Hs.302063	CDw116a	Hs.182378
exons CH1–4, secretory
P1	Hs.297681	HLA-DRB5	Hs.181366
immunoglobulin lambda joining 3	Hs.289110	major histocompatibility complex, class I, A	Hs.181244
major histocompatibility complex, class II,	Hs.289095	elongation factor 1-alpha (clone CEF4)	Hs.181165
DQ alpha 2
HSPCA	Hs.289088	CD119	Hs.180866
interleukin 22	Hs.287369		Hs.180804
ribosomal protein L4	Hs.286		Hs.180532
IgM	Hs.285823	POLB	Hs.180107
EST	Hs.283267	CD1d	Hs.1799
TREM1	Hs.283022	CD87	Hs.179657
HLA-DRB3	Hs.279930	minichromosome maintenance deficient	Hs.179565
		(S. cerevisiae) 3
LIFR	Hs.2798	RAD23B (HR23B)	Hs.178658
C4B	Hs.278625		Hs.178391
EST	Hs.276907		Hs.177781
CDw52	Hs.276770	ADPRT	Hs.177766
CD16 b	Hs.274467	IFNGR2	Hs.177559
heat shock 70 kD protein 1B	Hs.274402	CD16 a	Hs.176663
Th1	Hs.273385	CD4	Hs.17483
MIP-5/HCC-2	Hs.272493	SCYC2 (XCL2)	Hs.174228
TBX21	Hs.272409	CD115	Hs.174142
Homo sapiens mRNA; cDNA	Hs.272307	CD11a	Hs.174103
DKFZp434O2417 (from clone
DKFZp434O2417); partial cds
Human DNA sequence from clone RP1-	Hs.272295	IL-10Rb	Hs.173936
108C2 on chromosome 6p12.1–21.1.
Contains the MCM3 gene for
minichromosome maintenance deficient (S.
cerevisiae) 3 (DNA replication licensing
factor, DNA polymerase alpha holoenzyme-
associated protein P1, RLF beta subunit), a
CACT (carnitine/acylcarnitine translocase)
pseudogene, part of the gene for a
PUTATIVE novel protein similar to IL17
(interleukin 17 (cytotoxic T-lymphocyte-
associated serine esterase 8)) (cytotoxic T
lymphocyte-associated antigen 8, CTLA8),
ESTs, STSs, GSSs and a putative CpG
island
CD49b	Hs.271986	MSCF	Hs.173894
MCP-2	Hs.271387	TDG	Hs.173824
CD49c	Hs.265829	RAC1	Hs.173737
NBS1	Hs.25812	integrin cytoplasmic domain-associated	Hs.173274
		protein 1
CD120b = TNFRSF1B	Hs.256278	IL2R	Hs.1724
CDw75	Hs.2554	IL-1a	Hs.1722
CD82	Hs.25409		Hs.171872
MCP-3	Hs.251526		Hs.171118
xanthine oxidase	Hs.250	EST	Hs.171009
Human Ig rearranged lambda-chain mRNA,	Hs.247947	EST	Hs.170934
subgroup VL3, V-J region, partial cds
Eotaxin-2/MPIF-2	Hs.247838	EST	Hs.170587
CTLA-4	Hs.247824	IL-9R	Hs.1702
immunoglobulin kappa variable 1-9	Hs.247792	CD45	Hs.170121
CD68	Hs.246381	TGF-a	Hs.170009
OSMR	Hs.238648	CD44	Hs.169610
CDw127	Hs.237868	Fyn	Hs.169370
transcription factor 8 (represses interleukin	Hs.232068	MPIF-1	Hs.169191
2 expression)
CD8b	Hs.2299	ICAM-1	Hs.168383
EST	Hs.229374	IL-15	Hs.168132
TRF4-1	Hs.225951	STAT5A	Hs.167503
CD3g	Hs.2259	ESTs	Hs.167208
C2	Hs.2253	ESTs	Hs.165693
	Hs.116834		Hs.135750
	Hs.117741	DINB1 (POLK)	Hs.135756
Human MHC Class I region proline rich	Hs.118354	Human DNA sequence from clone RP1-	Hs.136141
protein mRNA, complete cds		238O23 on chromosome 6. Contains part of
		the gene for a novel protein similar to PIGR
		(polymeric immunoglobulin receptor), part
		of the gene for a novel protein similar to rat
		SAC (soluble adenylyl cyclase), ESTs,
		STSs and GSS
ESTs, Weakly similar to FCE2 MOUSE	Hs.118392		Hs.136254
LOW AFFINITY IMMUNOGLOBULIN
EPSILON FC RECEPTOR [M. musculus]
MKK6	Hs.118825		Hs.13646
	Hs.118895		Hs.136537
H. sapiens mRNA for ITBA4 gene.	Hs.119018	Histone H1 (F3)	Hs.136857
	Hs.119057	MGMT	Hs.1384
TNFRSF10c	Hs.119684		Hs.138563
	Hs.12064	IgG	Hs.140
	Hs.120907		Hs.140478
acid phosphatase 5, tartrate resistant	Hs.1211		Hs.14070
	Hs.121297		Hs.141153
Human immunoglobulin (mAb59) light	Hs.121508		Hs.143954
chain V region mRNA, partial sequence
IL12Rb1	Hs.121544	ESTs, Moderately similar to	Hs.144814
		I1BC_HUMAN INTERLEUKIN-1 BETA
		CONVERTASE PRECURSOR [H. sapiens]
Human MHC class II DO-alpha mRNA,	Hs.123041	CHK2 (Rad53)	Hs.146329
partial cds
Histone H4 (H4F2)	Hs.123053	EST	Hs.146591
TSHR	Hs.123078		Hs.147040
	Hs.123445	CD42b	Hs.1472
regulatory factor X, 1 (influences HLA class	Hs.123638		Hs.149235
II expression)
CD13	Hs.1239	AICD	Hs.149342
IL-15R	Hs.12503	Homo sapiens putative tumor suppressor	Hs.149443
		protein (101F6) mRNA, complete cds.
