Methods and compositions for molecular profiling for disease diagnosis

Abstract

To provide a method of diagnosing thyroid cancer with early and accurate diagnosis.SOLUTION: A method of diagnosing thyroid cancer comprises the steps of: (a) obtaining from the subject a biological sample 100 comprising gene expression products; (b) measuring an expression level for one or more gene expression products of the biological sample 100, where the one or more gene expression products correspond to genes selected from a specific list; and (c) identifying the biological sample as cancerous by correlating the gene expression level with the presence of thyroid cancer in the biological sample.SELECTED DRAWING: Figure 18A

Description

[Technical field]

[0001] cross reference This application claims the benefit of U.S. Provisional Application No. 61 / 199,585, filed November 17, 2008, entitled "Methods and Compositions of Molecular Profiling for Diagnosis of Cancer," and U.S. Provisional Application No. 61 / 270,812, filed July 13, 2009, entitled "Methods and Compositions of Molecular Profiling for Diagnosis of Cancer," both of which are incorporated by reference in their entireties. [Background technology]

[0002] 2. Background of the Invention Cancer is the second leading cause of death in the United States and one of the leading causes of death worldwide. Nearly 25 million people are currently living with cancer, with 11 million new cases diagnosed each year. Furthermore, as the general population continues to age, cancer is expected to become an even greater problem. The World Health Organization estimates that by 2020, global cancer rates are expected to increase by 50%.

[0003] Successful cancer treatment begins with an early and accurate diagnosis. Current diagnostic methods include cytological examination of tissue samples taken by biopsy, or imaging of tissues and organs for evidence of abnormal cell proliferation. Although these techniques have proven useful and low-cost, they have a number of drawbacks. First, cytological analysis and imaging techniques for cancer diagnosis often require subjective assessment to determine the likelihood of malignancy. Second, the increased use of these techniques has led to a sharp increase in the number of indeterminate results that fail to provide a definitive diagnosis. Third, these routine diagnostic methods lack a precise method for determining the probability of an accurate diagnosis. Fourth, these techniques cannot detect malignant proliferation at very early stages. Fifth, these techniques do not provide information regarding the basis of abnormal cell proliferation.

[0004] Many of the newer generation cancer treatments show greatly reduced side effects, but they specifically target certain metabolic or signaling pathways and are only effective against cancers that rely on those pathways. Furthermore, the cost of treatments can be prohibitive for individuals, insurance providers, or government entities. This cost could be at least partially offset by improved methods of accurately diagnosing cancers and the pathways they rely on at an early stage. These improved methods would be useful both to prevent unnecessary therapeutic interventions and to direct treatment.

[0005] In the case of thyroid cancer, of the approximately 130,000 thyroidectomy surgeries performed annually in the United States due to presumed malignancy, only about 54,000 are necessary. Thus, approximately 76,000 unnecessary surgeries are performed annually. In addition, there are ongoing treatment costs and complications due to the need for lifelong drug therapy to replace lost thyroid function. Thus, improved testing modalities and commercial practices are needed to improve current methods of cancer diagnosis.

[0006] The thyroid gland has at least two types of cells that make hormones. Follicular cells make thyroid hormones that affect heart rate, body temperature, and energy levels. C cells make calcitonin, a hormone that helps control blood calcium levels. Abnormal growths in the thyroid gland can result in the formation of nodules that can be either benign or malignant. Thyroid cancer includes at least four different types of malignant tumors of the thyroid gland: papillary, follicular, medullary, and anaplastic. Summary of the Invention

[0007] The present invention includes a method of diagnosing thyroid disease in a subject comprising the steps of: (a) providing a nucleic acid sample from a subject; (b) detecting the amount of one or more genes, gene products, or transcripts selected from the group consisting of the genes or transcripts listed in Table 2, or their complements; and (c) determining whether the subject has or is likely to have a malignant or benign thyroid condition based on the results of step (b).

[0008] The present invention also includes compositions comprising one or more binding agents that specifically bind to one or more polymorphisms selected from the group consisting of the polymorphisms listed in the table. The basic features and various aspects of the present invention are listed below. [1] 2. A method for diagnosing thyroid cancer in a subject, comprising the steps of: (a) obtaining a biological sample from said subject comprising a gene expression product; (b) determining the expression level of one or more gene expression products corresponding to a gene selected from Figure 6 from the biological sample; and (c) correlating the gene expression level with the presence of thyroid cancer in the biological sample, thereby identifying the biological sample as cancerous. [2] The method described in [1], wherein the biological sample comprises thyroid cells. [3] The method described in [1], wherein the biological sample is a tissue sample. [4] The method of [1], wherein the sample is obtained by needle aspiration, fine needle aspiration, core needle biopsy, aspiration biopsy, large core biopsy, incisional biopsy, excision biopsy, punch biopsy, shave biopsy, or skin biopsy. [5] The method described in [1], wherein the gene expression product is RNA. [6] The method described in [5], wherein the gene expression product is mRNA, rRNA, tRNA, or miRNA. [7] The method of [5], wherein RNA expression levels are measured by microarray, SAGE, blotting, RT-PCR, or quantitative PCR. [8] The method described in [1], wherein the gene expression product is a protein. [9] The method of [8], wherein protein gene expression levels are measured by ELISA, mass spectrometry, blotting, proteomic techniques, or immunohistochemistry.

[10] comparing said expression levels of said one or more gene expression products to a control expression level for each gene expression product in a control sample, wherein if there is a difference in gene expression level of a gene expression product between said biological sample and said control sample, said biological sample is identified as cancerous. The method of [1], further comprising:

[11] The method of

[10] , wherein the difference in mRNA levels between the biological sample and the control sample is compared using a trained algorithm.

[12] The method of [1], wherein the one or more gene expression products correspond to genes selected from Table 3, Table 4 or Table 5.

[13] The method of [1], wherein the one or more gene expression products correspond to genes selected from one or more of Lists 1 to 31 or Table 6.

[14] The method of [1], wherein the gene expression product is derived from an acute myeloid leukemia signaling pathway, a type 2 somatostatin receptor signaling pathway, a cAMP-mediated signaling pathway, a cell cycle and DNA damage checkpoint signaling pathway, a G protein-coupled receptor signaling pathway, an integrin signaling pathway, a melanoma cell signaling pathway, a relaxin signaling pathway, or a thyroid cancer signaling pathway.

[15] The method of [1], wherein the one or more gene expression products are derived from an adhesion pathway, an ECM pathway, a thyroid cancer pathway, a focal adhesion pathway, an apoptosis pathway, a p53 pathway, a tight junction pathway, a TGFβ pathway, an ErbB pathway, a Wnt pathway, a pathway in cancer overview, a cell cycle pathway, a VEGF pathway, a Jak / STAT pathway, a MAPK pathway, a PPAR pathway, an mTOR pathway, or an autoimmune thyroid pathway.

[16] The method described in

[15] , wherein the gene expression product is derived from an adhesion pathway.

[17] The method described in

[15] , wherein at least two gene expression products are measured and the gene expression products are derived from at least two different pathways.

[18] The method described in

[17] , wherein at least one gene expression product is derived from an adhesion pathway and at least one gene expression product is derived from a focal adhesion pathway.

[19] The method of

[18] , further comprising a second gene expression product from the tight junction pathway.

[20] The method of

[15] , wherein the gene expression products are categorized into multiple pathways.

[21] The method of

[20] , wherein the gene expression products are categorized into the adhesion pathway and the focal adhesion pathway.

[22] The method of

[21] , wherein the gene expression product is also categorized as the tight junction pathway.

[23] The method according to [1], wherein the gene expression product is derived from a cell senescence ontology group, a cell cortex ontology group, a cell cycle ontology group, a cell death / apoptosis ontology group, a cell differentiation ontology group, a cell division ontology group, a cell-cell junction ontology group, a cell migration ontology group, a cell morphogenesis ontology group, a cell motility ontology group, a cell process ontology group, a cell proliferation ontology group, a cell recognition ontology group, a cell body ontology group, a cell surface ontology group, a cell surface-bound receptor signaling ontology group, a cell adhesion ontology group, a transcription ontology group, a immune response ontology group, an angiogenesis ontology group, a cell membrane ontology group, a rare membrane component ontology group, a cell apical membrane ontology group, a cell basolateral or cell lateral membrane ontology group, an integrin ontology group, a cell surface ontology group, an extracellular space ontology group, a cytoskeleton ontology group, or an inflammation ontology group.

[24] The method of [1], wherein one or more genes listed in Figure 6 contain representative sequences corresponding to a set of metabolic or signaling pathways indicative of cancer.

[25] The method of [1], further comprising the results of step (d) providing a suggested therapeutic intervention.

[26] The method described in [1], wherein the results of the expression analysis provide a statistical confidence level of greater than 90% that a given diagnosis is accurate.

[27] The method described in [1], wherein the results of the expression analysis provide a statistical confidence level of greater than 95% that a given diagnosis is accurate.

[28] performing a cytological analysis on a portion of said biological sample after step (a) to obtain a preliminary diagnosis. The method of [1], further comprising:

[29] The method of

[28] , wherein samples with a preliminary diagnosis of indeterminate samples and non-diagnostic samples are further analyzed by the methods of steps (b) and (c).

[30] The method of [1], wherein the genes selected from Figure 6 are further selected from the genes listed in Figure 2, Figure 3, Figure 4, or Figure 5.