RAD51L3 (RAD51D)	Hs.125244	CD49e	Hs.149609
CDw90	Hs.125359	heparan sulfate proteoglycan (HSPG) core	Hs.1501
		protein
LYPLA1	Hs.12540	CD107a	Hs.150101
ESTs, Weakly similar to AF201951 1 high	Hs.126580	ESTs, Weakly similar to I57587 MHC HLA	Hs.150175
affinity immunoglobulin epsilon receptor		SX-alpha [H. sapiens]
beta subunit [H. sapiens]
	Hs.127128	ALK2	Hs.150402
	Hs.127444	WRN	Hs.150477
C5	Hs.1281	EST	Hs.150708
C8G	Hs.1285	XRCC4	Hs.150930
RAD54B	Hs.128501	IFN-a	Hs.1510
	Hs.129020	MAPK	Hs.151051
	Hs.129268		Hs.15200
	Hs.129332	immunoglobulin mu binding protein 2	Hs.1521
XRCC2	Hs.129727	4-1BBL	Hs.1524
potassium voltage-gated channel, Shaw-	Hs.129738		Hs.152818
related subfamily, member 3 (KCNC3)
interleukin 17 receptor	Hs.129751	HUS1	Hs.152983
CD134	Hs.129780	SWAP70	Hs.153026
TNFRSF10d	Hs.129844	DOM-3 (C. elegans) homolog Z	Hs.153299
POLL	Hs.129903		Hs.153551
GADD153 = growth arrest and DNA-damage	Hs.129913		Hs.15370
inducible gene/fus-chop fusion protein
solute carrier family 5 (neutral amino acid	Hs.130101	SMAD6	Hs.153863
transporters, system A), member 4
	Hs.130232	APEXL2	Hs.154149
	Hs.13034		Hs.154198
CD30L	Hs.1313		Hs.154366
SCYA26 (CCL26)	Hs.131342	BCL6	Hs.155024
CD30	Hs.1314		Hs.155150
	Hs.131885		Hs.155402
	Hs.131887	RAIDD	Hs.155566
	Hs.13256	POLH	Hs.155573
ESTs	Hs.132775		Hs.15589
Homo sapiens (clone 3.8-1) MHC class I	Hs.132807	Homo sapiens mRNA for KIAA0695	Hs.155976
mRNA fragment		protein, complete cds.
	Hs.13288	SNM1 (PS02)	Hs.1560
	Hs.132943	Topo2A	Hs.156346
EST	Hs.133261	ESTs, Highly similar to MHC class II	Hs.156811
		antigen [H. sapiens]
	Hs.133388	Histamine H1 receptor	Hs.1570
EST	Hs.133393		Hs.157118
EST	Hs.133930		Hs.157267
ESTs	Hs.133947	EST	Hs.157279
ESTs	Hs.133949	EST	Hs.157280
EST	Hs.134017	EST	Hs.157308
EST	Hs.134018	EST	Hs.157309
EST	Hs.134590	EST	Hs.157310
	Hs.135135	EST	Hs.157311
immunoglobulin superfamily, member 6	Hs.135194	ESTs	Hs.157344
	Hs.135570	ret finger protein-like 2	Hs.157427
Homo sapiens arrestin, beta 2 (ARRB2)	Hs.18142		Hs.214956
mRNA.
myeloperoxidase	Hs.1817	WASP	Hs.2157
APO-1	Hs.182359	CD88	Hs.2161
TRAP1	Hs.182366		Hs.21618
	Hs.182594	ring finger protein 5	Hs.216354
TNFRSF16	Hs.1827	class II cytokine receptor ZCYTOR7	Hs.21814
	Hs.182817		Hs.219149
regulatory factor X, 4 (influences HLA class	Hs.183009	cyclophilin-related protein	Hs.219153
II expression)
Homo sapiens killer cell lectin-like receptor	Hs.183125	Homo sapiens mannosyl (alpha-1,6-)-	Hs.219479
F1 (KLRF1), mRNA.		glycoprotein beta-1,2-N-
		acetylglucosaminyltransferase (MGAT2)
		mRNA.
	Hs.183171	perforin	Hs.2200
EST	Hs.183386		Hs.220154
	Hs.183656	ESTs, Weakly similar to FCE2 MOUSE	Hs.220649
		LOW AFFINITY IMMUNOGLOBULIN
		EPSILON FC RECEPTOR [M. musculus]
	Hs.18368		Hs.220868
advanced glycosylation end product-specific	Hs.184		Hs.220960
receptor
CDK7	Hs.184298	immunoglobulin superfamily, member 1	Hs.22111
	Hs.184376		Hs.221539
CCR4	Hs.184926	ESTs	Hs.221694
EST, Weakly similar to A27307 proline-	Hs.185463		Hs.222921
rich phosphoprotein [H. sapiens]
EST	Hs.185498		Hs.222942
EST, Weakly similar to B39066 proline-	Hs.186243	EST	Hs.223520
rich protein 15 - rat [R. norvegicus]
EST, Weakly similar to salivary proline-rich	Hs.186265	EST	Hs.223935
protein [R. norvegicus]
EST	Hs.187200	EST, Moderately similar to SMO_HUMAN	Hs.224178
		SMOOTHENED HOMOLOG
		PRECURSOR [H. sapiens]
	Hs.188048	Blk	Hs.2243
EST	Hs.188075	EST	Hs.224344
EST	Hs.188194	EST	Hs.224408
EST	Hs.188300	EST	Hs.224409
	Hs.190251	CPN1	Hs.2246
	Hs.19056	MMP7	Hs.2256
EST	Hs.190831	MMP10	Hs.2258
MAPK8	Hs.190913	CCR9	Hs.225946
EST	Hs.190921	toll-like receptor 6 (TLR6)	Hs.227105
EST, Weakly similar to S39206	Hs.190924	XPR1	Hs.227656
hypothetical protein 1 - rat□ [R. norvegicus]
GTF2H2	Hs.191356	CD49f	Hs.227730
	Hs.191367		Hs.22790
	Hs.191914	EST	Hs.228337
ESTs, Weakly similar to immunoglobulin	Hs.192078	EST, Highly similar to 1409218A elastase	Hs.228525
superfamily member [D. melanogaster]		[H. sapiens]
XPA	Hs.192803	EST	Hs.228528
CD89	Hs.193122	EST, Moderately similar to	Hs.228874
		R37A_HUMAN 60S RIBOSOMAL
		PROTEIN L37A□ [H. sapiens]
DFFRY	Hs.193145	EST	Hs.228891
CD35	Hs.193716	EST	Hs.228926
REV7 (MAD2L2)	Hs.19400	EST	Hs.229071
	Hs.194082	EST	Hs.229405
	Hs.194110	EST	Hs.229494
BRCA1	Hs.194143	EST, Weakly similar to ALU1_HUMAN	Hs.229560
		ALU SUBFAMILY J SEQUENCE
		CONTAMINATION WARNING
		ENTRY□ [H. sapiens]
ESTs, Moderately similar to MHC Class I	Hs.194249	EST, Moderately similar to AAD18086	Hs.229901
region proline rich protein [H. sapiens]		BAT2 [H. sapiens]
	Hs.194534	EST	Hs.229902
Topo3B	Hs.194685	EST, Highly similar to 1409218A elastase	Hs.230053
		[H. sapiens]
Human DNA sequence from clone 1170K4	Hs.194750	RAD51	Hs.23044
on chromosome 22q12.2–13.1. Contains
three novel genes, one of which codes for a
Trypsin family protein with class A LDL
receptor domains, and the IL2RB gene for
Interleukin 2 Receptor, Beta (IL-2 Receptor,
CD122 antigen). Contains a putative CpG
island, ESTs, and GSSs
major histocompatibility complex, class II,	Hs.194764	EST, Moderately similar to A54746 adhalin	Hs.230485
DP alpha 2 (pseudogene)		precursor - human□ [H. sapiens]
Human DNA sequence from clone RP11-	Hs.194976	EST	Hs.230691
367J7 on chromosome 1. Contains (part of)
two or more genes for novel
Immunoglobulin domains containing
proteins, a SON DNA binding protein
(SON) pseudogene, a voltage-dependent
anion channel 1 (VDAC1) (plasmalemmal
porin) pseudogene, ESTs, STSs and GSSs
	Hs.195447	EST	Hs.230775
PDGF-B	Hs.1976	EST	Hs.230805
CXCR3	Hs.198252	EST	Hs.230848
	Hs.198694	EST	Hs.230862
	Hs.198738	EST	Hs.230874
MAR/SAR DNA binding protein (SATB1)	Hs.198822	EST	Hs.230931
CHUK	Hs.198998	EST	Hs.231031
hemochromatosis	Hs.20019	EST	Hs.231261
T-cell receptor active beta-chain	Hs.2003	EST	Hs.231284
APO-1	Hs.2007,	EST	Hs.231285
RXRA	Hs.20084	EST	Hs.231292
EST	Hs.200876	EST, Weakly similar to putative	Hs.231512
		mitochondrial outer membrane protein
		import receptor [H. sapiens]
	Hs.201194	Homo sapiens mRNA for KIAA0529	Hs.23168
		protein, partial cds.