[31] determining the expression levels of at least two of the gene expression products in said biological sample. The method described in [1], comprising:

[32] determining the expression levels of at least five of the gene expression products in said biological sample. The method described in [1], comprising:

[33] determining the expression levels of at least ten of the gene expression products of said biological sample; The method described in [1], comprising:

[34] The method described in [1], wherein the difference in gene expression levels is at least 10%.

[35] The method described in [1], wherein the difference in gene expression levels is at least 25%.

[36] The method described in [1], wherein the difference in gene expression levels is at least 50%.

[37] The method described in [1], wherein the difference in gene expression levels is at least two-fold.

[38] The method described in [1], wherein the difference in gene expression levels is at least 5-fold.

[39] The method described in [1], wherein the difference in gene expression levels is at least 10-fold.

[40] The method of [1], wherein the biological sample is identified as cancerous with greater than 75% accuracy.

[41] The method of [1], wherein the biological sample is identified as cancerous with a sensitivity of greater than 70%.

[42] The method of [1], wherein the biological sample is identified as cancerous with a specificity of greater than 70%.

[43] The method of [1], wherein the biological sample is identified as cancerous with a sensitivity of greater than 70% and a specificity of greater than 70%.

[44] The method of [1], wherein the biological sample is identified as cancerous with a positive predictive value of greater than 95%.

[45] The method of [1], wherein the biological sample is identified as cancerous with a negative predictive value of greater than 95%.

[46] The method of any one of

[40] to

[45] , wherein the accuracy is calculated using a trained algorithm.

[47] The method of [1], wherein the control sample is obtained from the same individual as the test sample, a different individual from the test sample, a tissue bank, or a cell bank.

[48] The method according to [1], wherein the gene level of the gene expression product of the control sample is derived from a database.

[49] The method described in [1], wherein the sample is a preoperative specimen.

[50] The method according to [1], wherein the sample is a post-operative specimen.

[51] The method described in [1], which distinguishes malignant thyroid cancer from benign thyroid disease.

[52] The method of claim 1, wherein the biological sample is identified as follicular adenoma, follicular carcinoma, lymphocytic thyroiditis, follicular variant papillary thyroid carcinoma, papillary thyroid carcinoma, nodular hyperplasia, medullary thyroid carcinoma, Hurthle cell carcinoma, Hurthle cell adenoma, or anaplastic thyroid carcinoma.

[53] The method described in [1], wherein metastatic non-thyroid cancer is identified in the biological sample.

[54] The method of [1], wherein metastatic parathyroid carcinoma, metastatic melanoma, metastatic renal carcinoma, metastatic breast cancer, or metastatic B-cell lymphoma is identified in the biological sample.

[55] communicating the result set having the identified information; The method of [1], further comprising:

[56]

[55] The method according to claim 1, wherein the information is transmitted via the Internet.

[57] A method of classifying cancer, comprising the steps of: (a) obtaining a biological sample containing a gene expression product; (b) determining the expression level of one or more gene expression products corresponding to a gene selected from Figure 6 from the biological sample; (c) comparing the expression level of the one or more gene expression products to a control expression level for each gene expression product in a control sample; and (d) identifying the biological sample as positive for the cancer subtype if there is a difference in gene expression levels between the biological sample and the control sample at a specified confidence level.

[58] The method of claim 57, further comprising identifying follicular adenoma, follicular carcinoma, lymphocytic thyroiditis, follicular type papillary thyroid carcinoma, papillary thyroid carcinoma, nodular hyperplasia, medullary thyroid carcinoma, Hurthle cell carcinoma, Hurthle cell adenoma, or anaplastic thyroid carcinoma in the biological sample.

[59] The method of

[57] , further comprising identifying metastatic non-thyroid cancer in the biological sample.

[60] The method of

[57] , wherein metastatic parathyroid carcinoma, metastatic melanoma, metastatic renal carcinoma, metastatic breast cancer, or metastatic B-cell lymphoma is identified in the biological sample.

[61] A composition for diagnosing cancer comprising a probe which binds to a portion of one or more of the gene expression products listed in FIG. 6, or to their complements.

[62] The composition described in

[61] , wherein the probe is an antibody.

[63] The composition described in

[61] , wherein the probe is an oligonucleotide.

[64] a substrate to which the probes are covalently attached;

[61] The composition according to

[61] , further comprising:

[65]

[64] The composition described in

[64] , which is suitable for use in diagnosing cancer at a specified confidence level using a trained algorithm.

[66] The composition described in

[65] , wherein the cancer is thyroid cancer.

[67] The composition of

[61] , wherein the probe binds to one or more gene expression products corresponding to genes selected from Table 3, Table 4, or Table 5.

[68] The composition described in

[61] , wherein the probe binds to one or more gene expression products corresponding to genes selected from one or more of Lists 1 to 31 or Table 6.

[69] The composition of claim 61, wherein the probe binds to one or more portions of genes from an acute myeloid leukemia pathway, a type 2 somatostatin receptor pathway, a cAMP-mediated pathway, a cell cycle and DNA damage checkpoint pathway, a G protein-coupled receptor pathway, an integrin pathway, a melanoma cell pathway, a relaxin pathway, or a thyroid cancer signaling pathway.

[70] The composition of claim 61, wherein the probe binds to one or more portions of genes from an adhesion pathway, an ECM pathway, a thyroid cancer pathway, a focal adhesion pathway, an apoptosis pathway, a p53 pathway, a tight junction pathway, a TGFβ pathway, an ErbB pathway, a Wnt pathway, a cancer profile pathway, a cell cycle pathway, a VEGF pathway, a Jak / STAT pathway, a MAPK pathway, a PPAR pathway, an mTOR pathway, or an autoimmune thyroid pathway.

[71] The composition of

[61] , wherein there are at least two probes that bind to gene expression products from at least two different pathways.

[72] The composition described in

[71] , wherein at least one gene expression product is derived from the adhesion pathway and at least one gene expression product is derived from the focal adhesion pathway.

[73] The composition described in

[72] , further comprising a gene expression product from the tight junction pathway.

[74] The composition described in

[61] , wherein the gene expression products are categorized into multiple pathways.

[75] The composition described in

[74] , wherein the gene expression products are categorized into adhesion pathways and focal adhesion pathways.

[76] The composition described in

[75] , wherein the gene expression product is also categorized as a tight junction pathway.

[77] The composition according to

[61] , wherein the gene expression product comprises one or more parts of genes from a cellular senescence ontology group, a cell cortex ontology group, a cell cycle ontology group, a cell death / apoptosis ontology group, a cell differentiation ontology group, a cell division ontology group, a cell-cell junction ontology group, a cell migration ontology group, a cell morphogenesis ontology group, a cell motility ontology group, a cell process ontology group, a cell proliferation ontology group, a cell recognition ontology group, a cell body ontology group, a cell surface ontology group, a cell surface-bound receptor signaling ontology group, a cell adhesion ontology group, a transcription ontology group, a immune response ontology group, an angiogenesis ontology group, a cell membrane ontology group, a rare membrane component ontology group, a cell apical membrane ontology group, a cell basolateral or cell lateral membrane ontology group, an integrin ontology group, a cell surface ontology group, an extracellular space ontology group, a cytoskeleton ontology group, or an inflammation ontology group.

[78] A method of conducting a molecular profiling business, comprising the steps of: (a) providing a cancer diagnostic or monitoring service by measuring the level of a gene expression product according to [1]; or (b) using the composition according to

[61] to measure the amount of one or more gene expression products according to [1]; (c) advising on diagnosis, prognosis, and required treatment based on the data obtained in (a) or (b); and (d) marketing the services of (a), (b) or (c) to physicians, hospitals and clinics, either collectively or independently;

[0009] INCORPORATION BY REFERENCE All publications and patent applications mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication or patent application was specifically and individually indicated to be incorporated by reference.

[0010] The novel features of the invention are set forth with particularity in the appended claims. A better understanding of the features and advantages of the present invention will be obtained by reference to the following detailed description that sets forth illustrative embodiments in which the principles of the invention are utilized, and the accompanying drawings, as follows: [Brief description of the drawings]