TCRd	Hs.2014	EST	Hs.235042
ESTs, Highly similar to TNF-alpha	Hs.202407	EST	Hs.235826
converting enzyme [H. sapiens]
	Hs.202608	TREX1 (Dnase III)	Hs.23595
Integrin b1 = CD29	Hs.202661	EST	Hs.237126
thrombomodulin	Hs.2030		Hs.23860
	Hs.203064	RAD9	Hs.240457
	Hs.203184	1-acylglycerol-3-phosphate O-	Hs.240534
		acyltransferase 1 (lysophosphatidic acid
		acyltransferase, alpha)
	Hs.203584	EST	Hs.240635
EST	Hs.204477	EST, Weakly similar to ALU8_HUMAN	Hs.241136
		ALU SUBFAMILY SX SEQUENCE
		CONTAMINATION WARNING
		ENTRY□ [H. sapiens]
EST	Hs.204480	TNFSF15	Hs.241382
EST, Weakly similar to CA13_HUMAN	Hs.204483	interleukin 1 receptor accessory protein-like	Hs.241385
COLLAGEN ALPHA 1(III) CHAIN		1
PRECURSOR□ [H. sapiens]
ESTs	Hs.204588	RANTES	Hs.241392
EST, Weakly similar to salivary proline-rich	Hs.204598	sema domain, immunoglobulin domain (Ig),	Hs.2414
protein 1 [H. sapiens]		short basic domain, secreted, (semaphorin)
		3A
EST	Hs.204610	POLQ	Hs.241517
ESTs	Hs.204703	TNF-a	Hs.241570
ESTs	Hs.204751	Homo sapiens genes encoding RNCC	Hs.241586
		protein, DDAH protein, Ly6-C protein, Ly6-
		D protein and immunoglobulin receptor
EST	Hs.204760	megakaryocyte-enhanced gene transcript 1	Hs.241587
		protein
EST	Hs.204771	EST, Moderately similar to 1409218A	Hs.241981
		elastase [H. sapiens]
ESTs	Hs.204873	EST	Hs.241982
ESTs	Hs.204932	EST	Hs.241983
EST	Hs.204954	EST	Hs.242605
EST	Hs.205158	ADPRT2	Hs.24284
ESTs	Hs.205159	EST	Hs.243284
ESTs	Hs.205327	EST	Hs.243286
CD39	Hs.205353	ESTs	Hs.243288
ESTs	Hs.205435	SCYB14	Hs.24395
EST	Hs.205438	EST	Hs.244046
EST, Highly similar to elastic titin	Hs.205452	EST	Hs.244048
[H. sapiens]
EST	Hs.205456	EST	Hs.244049
MRE11A	Hs.20555	EST	Hs.244050
HLA class II region expressed gene KE2	Hs.205736	RFXAP	Hs.24422
EST	Hs.205788		Hs.24435
ESTs	Hs.205789	STAT5B	Hs.244613
EST	Hs.205803	EST	Hs.244666
EST	Hs.205815	EST	Hs.245586
ESTs	Hs.206160	CDw108	Hs.24640
	Hs.206654	ESTs	Hs.246796
EST	Hs.207060	dimethylarginine dimethylaminohydrolase 2	Hs.247362
EST	Hs.207062	Homo sapiens clone mcg53–54	Hs.247721
		immunoglobulin lambda light chain variable
		region 4a mRNA, partial cds
EST	Hs.207063	Homo sapiens ELK1 pseudogene (ELK2)	Hs.247775
		and immunoglobulin heavy chain gamma
		pseudogene (IGHGP)
EST	Hs.207473	immunoglobulin kappa variable 1/OR2-108	Hs.247804
ESTs	Hs.207474	butyrophilin-like 2 (MHC class II	Hs.247808
		associated)
ESTs	Hs.207971	Homo sapiens genes encoding RNCC	Hs.247879
		protein, DDAH protein, Ly6-C protein, Ly6-
		D protein and immunoglobulin receptor
EST	Hs.207993	Histamine H2 receptor	Hs.247885
EST	Hs.208153	Human anti-streptococcal/anti-myosin	Hs.247898
		immunoglobulin lambda light chain variable
		region mRNA, partial cds
EST, Weakly similar to S10889 proline-rich	Hs.208667	Homo sapiens isolate donor Z clone Z55K	Hs.247907
protein - human□ [H. sapiens]		immunoglobulin kappa light chain variable
		region mRNA, partial cds
ESTs	Hs.209142	Homo sapiens isolate donor D clone D103L	Hs.247908
		immunoglobulin lambda light chain variable
		region mRNA, partial cds
EST	Hs.209261	Homo sapiens isolate 459 immunoglobulin	Hs.247909
		lambda light chain variable region (IGL)
		gene, partial cds
ESTs	Hs.209306	Homo sapiens isolate donor N clone N88K	Hs.247910
		immunoglobulin kappa light chain variable
		region mRNA, partial cds
	Hs.209362	Homo sapiens isolate donor N clone N8K	Hs.247911
		immunoglobulin kappa light chain variable
		region mRNA, partial cds
EST, Weakly similar to FCEB MOUSE	Hs.209540	Human Ig rearranged mu-chain V-region	Hs.247923