[0011] [Figure 1-1] 1 is a table listing 75 thyroid samples that were interrogated for gene expression analysis using the Affymetrix Human Exon 10ST array to identify genes that are significantly differentially expressed or alternatively spliced ​​between malignant, benign, and normal samples. The name and pathological classification of each sample are listed. [Figure 1-2] This figure shows a continuation of Figure 1-1. [Figure 1-3] This figure shows a continuation of Figure 1-2. [Figure 2-1] Table listing the top 100 differentially expressed genes at the gene level. Data is from a dataset where benign, malignant and normal thyroid samples were compared at the gene level. Markers were selected based on statistical significance after Benjamini and Hochberg correction for false discover rate (FDR). Positive numbers indicate upregulation of expression and negative numbers indicate downregulation. [Figure 2-2] This figure shows a continuation of Figure 2-1. [Figure 2-3] This figure shows the continuation of Figure 2-2. [Figure 2-4] This figure shows the continuation of Figure 2-3. [Figure 3-1] Table listing the top 100 alternatively spliced ​​genes. Data is from a dataset in which benign, malignant, and normal thyroid samples were compared at the gene level. Markers were selected based on statistical significance after Benjamini and Hochberg correction for false discovery rate (FDR). [Figure 3-2] This figure shows a continuation of Figure 3-1. [Figure 3-3] This figure shows the continuation of Figure 3-2. [Diagram 3-4] This figure shows the continuation of Figure 3-3. [Figure 4-1] 1 is a table listing the top 100 differentially expressed genes at the probe set level. Data is from the probe set dataset. Positive numbers indicate upregulation of expression, and negative numbers indicate downregulation. [Figure 4-2] This figure shows a continuation of Figure 4-1. [Figure 4-3] This figure shows the continuation of Figure 4-2. [Figure 4-4] This figure shows the continuation of Figure 4-3. [Figure 5-1] This table lists the top 100 significant diagnostic markers determined by gene level analysis.The markers in this list show both differential gene expression and alternative exon splicing.Positive numbers mean upregulation, and negative numbers mean downregulation.This table lists three sets of calculated fold changes for a given marker, which allow comparison between malignant and benign groups, between benign and normal groups, and between malignant and normal groups. [Figure 5-2] This figure shows a continuation of Figure 5-1. [Figure 5-3] This figure shows the continuation of Figure 5-2. [Figure 5-4]This figure shows the continuation of Figure 5-3. [Figure 6-1] 1 is a table listing genes identified by molecular profiling of gene expression levels and / or alternative exon splicing as contributing to thyroid cancer diagnosis. Markers identified from datasets in which benign, malignant, and normal samples are analyzed at the gene level are referred to as BMN in the data source column; similarly, markers identified from datasets in which benign and malignant samples are analyzed at the gene level are referred to as BM in the data source column. Similarly, markers identified from datasets in which benign and malignant samples are analyzed at the probe set level are referred to as probe sets in the data source column. [Figure 6-2] This figure shows the continuation of Figure 6-1. [Figure 6-3] This figure shows the continuation of Figure 6-2. [Figure 6-4] This figure shows the continuation of Figure 6-3. [Figure 6-5] This figure shows the continuation of Figure 6-4. [Figure 6-6] This figure shows the continuation of Figure 6-5. [Figure 6-7] This figure shows the continuation of Figure 6-6. [Figure 6-8] This is a continuation of Figure 6-7. [Figure 6-9] This figure shows the continuation of Figure 6-8. [Figure 6-10] This figure shows a continuation of Figure 6-9. [Figure 6-11] This figure shows the continuation of Figure 6-10. [Figure 6-12] This figure shows a continuation of Figure 6-11. [Figure 6-13] This figure shows a continuation of Figure 6-12. [Figure 6-14] This figure shows the continuation of Figure 6-13. [Figure 6-15] This figure shows the continuation of Figure 6-14. [Figure 6-16] This figure shows the continuation of Figure 6-15. [Figure 6-17] This is a continuation of Figure 6-16. [Figure 6-18] This is a continuation of Figure 6-17. [Figure 6-19] This is a continuation of Figure 6-18. [Figure 6-20] This is a continuation of Figure 6-19. [Figure 6-21] This is a continuation of Figure 6-20. [Figure 6-22] This is a continuation of Figure 6-21. [Figure 6-23] This is a continuation of Figure 6-22. [Figure 6-24] This is a continuation of Figure 6-23. [Figure 6-25] This is a continuation of Figure 6-24. [Figure 6-26] This is a continuation of Figure 6-25. [Figure 6-27] This is a continuation of Figure 6-26. [Figure 6-28] This is a continuation of Figure 6-27. [Figure 6-29] This is a continuation of Figure 6-28. [Figure 6-30] This is a continuation of Figure 6-29. [Figure 6-31] This is a continuation of Figure 6-30. [Figure 6-32] This is a continuation of Figure 6-31. [Figure 6-33] This is a continuation of Figure 6-32. [Figure 6-34] This is a continuation of Figure 6-33. [Figure 6-35] This is a continuation of Figure 6-34. [Figure 6-36] This is a continuation of Figure 6-35. [Figure 6-37] This is a continuation of Figure 6-36. [Figure 6-38] This is a continuation of Figure 6-37. [Figure 6-39] This is a continuation of Figure 6-38. [Figure 6-40] This is a continuation of Figure 6-39. [Figure 6-41]This is a continuation of Figure 6-40. [Figure 6-42] This is a continuation of Figure 6-41. [Figure 6-43] This is a continuation of Figure 6-42. [Figure 6-44] This is a continuation of Figure 6-43. [Figure 6-45] This is a continuation of Figure 6-44. [Figure 7-1] Table listing tissue samples investigated for gene expression analysis. Samples were classified as benign (B) or malignant (M) by pathological analysis. Benign samples were further classified as follicular adenoma (FA), lymphocytic thyroiditis (LCT), or nodular hyperplasia (NHP). Malignant samples were further classified as Hurthle cell carcinoma (HC), follicular carcinoma (FC), follicular variant papillary thyroid carcinoma (FVPTC), papillary thyroid carcinoma (PTC), medullary thyroid carcinoma (MTC), or anaplastic carcinoma (ATC). [Figure 7-2] This figure shows a continuation of Figure 7-1. [Figure 7-3] This figure shows the continuation of Figure 7-2. [Figure 7-4] This figure shows the continuation of Figure 7-3. [Figure 7-5] This figure shows the continuation of Figure 7-4. [Figure 7-6] This figure shows the continuation of Figure 7-5. [Figure 7-7] This figure shows the continuation of Figure 7-6. [Figure 8-1] 1 is a table listing fine needle aspirate samples examined for gene expression analysis. Samples were classified as benign (B) or malignant (M) by pathological analysis. Benign samples were further classified as follicular adenoma (FA), lymphocytic thyroiditis (LCT), Hurthle cell adenoma (HA) or nodular hyperplasia (NHP). Malignant samples were further classified as Hurthle cell carcinoma (HC), follicular carcinoma (FC), follicular type papillary thyroid carcinoma (FVPTC), papillary thyroid carcinoma (PTC), medullary thyroid carcinoma (MTC), or anaplastic carcinoma (ATC). [Figure 8-2]This figure shows a continuation of Figure 8-1. [Figure 8-3] This figure shows the continuation of Figure 8-2. [Figure 9-1] 8 is a table listing genes identified from expression analysis of tissue samples listed in FIG. 7 that show significant differences in expression between malignant and benign samples as determined by feature selection using LIMMA (linear models for micro array data) and SVM (support vector machine) for classification of malignant and benign samples. Rank refers to the significance of the marker after Benjamini and Hochberg correction for false discovery rate (FDR) (lower rank means higher significance). Gene symbol refers to the name of the gene. TCID refers to the transcript cluster ID of the gene used in the Affymetrix Human Exon 10ST array. Ref Seq refers to the name of the reference sequence corresponding to the gene. The column labeled "Newly discovered markers" refers to gene expression markers not previously described to be differentially expressed in malignant and benign thyroid tissues. [Figure 9-2] This figure shows a continuation of Figure 9-1. [Figure 9-3] This figure shows the continuation of Figure 9-2. [Figure 10-1]9 is a table listing genes identified from expression analysis of tissue samples listed in FIG. 8 that show significant differences in expression between medullary thyroid carcinoma (MTC) and other pathologies as determined by feature selection using LIMMA (Linear Model for Microarray Data) and SVM (Support Vector Machine) for classification of MTC samples from other samples. Rank refers to the significance of the marker after Benjamini and Hochberg correction for false discovery rate (FDR) (lower rank means higher significance). Gene symbol refers to the name of the gene. TCID refers to the transcript cluster ID of the gene used in the Affymetrix Human Exon 10ST array. P-values ​​indicate the statistical significance of differential expression between MTC and non-MTC samples. Fold change indicates the degree of differential expression between MTC and non-MTC samples. The column labeled "Newly discovered markers" refers to gene expression markers not previously described to be differentially expressed in malignant and benign thyroid tissues. [Figure 10-2] This figure shows a continuation of Figure 10-1. [Figure 10-3] This figure shows a continuation of Figure 10-2. [Figure 11-1] FIG. 9 is a table listing genes identified from expression analysis of the samples listed in FIGS. 7 and 8 that show significant differences in expression between benign and malignant samples as determined by a reproducibility-based meta-analysis classification algorithm. [Figure 11-2] This figure shows a continuation of Figure 11-1. [Figure 11-3] This figure shows a continuation of Figure 11-2. [Figure 11-4] This figure shows a continuation of Figure 11-3. [Figure 12-1]1 is a table listing genes identified from expression analysis of samples listed in FIGS. 7 and 8 that exhibit significant differences in expression (posterior probability >.9) between benign and malignant samples as determined by Bayesian ranking of differentially expressed genes. To determine prior probabilities, type I and type II error rates were derived from previously published studies, and these prior probabilities were combined with the output of the dataset derived from expression analysis of samples listed in FIG. 10 to estimate the posterior probability of differential gene expression, and then the results of expression analysis of samples listed in FIG. 11 were combined with the estimated posterior probability to calculate the final posterior probability of differential gene expression. These posterior probabilities were then used to rank the differentially expressed genes. [Figure 12-2] This figure shows a continuation of Figure 12-1. [Figure 12-3] This figure shows a continuation of Figure 12-2. [Figure 13-1] 8 is a table listing genes identified from expression analysis of the samples listed in FIG. 7 that show differential expression among samples categorized as FA, LCT, NHP, HC, FC, FVPTC, PTC, MTC, or ATC as determined by feature selection using LIMMA (Linear Models for Microarray Data) and SVM (Support Vector Machine) for classification. [Figure 13-2] This figure shows a continuation of Figure 13-1. [Figure 13-3] This figure shows a continuation of Figure 13-2. [Figure 13-4] This figure shows a continuation of Figure 13-3. [Figure 13-5] This figure shows a continuation of Figure 13-4. [Figure 13-6] This figure shows a continuation of Figure 13-5. [Figure 13-7] This figure shows a continuation of Figure 13-6. [Figure 13-8] This figure shows a continuation of Figure 13-7. [Figure 14-1]Table listing fine needle aspirate samples interrogated for microRNA (miRNA) expression analysis using the Agilent Human v2 miRNA microarray chip. Samples were classified as benign (B) or malignant (M) by pathological analysis. Benign samples were further classified as follicular adenoma (FA) or nodular hyperplasia (NHP). Malignant samples were further classified as follicular carcinoma (FC), follicular type papillary thyroid carcinoma (FVPTC), papillary thyroid carcinoma (PTC), or medullary thyroid carcinoma (MTC). [Figure 14-2] This figure shows a continuation of Figure 14-1. [Figure 14-3] This figure shows the continuation of Figure 14-2. [Figure 15] Table listing fine needle aspirate samples interrogated for microRNA (miRNA) expression analysis using Illumina Human v2 miRNA arrays. Samples were classified as benign (B), nondiagnostic, or malignant (M) by pathology analysis. Benign samples were further classified as benign nodule (BN), follicular neoplasm (FN), (LCT), or (NHP). Malignant samples were further classified as FVPTC or PTC. [Figure 16] FIG. 15 is a table listing microRNAs (miRNAs) identified from the analysis of the samples listed in FIG. 14 that show differential expression between samples categorized as benign or malignant. The miRNA column refers to the name of the miRNA. The CHR column refers to the chromosome in which the miRNA is located. The P column refers to the statistical confidence or p-value provided by the analysis. The DE column indicates whether the listed miRNA is upregulated in malignant samples (1) or downregulated in malignant samples (-1). The Patent column refers to patents or applications that describe these miRNAs. [Figure 17-1]FIG. 16 is a table listing microRNAs (miRNAs) identified from the analysis of samples listed in FIG. 15 that show differential expression between samples categorized as benign or malignant. The miRNA column refers to the name of the miRNA. The Probe ID column refers to the corresponding probe ID in the Illumina array. The CHR column refers to the chromosome where the miRNA is located. The P column refers to the statistical confidence or p-value provided by the analysis. The DE column indicates whether the listed miRNA is upregulated in malignant samples (no symbol) or downregulated in malignant samples (minus symbol). The Rep column refers to the reproducibility score provided by a "hot probe" type analysis of the hybridization data. The Patent column refers to patents or applications that describe these miRNAs. [Figure 17-2] This figure shows a continuation of Figure 17-1. [Figure 17-3] This figure shows the continuation of Figure 17-2. [Figure 17-4] This figure shows the continuation of Figure 17-3. [Figure 17-5] This figure shows the continuation of Figure 17-4. [Figure 18A] 1 is a flow chart describing how molecular profiling can be used to improve the accuracy of routine cytological investigations. Another aspect of the molecular profiling business is described. [Figure 18B] 1 is a flow chart describing how molecular profiling can be used to improve the accuracy of routine cytological investigations. Another aspect of the molecular profiling business is described. [Figure 19] 1 is an illustration of a kit offered by a molecular profiling business. [Figure 20] 1 is an example of a molecular profiling result report. [Figure 21]Shown are computers useful in the methods of the invention for displaying, storing, retrieving, or calculating diagnostic results from molecular profiling; displaying, storing, retrieving, or calculating raw data from genomic or nucleic acid expression analysis; or displaying, storing, retrieving, or calculating sample information or customer information. [Figure 22] Titration curves of error rate versus number of genes using the SVM-based classification algorithm are shown. The titration curves reach a plateau when the classification algorithm examines 200-250 genes. These data indicate that the overall error rate of the algorithm was 4% (5 / 138). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Detailed Description of the Invention I. Introduction The present disclosure provides a novel method for diagnosing abnormal cell proliferation from biological test samples, and related kits and compositions.The present invention also provides methods and compositions for the differential diagnosis of abnormal cell proliferation types, such as follicular carcinoma (FC), follicular type papillary thyroid carcinoma (FVPTC), Hurthle cell carcinoma (HC), Hurthle cell adenoma (HA); cancer, including papillary thyroid carcinoma (PTC), medullary thyroid carcinoma (MTC), and anaplastic carcinoma (ATC); adenoma, including follicular adenoma (FA); nodular hyperplasia (NHP); colloid nodule (CN); benign nodule (BN); follicular neoplasm (FN); lymphocytic thyroiditis (LCT), including lymphocytic autoimmune thyroiditis; parathyroid tissue; renal cancer metastasis to thyroid; melanoma metastasis to thyroid; B-cell lymphoma metastasis to thyroid; breast cancer metastasis to thyroid; benign (B) tumor, malignant (M) tumor, and normal (N) tissue. The present invention further provides novel markers, including microRNA (miRNA) and gene expression product markers, and novel groups of genes and markers, useful for the diagnosis, characterization, and treatment of cell proliferation. Additionally, the present invention provides business methods that provide enhanced diagnosis, differential diagnosis, monitoring, and treatment of cell proliferation.

[0013] Cancer is the leading cause of death in the United States. Early and accurate diagnosis of cancer is crucial for effective management of the disease. Therefore, it is important to develop testing modalities and commercial practices to enable more accurate and earlier cancer diagnosis. Expression product profiling, also called molecular profiling, provides a powerful method for early and accurate diagnosis of tumors or other types of cancer from biological samples.

[0014] Typically, screening for the presence of tumors or other types of cancer involves analyzing biological samples taken by various methods, such as biopsy. The biological samples are then prepared and examined by a person skilled in the art. Preparation methods can include, but are not limited to, various cytological staining and immunohistochemical methods. Unfortunately, traditional methods of cancer diagnosis have a number of deficiencies. These deficiencies include: (1) diagnosis requires subjective assessment and therefore may be prone to inaccuracy and lack of reproducibility; (2) the methods may not determine the genetic, metabolic, or signaling pathways underlying the resulting pathogenesis; (3) the methods may not provide a quantitative assessment of the test results; and (4) the methods may not provide a definitive diagnosis for certain samples.

[0015] One hallmark of cancer is the dysregulation of normal transcriptional control, resulting in the abnormal expression of genes or other RNA transcripts, such as miRNAs. Aberrantly expressed transcripts include genes involved in cell transformation, such as tumor suppressors and oncogenes. Tumor suppressor genes and oncogenes may be upregulated or downregulated in tumors compared to normal tissues. Known tumor suppressors and oncogenes include, but are not limited to, brca1, brca2, bcr-abl, bcl-2, HER2, N-myc, C-myc, BRAF, RET, Ras, KIT, Jun, Fos, and p53. This abnormal expression may occur through a variety of different mechanisms. It is not necessary for the present invention to understand the mechanism of abnormal expression, or the mechanism by which carcinogenesis occurs. Nevertheless, finding a marker or set of markers in a sample whose expression is upregulated or downregulated compared to a normal sample can indicate cancer. Furthermore, a particular abnormally expressed marker or set of markers can indicate a particular type of cancer, or even a recommended treatment protocol. Furthermore, it is understood that the methods of the present invention are not limited solely to canonically defined tumor suppressors or oncogenes, but rather any marker, gene, or set of genes or markers that have been determined to have a statistically significant correlation, in terms of expression levels or alternative gene splicing, with a benign, malignant, or normal diagnosis is encompassed by the present invention.

[0016] In one embodiment, the method of the present invention is for improving the accuracy of current methods of cancer diagnosis.Improved accuracy can be due to the measurement of multiple genes and / or expression markers, the identification of gene expression products such as miRNA, rRNA, tRNA and mRNA gene expression products with high diagnostic power or statistical significance, or the identification of a group of genes and / or expression products with high diagnostic power or statistical significance, or any combination thereof.

[0017] For example, the increased expression of multiple receptor tyrosine kinases is associated with carcinogenesis. Measurement of gene expression product levels of specific receptor tyrosine kinases known to be differentially expressed in cancer cells may provide inaccurate diagnostic results resulting in low accuracy. Measurement of multiple receptor tyrosine kinases can increase accuracy by requiring the occurrence of alternative expressed gene combinations. Thus, in some cases, measurement of multiple genes can increase the accuracy of diagnosis by reducing the probability that a sample will show an abnormal gene expression profile at random probability.

[0018] Similarly, some gene expression products in a group, such as receptor tyrosine kinases, can indicate a disease or condition when their expression levels are higher or lower than normal. Measurement of the expression levels of other gene products in the same group may provide diagnostic utility. Thus, in one embodiment, the present invention measures two or more gene expression products in a group. For example, in some embodiments, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, or 50 gene expression products are measured from a group. Various groups are defined herein, such as groups useful for diagnosing thyroid cancer subtypes, or groups of gene expression products belonging to a particular ontology group. In another embodiment, it may be advantageous to measure the expression levels of a set of genes from multiple groups that pinpoint the presence or absence of cancer. For example, the present invention contemplates using 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, or 50 gene expression groups and measuring 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, or 50 gene expression products, respectively.

[0019] In addition, the increased expression of other cancer genes, such as Ras, in biological samples can also indicate the presence of cancerous cells.In some cases, it may be advantageous to measure the expression levels of several different classes of cancer genes, such as receptor tyrosine kinases, cytoplasmic tyrosine kinases, GTPases, serine / threonine kinases, lipid kinases, mitogens, growth factors, and transcription factors.Determining the expression levels and / or exon usage of different classes or groups of genes involved in cancer progression increases the diagnostic power of the present invention in some cases.