HIGH AFFINITY IMMUNOGLOBULIN		gene, subgroup VH-III, exon 1 and 2
EPSILON RECEPTOR BETA-SUBUNIT
[M. musculus]
EST	Hs.209913	Epsilon, IgE = membrane-bound IgE,	Hs.247930
		epsilon m/s isoform {alternative splicing}
		[human, mRNA Partial, 216 nt]
EST	Hs.209989	H. sapiens (T1.1) mRNA for IG lambda light	Hs.247949
		chain
EST	Hs.210049	H. sapiens mRNA for Ig light chain, variable	Hs.247950
		region (ID:CLL001VL)
EST, Moderately similar to probable	Hs.210276	Human interleukin 2 gene, clone pATtacIL-	Hs.247956
sodium potassium ATPase gamma chain		2C/2TT, complete cds, clone pATtacIL-
[H. sapiens]		2C/2TT
EST, Weakly similar to N-WASP	Hs.210306	pre-B lymphocyte gene 1	Hs.247979
[H. sapiens]
EST	Hs.210307	Human immunoglobulin heavy chain	Hs.247987
		variable region (V4-31) gene, partial cds
EST	Hs.210385	Human immunoglobulin heavy chain	Hs.247989
		variable region (V4-30.2) gene, partial cds
interleukin 21 receptor	Hs.210546	Human DNA sequence from phage LAW2	Hs.247991
		from a contig from the tip of the short arm
		of chromosome 16, spanning 2 Mb of
		16p13.3 Contains Interleukin 9 receptor
		pseudogene
EST	Hs.210727	Homo sapiens HLA class III region	Hs.247993
		containing NOTCH4 gene, partial sequence,
		homeobox PBX2 (HPBX) gene, receptor
		for advanced glycosylation end products
		(RAGE) gene, complete cds, and 6
		unidentified cds
	Hs.211266	Homo sapiens immunoglobulin lambda	Hs.248010
		gene locus DNA, clone:61D6
SMAD3	Hs.211578	immunoglobulin lambda variable 9-49	Hs.248011
MHC class I polypeptide-related sequence	Hs.211580	immunoglobulin lambda variable 4-3	Hs.248012
B
ESTs, Weakly similar to CA1B_MOUSE	Hs.211744	H. sapiens mRNA for IgG lambda light	Hs.248030
COLLAGEN ALPHA 1(XI) CHAIN		chain V-J-C region (clone Tgl11)
PRECURSOR□ [M. musculus]
sema domain, immunoglobulin domain (Ig),	Hs.212414	Human immunoglobulin (mAb56) light	Hs.248043
short basic domain, secreted, (semaphorin)		chain V region mRNA, partial sequence
3E
TNFRSF18	Hs.212680	Homo sapiens lymphocyte-predominant	Hs.248077
		Hodgkin's disease case #4 immunoglobulin
		heavy chain gene, variable region, partial
		cds
Homo sapiens general transcription factor 2-	Hs.212939	Homo sapiens lymphocyte-predominant	Hs.248078
I pseudogene 1 (GTF2IP1) mRNA.		Hodgkin's disease case #7 immunoglobulin
		heavy chain gene, variable region, partial
		cds
RAD18	Hs.21320	Homo sapiens clone ASMneg1-b3	Hs.248083
		immunoglobulin lambda chain VJ region,
		(IGL) mRNA, partial cds
	Hs.213226	OSM	Hs.248156
ESTs	Hs.279090		Hs.29128
ESTs	Hs.279091	Homo sapiens clone 24659 mRNA	Hs.29206
		sequence.
ESTs	Hs.279092	EST	Hs.292235
EST	Hs.279093	EST	Hs.292450
ESTs	Hs.279094	EST, Moderately similar to Ewing sarcoma	Hs.292455
		breakpoint region 1, isoform EWS
		[H. sapiens]
ESTs	Hs.279095	EST	Hs.292461
ESTs, Weakly similar to AF279265_1	Hs.279096	ESTs	Hs.292501
putative anion transporter 1 [H. sapiens]
ESTs	Hs.279097	EST	Hs.292516
EST	Hs.279098	EST	Hs.292517
ESTs	Hs.279099	EST	Hs.292520
ESTs	Hs.279100	EST, Moderately similar to RL13_HUMAN	Hs.292540
		60S RIBOSOMAL PROTEIN L13
		[H. sapiens]
ESTs	Hs.279101	EST	Hs.292545
ESTs	Hs.279102	EST, Weakly similar to ORFII [H. sapiens]	Hs.292704
ESTs	Hs.279103	EST	Hs.292761
ESTs	Hs.279104	ESTs	Hs.292803
ESTs	Hs.279105	ESTs	Hs.293183
ESTs	Hs.279106	ESTs	Hs.293280
EST	Hs.279107	ESTs	Hs.293281
ESTs	Hs.279108	ESTs, Moderately similar to 0501254A	Hs.293441
		protein Tro alpha1 H,myeloma [H. sapiens]
EST	Hs.279109	MMP13	Hs.2936