[0020] A group of expression markers may include markers within a metabolic or signal transduction pathway, or markers that are genetically or functionally homologous. For example, one group of markers may include genes involved in the epidermal growth factor signal transduction pathway. Another group of markers may include mitogen-activated protein kinases. The present invention also provides methods and compositions for detecting (i.e., measuring) and measuring gene expression markers from multiple and / or independent metabolic or signal transduction pathways.

[0021] In one embodiment, the expression product markers of the present invention can increase the accuracy of cancer diagnosis through the use of multiple expression product markers and statistical analysis.In particular, but not limited to, the present invention provides an RNA expression profile associated with thyroid cancer.The present invention also provides a method for characterizing thyroid tissue samples, as well as kits and compositions useful for the application of the method.The disclosure further includes a method for carrying out molecular profiling business.

[0022] The present disclosure provides methods and compositions to improve upon the current state of the art for diagnosing cancer.

[0023] In some embodiments, the present invention provides a method of diagnosing cancer comprising the steps of: obtaining a biological sample comprising gene expression products; measuring expression levels for one or more gene expression products of the biological sample; and identifying the biological sample as cancerous if the gene expression levels indicate the presence of thyroid cancer in the biological sample. This can be done by correlating the gene expression levels with the presence of thyroid cancer in the biological sample. In one embodiment, the gene expression products are selected from Figure 6. In some embodiments, the method further comprises comparing the expression levels of one or more gene expression products to a control expression level for each gene expression product in a control sample, where if there is a difference in gene expression levels between the gene expression products in the biological sample and the gene expression products in the control sample, the biological sample is identified as cancerous.

[0024] In some embodiments, the present invention provides a method for diagnosing cancer, comprising the steps of: obtaining a biological sample containing alternatively spliced ​​gene expression products; measuring expression levels for one or more gene expression products of the biological sample; and identifying the biological sample as cancerous if the gene expression levels indicate the presence of thyroid cancer in the biological sample. This can be done by correlating the gene expression levels with the presence of thyroid cancer in the biological sample. In one embodiment, the alternatively spliced ​​gene expression products are selected from FIG. 6, and differential gene expression product alternative exon usage is compared between the biological sample and the control sample; if there is a difference in gene expression product alternative exon usage between the biological sample and the control sample at a specified confidence level, the biological sample is identified as cancerous. In some embodiments, the genes selected from FIG. 6 are further selected from the genes listed in FIG. 2, FIG. 3, FIG. 4, or FIG. 5.

[0025] In some embodiments, the present invention provides a method of diagnosing cancer that provides a specificity or sensitivity of greater than 70% using the methods of the present invention described herein, in which gene expression product levels are compared between a biological sample and a control sample; if there is a difference in gene expression levels between the biological sample and the control sample at a specified confidence level, the biological sample is identified as cancerous. In some embodiments, the specificity and / or sensitivity of the method is at least 70%, 75%, 80%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or more.

[0026] In some embodiments, the nominal specificity is 70% or greater. The nominal negative predictive value (NPV) is 95% or greater. In some embodiments, the NPV is at least 95%, 95.5%, 96%, 96.5%, 97%, 97.5%, 98%, 98.5%, 99%, 99.5%, or greater.

[0027] Sensitivity typically refers to TP / (TP+FN), where TP is true positive and FN is false negative. The number of Continued Indeterminate results is divided by the total number of malignant results based on the adjudicated histopathology diagnosis. Specificity typically refers to TN / (TN+FP), where TN is true negative and FP is false positive. The number of benign results is divided by the total number of benign results based on the adjudicated histopathology diagnosis. Positive predictive value (PPV): TP / (TP+FP); Negative predictive value (NPV): TN / (TN+FN).

[0028] The marker panel is chosen to accommodate reasonable separation of benign from non-benign expression profiles. Training of this multidimensional classifier (i.e., algorithm) was performed on over 500 thyroid samples, including >300 thyroid FNAs. A number of training / test sets were used to develop the preliminary algorithm. An exemplary data set is shown in FIG. 22. First, the overall algorithm error rate is shown as a function of the number of genes for benign versus non-benign samples. All results are obtained using a support vector machine model trained and tested on samples in a cross-validated mode (30-fold).

[0029] In some embodiments, the difference in gene expression levels is at least 10%, 15%, 20%, 25%, 30%, 35%, 40%, 45%, or 50% or more. In some embodiments, the difference in gene expression levels is at least 2-fold, 3-fold, 4-fold, 5-fold, 6-fold, 7-fold, 8-fold, 9-fold, 10-fold or more. In some embodiments, the biological sample is identified as cancerous with greater than 75%, 80%, 85%, 90%, 95%, 99% or more accuracy. In some embodiments, the biological sample is identified as cancerous with greater than 95% sensitivity. In some embodiments, the biological sample is identified as cancerous with greater than 95% specificity. In some embodiments, the biological sample is identified as cancerous with greater than 95% sensitivity and greater than 95% specificity. In some embodiments, the accuracy is calculated using a trained algorithm.

[0030] In some embodiments, the present invention provides gene expression products corresponding to a gene selected from Table 3, Table 4, and / or Table 5.

[0031] In some embodiments, the present invention provides a method of diagnosing cancer, comprising using gene expression products from one or more of the following signaling pathways. Signaling pathways from which genes may be selected include, but are not limited to, acute myeloid leukemia signaling, type 2 somatostatin receptor signaling, cAMP-mediated signaling, cell cycle and DNA damage checkpoint signaling, G protein-coupled receptor signaling, integrin signaling, melanoma cell signaling, relaxin signaling, and thyroid cancer signaling. In some embodiments, multiple genes are selected from a single signaling pathway to measure and compare differential gene expression product levels between a biological sample and a control sample. Other signaling pathways include, but are not limited to, adhesion pathway, ECM pathway, thyroid cancer pathway, focal adhesion pathway, apoptosis pathway, p53 pathway, tight junction pathway, TGFβ pathway, ErbB pathway, Wnt pathway, pathway in cancer overview, cell cycle pathway, VEGF pathway, Jak / STAT pathway, MAPK pathway, PPAR pathway, mTOR pathway, or autoimmune thyroid pathway. In other embodiments, at least two genes are selected from at least two different signaling pathways to measure and compare the differential gene expression product levels between biological samples and control samples.The methods and compositions of the present invention can have genes selected from 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50 or more signaling pathways, and can have 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50 or more gene expression products from each signaling pathway in any combination.In some embodiments, the set of combined genes provides a specificity or sensitivity of greater than 70%, 75%, 80%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 99.5%, or a positive or negative predictive value of at least 95%, 95.5%, 96%, 96.5%, 97%, 97.5%, 98%, 98.5%, 99%, 99.5% or more.

[0032] In some embodiments, the present invention provides a method for diagnosing cancer, comprising a gene selected from at least two different ontology groups.In some embodiments, the ontology groups from which genes can be selected include, but are not limited to, cell senescence, cell cortex, cell cycle, cell death / apoptosis, cell differentiation, cell division, cell-cell junctions, cell migration, cell morphogenesis, cell movement, cell process, cell proliferation, cell recognition, cell body, cell surface, cell surface-bound receptor signaling, cell adhesion, transcription, immune response, or inflammation.In some embodiments, a plurality of genes are selected from a single ontology group to measure and compare the differential gene expression product levels between biological samples and control samples.In other embodiments, at least two genes are selected from at least two different ontology groups to measure and compare the differential gene expression product levels between biological samples and control samples. The methods and compositions of the invention can have genes selected from 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50 or more Gene Ontology Groups and can have 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50 or more gene expression products from each Gene Ontology Group in any combination. In some embodiments, the set of combined genes provides a specificity or sensitivity of 70%, 75%, 80%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 99.5%, or a positive or negative predictive value of at least 95%, 95.5%, 96%, 96.5%, 97%, 97.5%, 98%, 98.5%, 99%, 99.5% or more.

[0033] In some embodiments, the present invention provides a method for classifying cancer, comprising the steps of: obtaining a biological sample containing gene expression products; measuring the expression level of one or more gene expression products of the biological sample that are differentially expressed in different subtypes of cancer; and identifying the biological sample as cancerous if the gene expression level indicates the subtype of cancer. In some embodiments, the method further comprises comparing the expression level of the one or more gene expression products with a control expression level for each gene expression product in a control sample, and identifying the biological sample as cancerous if there is a difference in gene expression level between the gene expression products in the biological sample and the gene expression products in the control sample. In some embodiments, the method of the present invention distinguishes follicular carcinoma from medullary carcinoma. In some embodiments, the method of the present invention distinguishes benign thyroid disease from malignant thyroid tumor / cancer.

[0034] In some embodiments, the gene expression products of the methods of the invention are proteins and the amounts of proteins are compared. The amounts of proteins can be measured by one or more of the following: ELISA, mass spectrometry, blotting, or immunohistochemistry. RNA can be measured by one or more of the following: microarray, SAGE, blotting, RT-PCR, or quantitative PCR.

[0035] In some embodiments, the difference in gene expression levels, e.g., mRNA, protein, or alternatively spliced ​​gene products, between a biological sample and a control sample that can be used to diagnose cancer is at least 1.5-fold, 2-fold, 2.5-fold, 3-fold, 3.5-fold, 4-fold, 4.5-fold, 5-fold, 5.5-fold, 6-fold, 6.5-fold, 7-fold, 7.5-fold, 8-fold, 8.5-fold, 9-fold, 9.5-fold, 10-fold, or more.

[0036] In some embodiments, the biological sample is classified as cancerous or classified as positive for a cancer subtype with greater than 75%, 80%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 99.5% accuracy. Diagnostic accuracy as used herein includes specificity, sensitivity, positive predictive value, negative predictive value, and / or false detection rate.

[0037] When classifying biological samples for cancer diagnosis, there are typically four possible outcomes from a binary classifier. If the outcome from the prediction is p and the actual value is also p, it is called a true positive (TP). However, if the actual value is n, it is said to be a false positive (FP). Conversely, if both the predicted outcome and the actual value are n, a true negative occurs, and a false negative is when the predicted outcome is n but the actual value is p. In one embodiment, consider a diagnostic test to determine whether or not a person has a certain disease. A false positive in this case occurs when the person tests positive but actually does not have the disease. On the other hand, a false negative occurs when the person tests negative, suggesting that they are healthy, when in fact they do have the disease. In some embodiments, a ROC curve that assumes the real prevalence of the subtype can be generated by re-extracting the errors achieved in the samples available at the relevant rates.

[0038] The positive predictive value (PPV) of a disease, or accuracy, or post-test probability, is the proportion of patients with a positive test result who are correctly diagnosed. It is the most important measurement of a diagnostic method because it reflects the probability that a positive test reflects the underlying condition being tested for. However, its value depends on the prevalence of the disease, which can vary. In one example, FP (false positive); TN (true negative); TP (true positive); FN (false negative).

[0039] False positive rate (α)=FP / (FP+TN)-specificity

[0040] False negative rate (β)=FN / (TP+FN)--Sensitivity

[0041] Force = Sensitivity = 1-β

[0042] Positive likelihood ratio = sensitivity / (1-specificity)

[0043] Negative likelihood ratio = (1-sensitivity) / specificity

[0044] Negative predictive value is the proportion of patients with negative test results who are correctly diagnosed.Measurements of PPV and NPV can be derived using appropriate disease subtype prevalence estimates.Pooled malignant disease prevalence estimates can be calculated from indeterminate pools that are broadly classified as B or M by surgery.For subtype-specific estimates, in some embodiments, disease prevalence may not be calculable due to lack of available samples.In these cases, pooled disease prevalence estimates can be used instead of subtype disease prevalence.

[0045] In some embodiments, the level of expression product or alternative exon usage is indicative of one of the following: follicular cell carcinoma, undifferentiated carcinoma, medullary carcinoma, or sarcoma. In some embodiments, the one or more genes selected using the methods of the invention for diagnosing cancer contain representative sequences corresponding to a set of metabolic or signaling pathways indicative of cancer.

[0046] In some embodiments, the results of the expression analysis of the methods of the present invention provide a statistical confidence level that a given diagnosis is accurate, hi some embodiments, such a statistical confidence level is greater than 85%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 99.5%.

[0047] In another aspect, the present invention provides a composition for diagnosing cancer, comprising an oligonucleotide comprising a portion of one or more of the genes listed in Figure 6, or a complement thereof, and a substrate to which the oligonucleotide is covalently attached. The composition of the present invention is suitable for use in diagnosing cancer at a specified confidence level using a trained algorithm. In one example, the composition of the present invention is used to diagnose thyroid cancer.

[0048] In one aspect of the disclosure, samples processed by the cytology industry, subjected to routine methods and staining, diagnosed and categorized are then subjected to molecular profiling as a secondary diagnostic screen that allows for (1) a significant reduction in false positives and false negatives, (2) the determination of the genetic, metabolic or signaling pathways underlying the resulting pathology, (3) the ability to assign statistical probability to the accuracy of the diagnosis, (4) the ability to resolve ambiguous results, and (5) the ability to distinguish between cancer subtypes.

[0049] For example, in the particular case of thyroid cancer, the molecular profiling of the present invention can further provide a diagnosis of a particular type of thyroid cancer (e.g., papillary, follicular, medullary, or undifferentiated). The results of molecular profiling can further enable a person skilled in the art, such as a scientist or medical practitioner, to suggest or prescribe a particular therapeutic intervention. Molecular profiling of biological samples can also be used to monitor the efficacy of a particular treatment after initial diagnosis. It is further understood that in some cases, molecular profiling can be used in place of established cancer diagnostic methods, rather than in addition to established cancer diagnostic methods.

[0050] In one aspect, the present invention provides algorithms and methods that can be used for the diagnosis and monitoring of genetic disorders. Genetic disorders are diseases caused by abnormalities in genes or chromosomes. Some diseases, such as cancer, are in part caused by genetic disorders, but can also be caused by environmental factors. In some embodiments, the algorithms and methods disclosed herein are used for the diagnosis and monitoring of cancer, such as thyroid cancer.

[0051] Genetic disorders can typically be classified into two categories: single gene disorders and multifactorial polygenic (complex) disorders. Single gene disorders are the result of a single mutated gene. It is estimated that there are more than 4000 human diseases caused by defects in a single gene. Single gene disorders can be passed on to subsequent generations in several ways. There are several types of inheritance of single gene disorders, including but not limited to autosomal dominant inheritance, autosomal recessive inheritance, X-linked dominant inheritance, X-linked recessive inheritance, Y-linked inheritance, and mitochondrial inheritance. It is believed that only one mutated copy of a gene is necessary for a person to be affected by an autosomal dominant disorder. Examples of autosomal dominant disorders include but are not limited to Huntington's disease, neurofibromatosis type 1, Marfan syndrome, hereditary nonpolyposis colorectal cancer, and hereditary multiple exostoses. In an autosomal recessive disorder, two copies of a gene must be mutated for a person to be affected by an autosomal recessive disorder. Examples of this type of disorder include, but are not limited to, cystic fibrosis, sickle cell disease (also partial sickle cell disease), Tay-Sachs disease, Niemann-Pick disease, spinal muscular atrophy, and dry ear wax. X-linked dominant disorders are caused by mutations in genes on the X chromosome. Only a few disorders have this inheritance pattern, the primary example being X-linked hypophosphatemic rickets. Both sexes are affected in these disorders, but typically males are affected more severely than females. Some X-linked dominant conditions, such as Rett syndrome, incontinentia pigmenti type 2, and Aicardi syndrome, are usually lethal in males in utero or shortly after birth, and therefore are seen primarily in females. X-linked recessive disorders are also caused by mutations in genes on the X chromosome. Examples of this type of disorder include, but are not limited to, hemophilia A, Duchenne muscular dystrophy, red-green color blindness, muscular dystrophy, and male pattern baldness. Y-linked disorders are caused by mutations on the Y chromosome. Examples include, but are not limited to, male infertility and hypertrichosis pinnae.Mitochondrial inheritance, also known as maternal inheritance, refers to genes within the mitochondrial DNA.An example of this type of disorder is Leber's hereditary optic neuropathy.

[0052] Genetic disorders can also be complex, multifactorial, or polygenic. This means that they are likely to be related to the effects of multiple genes combined with lifestyle and environmental factors. Complex disorders often run in families, but do not have a clear inheritance pattern. Therefore, it is difficult to determine the risk that someone will inherit or transmit these disorders. Complex disorders are also difficult to study and treat, since the specific factors that cause most of these disorders have not yet been identified. Multifactorial or polygenic disorders that can be diagnosed, characterized, and / or monitored using the algorithms and methods of the present invention include, but are not limited to, heart disease, diabetes, asthma, autism, autoimmune diseases such as multiple sclerosis, cancer, cilia-related diseases, cleft palate, hypertension, inflammatory bowel disease, mental retardation, and obesity.