ESTs	Hs.279110	major histocompatibility complex, class II,	Hs.293934
		DR beta 4
ESTs	Hs.279111	Human MHC class III serum complement	Hs.294163
		factor B, mRNA
ESTs	Hs.279112	EST	Hs.294315
EST	Hs.279113	EST	Hs.294316
ESTs	Hs.279114	EST, Highly similar to Y196_HUMAN	Hs.295582
		HYPOTHETICAL PROTEIN KIAA0196□
		[H. sapiens]
ESTs	Hs.279115	EST	Hs.295583
ESTs	Hs.279116	EST, Highly similar to ZN07_HUMAN	Hs.295584
		ZINC FINGER PROTEIN 7 [H. sapiens]
ESTs	Hs.279117	EST	Hs.295585
ESTs	Hs.279118	EST	Hs.295586
ESTs	Hs.279119	EST, Moderately similar to angiotensin	Hs.295595
		converting enzyme [H. sapiens]
ESTs	Hs.279120	EST	Hs.295621
ESTs	Hs.279121	EST	Hs.295622
ESTs	Hs.279122	EST, Moderately similar to RL13_HUMAN	Hs.295629
		60S RIBOSOMAL PROTEIN L13
		[H. sapiens]
ESTs	Hs.279123	EST	Hs.295724
ESTs	Hs.279124	EST	Hs.296064
ESTs	Hs.279125	EST, Moderately similar to IDS_HUMAN	Hs.296070
		IDURONATE 2-SULFATASE
		PRECURSOR□ [H. sapiens]
ESTs	Hs.279126	EST	Hs.296073
ESTs	Hs.279127	interleukin enhancer binding factor 1	Hs.296281
EST	Hs.279128	similar to rat integral membrane	Hs.296429
		glycoprotein POM121
ESTs, Weakly similar to aconitase	Hs.279129	Human histocompatibility antigen mrna	Hs.296476
[H. sapiens]		clone phla-1
ESTs	Hs.279130	immunoglobulin lambda-like polypeptide 3	Hs.296552
ESTs	Hs.279131	RFXANK	Hs.296776
ESTs	Hs.279132		Hs.29826
ESTs	Hs.279133		Hs.29871
ESTs, Weakly similar to PYRG_HUMAN	Hs.279134	MEKK1	Hs.298727
CTP SYNTHASE [H. sapiens]
ESTs, Weakly similar to RIR1_HUMAN	Hs.279135		Hs.30029
RIBONUCLEOSIDE-DIPHOSPHATE
REDUCTASE M1 CHAIN [H. sapiens]
ESTs	Hs.279136	CD3e	Hs.3003
ESTs	Hs.279137	ESTs, Weakly similar to CA13_HUMAN	Hs.300697
		COLLAGEN ALPHA 1(III) CHAIN
		PRECURSOR [H. sapiens]
ESTs	Hs.279138	Homo sapiens clone BCSynL38	Hs.300865
		immunoglobulin lambda light chain variable
		region mRNA, partial cds
ESTs	Hs.279139	FCGR3A	Hs.300983
ESTs	Hs.279140	Homo sapiens DP47 gene for	Hs.301365
		immunoglobulin heavy chain, partial cds
ESTs	Hs.279141	PMS2L9	Hs.301862
EST	Hs.279142	CCR1	Hs.301921
ESTs	Hs.279143	FANCE	Hs.302003
ESTs	Hs.279144	interleukin 21	Hs.302014
ESTs	Hs.279145	interleukin 17E	Hs.302036
ESTs	Hs.279146		Hs.30446
EST	Hs.279147	EST	Hs.30709
ESTs	Hs.279148	EST	Hs.30731
ESTs	Hs.279149	MHC class II transactivator	Hs.3076
ESTs	Hs.279150	EST	Hs.30766
ESTs, Weakly similar to PUR2_HUMAN	Hs.279151	EST	Hs.30793
TRIFUNCTIONAL PURINE
BIOSYNTHETIC PROTEIN ADENOSINE
3 [H. sapiens]
ESTs	Hs.279152		Hs.30818
ESTs	Hs.279153	CD97	Hs.3107
ESTs	Hs.279154	RAR-beta2	Hs.31408
ESTs	Hs.279155	RECQL4	Hs.31442
ESTs	Hs.279156	XPC	Hs.320
ESTs	Hs.279157	ERK2	Hs.324473
ESTs	Hs.279158		Hs.32456
ESTs	Hs.279159	MSH6	Hs.3248
ESTs	Hs.279160	ribosomal protein L23-related	Hs.3254
ESTs, Weakly similar to IDHA_HUMAN	Hs.279161	PI3CG	Hs.32942
ISOCITRATE DEHYDROGENASE
[H. sapiens]
ESTs	Hs.279162	CSA (CKN1)	Hs.32967
ESTs	Hs.279163	sema domain, immunoglobulin domain (Ig),	Hs.32981
		short basic domain, secreted, (semaphorin)
		3F
ESTs	Hs.279164	BRCA2	Hs.34012
ESTs	Hs.279165	MEK1	Hs.3446
ESTs	Hs.279166	STRL33 (CXCR6)	Hs.34526
ESTs	Hs.279167	MBD4	Hs.35947
ESTs	Hs.279168	immunoglobulin (CD79A) binding protein 1	Hs.3631
EST	Hs.279169	CD7	Hs.36972
ESTs	Hs.279170	IFNA1	Hs.37026
ESTs	Hs.279171	PDGF-A	Hs.37040
EST	Hs.279172	immunoglobulin kappa variable 1-13	Hs.37089
ESTs	Hs.279174	DMC1	Hs.37181
ESTs	Hs.279175		Hs.37892
CD86	Hs.27954	Homo sapiens suppressor of variegation 3-9	Hs.37936
		(Drosophila) homolog (SUV39H) mRNA,
		and translated products.