[0053] Other genetic disorders that may be diagnosed, characterized, and / or monitored using the algorithms and methods of the present invention include 1p36 deletion syndrome, 21-hydroxylase deficiency, 22q11.2 deletion syndrome, 47,XYY syndrome, 48,XXXX, 49,XXXXX, aceruloplasminemia, achondroplasia type II, achondroplasia, acute intermittent porphyria, adenylosuccinate lyase deficiency, adrenoleukodystrophy, ALA-deficiency porphyria, ALA deficiency syndrome, and 47,XYY syndrome. Dolatase deficiency, Alexander disease, alkaptonuria, alpha-1 antitrypsin deficiency, Alström syndrome, Alzheimer's disease (types 1, 2, 3, and 4), enamel hypoplasia, amyotrophic lateral sclerosis, amyotrophic lateral sclerosis type 2, amyotrophic lateral sclerosis type 4, androgen insensitivity syndrome, anemia, Angelman syndrome, Apert syndrome, ataxia-telangiectasia, Beare-Stevenson cutis gyrata syndrome, Benjamin syndrome, beta thalassemia, biotinidase deficiency, Birt-Hogg-Dube syndrome, bladder cancer, Bloom syndrome, bone disease, breast cancer, CADASIL, kyphotic dysplasia, Canavan disease, cancer, celiac disease, CGD chronic granulomatous disorder, Charcot-Marie-Tooth disease, Charcot-Marie-Tooth disease type 1, Charcot-Marie-Tooth disease type 4, Charcot-Marie-Tooth disease type 2, Charcot-Marie-Tooth disease type 4, Cockayne syndrome, Coffin- Lowry's syndrome, collagenopathy types II and XI, colorectal cancer, congenital absence of the vas deferens, congenital bilateral absence of the vas deferens, congenital diabetes mellitus, congenital erythropoietic porphyria, congenital heart disease, congenital hypothyroidism, connective tissue disease, Cowden's syndrome, cat cry, Crohn's disease, fibrostenosing, Crouzon syndrome, Crouzondermoskeletal syndrome, cystic fibrosis, De Grouchy's syndrome,Grouchy syndrome, neurodegenerative diseases, Dent's disease, developmental disorders, DiGeorge syndrome, distal spinal muscular atrophy type V, Down syndrome, short stature, Ehlers-Danlos syndrome, Ehlers-Danlos syndrome arthrolytic type, Ehlers-Danlos syndrome classic type, Ehlers-Danlos syndrome cutis laxa, Ehlers-Danlos syndrome kyphoscoliosis type, vascular type, erythropoietic protoporphyria, Fabry disease, facial injuries and facial disorders, factor V Leiden thrombophilia, familial aortic polyposis, familial dysautonomia, Fanconi anemia, FG syndrome, fragile X syndrome, Friedreich's ataxia, ataxia), G6PD deficiency, galactosemia, Gaucher disease (types 1, 2, and 3), hereditary brain disorders, glycine encephalopathy, hemochromatosis type 2, hemochromatosis type 4, harlequin ichthyosis, head and brain malformations, hearing impairment and hearing loss, childhood hearing impairment, hemochromatosis (neonatal, 2, and 3), hemophilia, myelohepatic porphyria, hereditary coproporphyria, hereditary multiple exostoses, hereditary neuropathy with liability to pressure palsies, hereditary nonpolyposis colorectal cancer, homocystinuria, Huntington's disease, Hutchinson-Gilford progeria syndrome, primary hyperoxaluria, hyperphenylalaninemia, hypochondrogenesis, hypochondroplasia, Idic15, incontinentia pigmenti, infantile Gaucher disease , infantile-onset ascending hereditary spastic paralysis, infertility, Jackson-Weiss syndrome, Joubert syndrome, juvenile primary lateral sclerosis, Kennedy disease, Klinefelter syndrome, Kniest dysplasia, Krabbe disease, learning disabilities, Lesch-Nyhan syndrome, leukodystrophy, Li-Fraumeni syndrome, familial lipoprotein lipase deficiency, male reproductive disorders, Marfan syndrome, McEuen-Albright syndrome, McLeod syndrome, familial Mediterranean fever, MEDNIK, Menkes disease, Menkes syndrome, metabolic disorders, beta-globin methemoglobinemia, congenital methemoglobinemia, methylmalonic acidemia, Micro syndrome, microcephaly, movement disorders, Mowat-Wilson syndrome, mucopolysaccharidoses (MPS)I), Muenke syndrome, muscular dystrophy, Duchenne and Becker muscular dystrophy, Duchenne and Becker muscular dystrophy, myotonic dystrophy, myotonic dystrophy types 1 and 2, neonatal hemochromatosis, neurofibromatosis, neurofibromatosis type 1, neurofibromatosis type 2, neurofibromatosis type I, neurofibromatosis type II, neurological disorders, neuromuscular disorders, Niemann-Pick disease, nonketotic hyperglycinemia, nonsyndromic hearing loss, autosomal recessive nonsyndromic hearing loss, Noonan syndrome, osteogenesis imperfecta (Types I and III), otospondylomegalyepiphyseal dysplasia, pantothenate kinase-associated neurodegeneration, Patau syndrome (trisomy 13), Pendred syndrome, Peutz-Jeghers syndrome, Pfeiffer syndrome, phenylketonuria, porphyria, porphyria cutanea tarda, Prader-Willi syndrome, primary pulmonary hypertension, prion diseases, progeria, propionic acidemia, protein C deficiency, protein S deficiency, pseudo-Gaucher disease, pseudoxanthoma elasticum, retinal disorders, retinoblastoma, FA-Friedreich syndrome Ataxia, Rett syndrome, Rubinstein-Taybi syndrome, SADDAN, Sandhoff disease, Sensory autonomic neuropathy type III, Sickle cell anemia, Skeletal muscle regeneration, Dyschromatosis, Smith-Lemli-Opitz syndrome, Speech and communication disorders, Spinal muscular atrophy, Spinobulbar muscular atrophy, Spinocerebellar degeneration, Strudwick type spondyloepiphyseal dysplasia, Congenital spondyloepiphyseal dysplasia, Stickler syndrome, Stickler syndrome COL2A1, Tay-Sachs disease, Tetrahydrobiopterin deficiency, Thallistic bone dysplasia, Diabetes and sensorineural deformity These include, but are not limited to, thiamine-responsive megaloblastic anemia with hearing loss, thyroid disease, Tourette's syndrome, Treacher Collins syndrome, triple X syndrome, tuberous sclerosis, Turner syndrome, Usher syndrome, variegate porphyria, von Hippel-Lindau disease, Waardenburg syndrome, Weissenbacher-Zweymuller syndrome, Wilson disease, Wolf-Hirschhorn syndrome, xeroderma pigmentosum, X-linked severe combined immunodeficiency, X-linked sideroblastic anemia, and X-linked spinal and bulbar muscular atrophy.

[0054] In one embodiment, the methods and algorithms of the present invention are used to diagnose, characterize, and monitor thyroid cancer. Other types of cancer that can be diagnosed, characterized, and / or monitored using the algorithms and methods of the present invention include adrenocortical carcinoma, anal cancer, aplastic anemia, bile duct cancer, bladder cancer, bone cancer, bone metastases, central nervous system (CNS) cancer, peripheral nervous system (PNS) cancer, breast cancer, Castleman's disease, cervical cancer, childhood non-Hodgkin's lymphoma, colorectal cancer, endometrial cancer, esophageal cancer, Ewing's family of tumors, and thyroid cancer. tumors) (e.g., Ewing's sarcoma), eye cancer, gallbladder cancer, gastrointestinal carcinoid tumors, gastrointestinal stromal tumors, gestational trophoblastic disease, hairy cell leukemia, Hodgkin's disease, Kaposi's sarcoma, kidney cancer, laryngeal and hypopharyngeal cancer, acute lymphocytic leukemia, acute myeloid leukemia, childhood leukemia, chronic lymphocytic leukemia, chronic myeloid leukemia, liver cancer, lung cancer, pulmonary carcinoid tumors, non-Hodgkin's lymphoma, male breast cancer, malignant mesothelioma, multiple myeloma, myelodysplastic syndromes These include, but are not limited to, adult cancers, myeloproliferative disorders, nasal and paranasal sinus cancer, nasopharyngeal cancer, neuroblastoma, oral and oropharyngeal cancer, osteosarcoma, ovarian cancer, pancreatic cancer, penile cancer, pituitary tumors, prostate cancer, retinoblastoma, rhabdomyosarcoma, salivary gland cancer, sarcoma (adult soft tissue cancer), melanoma skin cancer, non-melanoma skin cancer, gastric cancer, testicular cancer, thymic cancer, uterine cancer (e.g., uterine sarcoma), vaginal cancer, vulvar cancer, and Waldenstrom's macroglobulinemia.

[0055] In some embodiments, the gene expression product markers of the present invention can increase the accuracy of diagnosis of genetic disorders or cancer through the use of multiple gene expression product markers of low quantity and quality and statistical analysis using the algorithms of the present invention.In particular, but not limited to, the present invention provides a method for diagnosing, characterizing, and classifying gene expression profiles associated with thyroid cancer.The present invention also provides an algorithm for characterizing and classifying thyroid tissue samples, as well as kits and compositions useful for the application of the method.The present disclosure further includes a method of carrying out a molecular profiling business.

[0056] In one embodiment of the present invention, markers and genes can be identified as having differential expression in thyroid cancer samples compared to thyroid benign samples. Illustrative examples of benign pathologies include follicular adenoma, Hurthle cell adenoma, lymphocytic thyroiditis, and nodular hyperplasia. Illustrative examples of malignant pathologies include follicular carcinoma, follicular type papillary thyroid carcinoma, medullary carcinoma, and papillary thyroid carcinoma.

[0057] Biological samples can be processed to extract nucleic acids such as DNA or RNA. The nucleic acid can be contacted with an array of probes of the invention under conditions that allow hybridization. The degree of hybridization can be quantitatively assayed using a number of methods known in the art. In some cases, the degree of hybridization at a probe position can be related to the intensity of the signal provided by the assay, and thus the amount of complementary nucleic acid sequence present in the sample. Software can be used to extract, normalize, summarize, and analyze array intensity data from probes across the human genome or transcriptome, including expressed genes, exons, introns, and miRNAs. In some embodiments, the intensity of a given probe in either benign or malignant samples can be compared to a reference set to determine whether differential expression is occurring in the sample. An increase or decrease in relative intensity at the marker position on the array that corresponds to the expressed sequence indicates an increase or decrease in expression of the corresponding expressed sequence, respectively. Alternatively, a decrease in relative intensity can indicate a mutation in the expressed sequence.

[0058] The intensity values ​​generated for each sample can be analyzed using feature selection techniques, including filter techniques that assess feature relevance by looking at intrinsic properties of the data, wrapper methods that embed model hypotheses into the feature subset search, and embedded techniques, in which the search for the optimal feature set is built into a classification algorithm.