CGI-81 protein	Hs.279583	C8B	Hs.38069
ESTs	Hs.279821	MTH1 (NUDT1)	Hs.388
ESTs	Hs.279823	Adrenomedullin	Hs.394
ESTs, Weakly similar to IRE1_HUMAN	Hs.279824		Hs.39441
IRON-RESPONSIVE ELEMENT
BINDING PROTEIN 1 [H. sapiens]
ESTs	Hs.279825	CD66b	Hs.41
ESTs	Hs.279826	RAD50	Hs.41587
MLH3	Hs.279843	CD94	Hs.41682
TNFRSF14	Hs.279899	HLJ1	Hs.41693
RPA4	Hs.283018	ESM1	Hs.41716
EST	Hs.283165	MSH3	Hs.42674
EST	Hs.283166	cAMP responsive element binding protein-	Hs.42853
		like 1
EST	Hs.283167	IKBKG	Hs.43505
EST	Hs.283168	Homo sapiens suppressor of white apricot	Hs.43543
		homolog 2 (SWAP2), mRNA.
ESTs	Hs.283169	LEU2	Hs.43628
EST	Hs.283245	Homo sapiens immunoglobulin lambda	Hs.43834
		gene locus DNA, clone:288A10
EST	Hs.283247	SIRT2	Hs.44017
ESTs	Hs.283248		Hs.44087
EST	Hs.283249	TREM2	Hs.44234
EST	Hs.283250	serine/threonine kinase 19	Hs.444
EST	Hs.283251		Hs.44512
EST	Hs.283252		Hs.44628
EST	Hs.283253		Hs.45063
EST	Hs.283254	LTC4 synthase	Hs.456
EST	Hs.283255	FUT2	Hs.46328
EST	Hs.283256	CCR6	Hs.46468
EST	Hs.283257	POLM	Hs.46964
EST	Hs.283258	EXO1 (HEX1)	Hs.47504
ESTs	Hs.283259	FEN1 (Dnase IV)	Hs.4756
EST	Hs.283261		Hs.4863
EST	Hs.283262	golgin-165	Hs.4953
EST	Hs.283263		Hs.50102
EST	Hs.283264	ATP-binding cassette, sub-family B	Hs.502
		(MDR/TAP), member 3
EST	Hs.283266		Hs.5057
ESTs	Hs.283268	corneodesmosin	Hs.507
EST	Hs.283269	Histone H2 (H2AFP)	Hs.51011
EST, Weakly similar to AF189011_1	Hs.283270	CCNH	Hs.514
ribonuclease III [H. sapiens]
EST	Hs.283271	EST	Hs.5146
EST	Hs.283272	SMUG1	Hs.5212
EST	Hs.283274	ABH (ALKB)	Hs.54418
EST	Hs.283275	CCR5	Hs.54443
EST	Hs.283276	CD81	Hs.54457
ESTs, Weakly similar to S32605 collagen	Hs.283392	TNFSF13	Hs.54673
alpha 3(VI) chain - mouse [M. musculus]
ESTs	Hs.283433	PRPS1	Hs.56
ESTs	Hs.283434		Hs.56156
ESTs	Hs.283438		Hs.56265
ESTs	Hs.283442	killer cell immunoglobulin-like receptor,	Hs.56328
		three domains, long cytoplasmic tail, 2
ESTs	Hs.283443	EST	Hs.5656
ESTs	Hs.283456		Hs.56845
ESTs	Hs.283457	MLH1	Hs.57301
ESTs, Weakly similar to similar to collagen	Hs.283458	testis specific basic protein	Hs.57692
[C. elegans]
ESTs	Hs.283459	ESTs	Hs.57841
ESTs	Hs.283460	Human 6Ckine	Hs.57907
ESTs	Hs.283462	EST	Hs.5816
ESTs	Hs.283463	Homo sapiens cell growth regulatory with	Hs.59106
		ring finger domain (CGR19) mRNA.
ESTs	Hs.283496	ERCC1	Hs.59544
ESTs	Hs.283497		Hs.61558
ESTs	Hs.283499	Homo sapiens GPI transamidase mRNA,	Hs.62187
		complete cds.
ESTs	Hs.283500		Hs.62699
ESTs, Weakly similar to ORF YDL014w	Hs.283504		Hs.63913
[S. cerevisiae]
ESTs, Weakly similar to S09646 collagen	Hs.283505	Homo sapiens chloride intracellular channel	Hs.64746
alpha 2(VI) chain precursor, medium splice		3 (CLIC3), mRNA.
form - human□ [H. sapiens]
ESTs	Hs.283608	FANCF	Hs.65328
CD42c	Hs.283743		Hs.6544
tenascin XA	Hs.283750	interleukin 1 receptor-like 1	Hs.66
immunoglobulin kappa variable 1D-8	Hs.283770	CD38	Hs.66052
protocadherin gamma subfamily A, 2	Hs.283801		Hs.6607
(PCDHGA2)
Homo sapiens mRNA; cDNA	Hs.283849	RAD54L	Hs.66718
DKFZp762F0616 (from clone
DKFZp762F0616)
Homo sapiens clone bsmneg3-t7	Hs.283876	SCYA17 (CCL17)	Hs.66742
immunoglobulin lambda light chain VJ
region, (IGL) mRNA, partial cds
Homo sapiens transgenic-JHD mouse #2357	Hs.283878	IL-12	Hs.673
immunoglobulin heavy chain variable
region (IgG VH251) mRNA, partial cds
Homo sapiens clone N97 immunoglobulin	Hs.283882	Human IL-12 p40	Hs.674
heavy chain variable region mRNA, partial
cds
Homo sapiens clone case06H1	Hs.283924	LILRB4	Hs.67846
immunoglobulin heavy chain variable
region gene, partial cds
Homo sapiens HSPC077 mRNA, partial cds	Hs.283929	interleukin 5 receptor, alpha	Hs.68876
Homo sapiens HSPC088 mRNA, partial cds	Hs.283931		Hs.6891
Homo sapiens HSPC097 mRNA, partial cds	Hs.283933		Hs.69233
Homo sapiens HSPC102 mRNA, partial cds	Hs.283934	FUT1	Hs.69747
Homo sapiens HSPC107 mRNA, partial cds	Hs.283935	B-factor, properdin	Hs.69771
CMKRL1	Hs.28408		Hs.70333
FANCA	Hs.284153		Hs.71618
Homo sapiens immunoglobulin mu chain	Hs.284277	RAD1	Hs.7179
antibody MO30 (IgM) mRNA, complete
cds
gamma-glutamyltransferase 1	Hs.284380	interleukin 19	Hs.71979
putative human HLA class II associated	Hs.285013	MEK2	Hs.72241
protein I
interleukin 13 receptor, alpha 1	Hs.285115	IL-7	Hs.72927
CDw131	Hs.285401	STAT2	Hs.72988
Homo sapiens VH2-D3.10-JH5b gene for	Hs.287403	CD42d	Hs.73734
immunoglobulin heavy chain variable
region
Homo sapiens cDNA: FLJ22546 fis, clone	Hs.287697	MIF	Hs.73798
HSI00290
Homo sapiens cDNA: FLJ23140 fis, clone	Hs.287728	ECP	Hs.73839
LNG09065
H. sapiens mRNA for HLA-C alpha chain	Hs.287811	CPN2	Hs.73858
(Cw*1701)
Homo sapiens clone ASMneg1-b1	Hs.287815	MMP8	Hs.73862
immunoglobulin lambda chain VJ region,
(IGL) mRNA, partial cds
Homo sapiens clone CPRF1-T2	Hs.287816	HLA-G histocompatibility antigen, class I,	Hs.73885
immunoglobulin lambda chain VJ region,		G
(IGL) mRNA, partial cds
EST	Hs.287817	TNFRSF9	Hs.73895
myelin protein zero-like 1	Hs.287832	IL-4	Hs.73917
immunoglobulin lambda-like polypeptide 1	Hs.288168	HLA-DQB1	Hs.73931
cathepsinB	Hs.288181	RAG1	Hs.73958
G18.2 protein	Hs.288316	LAG-3	Hs.74011
ESTs	Hs.288403		Hs.7402
EST	Hs.288431	CD163	Hs.74076
Homo sapiens partial IGVH2 gene for	Hs.288553	immunoglobulin superfamily, member 2	Hs.74115
immunoglobulin heavy chain V region, case
2, cell B 45
polymeric immunoglobulin receptor	Hs.288579	CD158b	Hs.74134
Human immunoglobulin heavy chain	Hs.288711		Hs.7434
variable region (V4-4) gene, partial cds
Human immunoglobulin heavy chain	Hs.289036	TCRa	Hs.74647
variable region (V4-4b) gene, partial cds
	Hs.28921	human immunodeficiency virus type I	Hs.75063
		enhancer-binding protein 2
EST	Hs.289577	MLN50	Hs.75080
EST	Hs.289836	lysyl hydroxylase (PLOD)	Hs.75093
EST	Hs.289878	TAK1	Hs.7510
GSN	Hs.290070	Homo sapiens transcription factor 6-like 1	Hs.75133
		(mitochondrial transcription factor 1-like)
		(TCF6L1) mRNA.
EST, Weakly similar to unnamed protein	Hs.290133	UBE2N (UBC13, BTG1)	Hs.75355
product [H. sapiens]
EST	Hs.290227		Hs.75450
ESTs	Hs.290315	HSPA2	Hs.75452
EST	Hs.290339	CD151	Hs.75564
EST	Hs.290340	RELA	Hs.75569
	Hs.29055	CD122	Hs.75596
EST	Hs.291125	CD14	Hs.75627
EST	Hs.291126	nuclear factor erythroid 2 isoform f = basic	Hs.75643
		leucine zipper protein {alternatively spliced}
CD91 = LRP	Hs.89137	C1QB	Hs.8986
XPF (ERCC4)	Hs.89296	superkiller viralicidic activity 2 (S.	Hs.89864
		cerevisiae homolog)-like
Carbonic anhydrase IV	Hs.89485	EST	Hs.90165
CETP	Hs.89538	EST	Hs.90171
RAD52	Hs.89571	GTF2H3	Hs.90304
GTF2H1	Hs.89578	protein tyrosine kinase related sequence	Hs.90314
Fc fragment of IgE, high affinity I, receptor	Hs.897		Hs.90463
for; alpha polypeptide
transcript ch138	Hs.94881	SGRF protein, Interleukin 23 p19 subunit	Hs.98309
	Hs.9578	XRCC1	Hs.98493
IL-9	Hs.960	Homo sapiens mRNA for KIAA0543	Hs.98507
		protein, partial cds.
NFATC1	Hs.96149		Hs.9893
OGG1	Hs.96398	DIR1 protein	Hs.99134
	Hs.96499	XRCC3	Hs.99742
NFKBIB	Hs.9731	Elastase(leukocyte)	Hs.99863
XAB2 (HCNP)	Hs.9822	JAK3	Hs.99877
CD40	Hs652