[0059] Filtering techniques useful in the methods of the invention include: (1) parametric methods such as two-sample t-tests, ANOVA analysis, Bayesian frameworks, and use of gamma distribution models; (2) model-free methods such as Wiscoxon rank sum tests, between-within class sum of squares tests, rank products methods, random permutation methods, or use of TNoM, which involves designing thresholds for the fold change in expression between two datasets and then finding the threshold in each gene that minimizes the number of misclassifications; and (3) multivariate methods such as bivariate methods, correlation based feature selection methods (CFS), minimum redundancy maximum relavance methods (MRMR), Markov blanket filters, and uncorrelated shrunken centroid methods. Wrapper methods useful in the methods of the present invention include sequential search methods, genetic algorithms, and estimation of distribution algorithms. Embedded methods useful in the methods of the present invention include random forest algorithms, weight vector of support vector machine algorithms, and weights of logistic regression algorithms. Bioinformatics. 2007 Oct 1; 23(19):2507-17 provides a summary of the relative merits of the above-listed filter techniques for the analysis of intensity data.

[0060] The selected features can then be classified using a classification algorithm. Exemplary algorithms include, but are not limited to, methods that reduce the number of variables, such as principal component analysis algorithms, partial least squares, and independent component analysis algorithms. Exemplary algorithms include, but are not limited to, methods that directly handle a large number of variables, such as methods based on statistical and machine learning techniques. Statistical methods include penalized logistic regression, prediction analysis of microarrays (PAM), methods based on shrunken centroids, support vector machine analysis, and regularized linear discriminant analysis. Machine learning techniques include bagging methods, boosting methods, random forest algorithms, and combinations thereof. Cancer Inform. 2008; 6: 77-97 provides an overview of the above classification techniques for the analysis of microarray intensity data.

[0061] The markers and genes of the present invention may be used to characterize the cancerous or non-cancerous state of cells or tissues. The present invention includes a method of diagnosing benign tissue or cells from malignant tissue or cells, comprising measuring the differential expression of a marker or gene listed in Figures 2-6, 9-13, 16, or 17 in a thyroid sample from a subject. The present invention also includes a method of diagnosing medullary thyroid cancer, comprising measuring the differential expression of a marker or gene listed in Figure 10 in a thyroid sample from a subject. The present invention also includes a method of diagnosing a subtype of thyroid pathology, comprising measuring the differential expression of a marker or gene listed in Figure 13 in a thyroid sample from a subject. The present invention also includes a method of diagnosing benign tissue or cells from malignant tissue or cells, comprising measuring the differential expression of a miRNA, which is a miRNA listed in Figures 16 or 17, in a thyroid sample from a subject.

[0062] In accordance with the above, differential expression of genes, markers, miRNAs, or combinations thereof as disclosed herein can be measured using Northern blotting and sequences as identified herein to develop probes for this purpose. Such probes can be composed of DNA or RNA or synthetic nucleotides or combinations thereof, and advantageously can be composed of a continuous stretch of nucleotide residues that match or are complementary to sequences as identified in Figures 2-6, 9-13, 16, or 17. Such probes will most usefully be those derived from one or more of the sequences identified in Figures 2-6, 9-13, 16, or 17, such as 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, It is contemplated that the sequence may comprise a contiguous stretch of at least 15 to 200 residues or more, including 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 80, 85, 90, 95, 100, 110, 120, 130, 140, 150, 160, 175, or 200 nucleotides or more. Thus, if a single probe binds multiple times to the transcriptome of a sample of cells that are cancerous, suspected to be cancerous, or predisposed to becoming cancerous, and binding of the same probe to a similar amount of transcriptome derived from the genome of non-cancerous cells of the same organ or tissue results in observably more or less binding, this indicates differential expression of a gene, genes, markers, or miRNA that includes or corresponds to a sequence identified in Figures 2-6, 9-13, 16, or 17 from which the sequenced probe is derived.

[0063] In one such embodiment, elevated expression compared to normal cells and / or tissues of the same organ is measured by measuring the relative rate of transcription of RNA, such as by making a corresponding cDNA, and then analyzing the resulting DNA using probes developed from gene sequences such as those identified in Figures 2-6, 9-13, 16, or 17. Thus, the level of cDNA generated by using reverse transcriptase with the complete RNA complement of a cell suspected to be cancerous can be amplified using polymerase chain reaction, or some other means such as linear amplification, isothermal amplification, NASB, or rolling circle amplification, to generate a corresponding amount of cDNA, which can then be used to measure the relative level of the resulting cDNA and thereby the relative level of gene expression.

[0064] Increased expression may be determined using agents that selectively bind to, and thereby detect the presence of, the expression products of the genes disclosed herein. For example, an antibody, perhaps an appropriately labeled antibody, such as an antibody conjugated to a fluorescent or radioactive label, can be generated against one of the polypeptides that include the sequences identified in Figures 2-6 and 9-13, which would then selectively or specifically bind and react with the polypeptides encoded by one of the genes corresponding to the sequences disclosed herein. The binding of such antibodies, particularly the relative extent of such binding in samples derived from cells and tissues suspected to be cancerous as opposed to other non-cancerous cells and tissues, can then be used as a measure of the degree of expression or overexpression of the cancer-associated genes identified herein. Thus, the genes identified herein as being overexpressed in cancerous cells and tissues may be overexpressed by increased copy number or by hypertranscription. For example, overexpression may result from overproduction of a transcription factor that activates the gene and results in repeated binding of RNA polymerase, thereby generating greater than normal amounts of RNA transcripts that are subsequently translated into polypeptides, such as those containing amino acid sequences as identified in Figures 2-6 and 9-13. Such analyses confirm expression of the genes identified by the present invention, thereby providing an additional means of determining the presence of a cancerous condition in a sample derived from a tested patient, and the presence of a predisposition in said patient to subsequently develop cancer.

[0065] In using the methods of the invention, it should be kept in mind that the expression of genes or markers indicative of a cancerous condition need not be characteristic of all cells found to be cancerous. Thus, the methods disclosed herein are useful for detecting the presence of a cancerous condition in tissues in which some cells show a perfect pattern of overexpression. For example, a set of selected genes or markers comprising sequences that are homologous under stringent conditions or at least 90%, preferably 95%, identical to at least one of the sequences as identified in Figures 2-6, 9-13, 16, or 17, can be found, using either a suitable DNA or RNA probe, to be present in as few as 60% of cells derived from a tumor sample or malignant tissue and absent in as many as 60% of cells derived from a corresponding non-cancerous or normal tissue (and thus present in as many as 40% of such normal tissue cells). In one embodiment, such expression patterns are found to be present in at least 70% of cells taken from the cancerous tissue and absent in at least 70% of the corresponding normal non-cancerous tissue sample. In another embodiment, such expression patterns are found to be present in at least 80% of cells taken from cancerous tissue and absent in at least 80% of corresponding normal non-cancerous tissue samples. In another embodiment, such expression patterns are found to be present in at least 90% of cells taken from cancerous tissue and absent in at least 90% of corresponding normal non-cancerous tissue samples. In another embodiment, such expression patterns are found to be present in at least 100% of cells taken from cancerous tissue and absent in at least 100% of corresponding normal non-cancerous tissue samples, although the latter embodiment may occur rarely.

[0066] In some embodiments, molecular profiling includes detection, analysis, or quantification of nucleic acid (DNA or RNA), protein, or a combination thereof. Diseases or conditions diagnosed by the methods of the present invention include, for example, abnormal proliferation conditions in one or more tissues of a subject, including, but not limited to, skin, heart, lung, kidney, breast, pancreas, liver, muscle, smooth muscle, bladder, gallbladder, colon, intestine, brain, esophagus, or prostate. In some embodiments, tissues analyzed by the methods of the present invention include thyroid tissue.

[0067] In some embodiments, the disease or condition diagnosed by the method of the present invention includes benign and malignant hyperproliferative disorders, including, but not limited to, cancer, hyperplasia, or neoplasm. In some cases, the hyperproliferative disorder diagnosed by the method of the present invention includes breast cancer, such as ductal carcinoma, medullary carcinoma, colloid carcinoma, tubular carcinoma, and inflammatory breast cancer in the ductal tissue of the breast; ovarian cancer, including ovarian epithelial tumors, such as adenocarcinoma in the ovary, and adenocarcinoma that has migrated from the ovary to the peritoneal cavity; uterine cancer; cervical cancer, such as adenocarcinoma in the cervical epithelium, including squamous cell carcinoma and adenocarcinoma; prostate cancer, such as prostate cancer selected from adenocarcinoma or adenocarcinoma that has migrated to bone; pancreatic cancer, such as epithelioid cell carcinoma in the pancreatic duct tissue and adenocarcinoma in the pancreatic duct; bladder cancer, such as ... bladder cancer, such as transitional cell carcinoma in the bladder, urothelial carcinoma (transitional cell carcinoma), tumors in the urothelial cells lining the bladder, squamous cell carcinoma, adenocarcinoma, and small cell carcinoma; leukemia, such as acute myeloid leukemia (AML), acute lymphocytic leukemia, chronic lymphocytic leukemia, chronic myelogenous leukemia, hairy cell leukemia, myelodysplasia, myeloproliferative disorders, acute myeloid leukemia (AML), chronic myelogenous leukemia (CML), mastocytosis, chronic lymphocytic leukemia (CLL), multiple myeloma (MM), and myelodysplastic syndromes (MDS); bone cancer; squamous cell Lung cancers such as non-small cell lung cancer (NSCLC), which is divided into carcinoma, adenocarcinoma, and large cell anaplastic carcinoma, and small cell lung cancer; skin cancers such as basal cell carcinoma, melanoma, squamous cell carcinoma, and actinic keratosis, a skin condition that can progress to squamous cell carcinoma; ocular retinoblastoma; cutaneous melanoma or intraocular (eye) melanoma; primary liver cancer (cancer that begins in the liver); kidney cancer; AIDS-related lymphomas such as diffuse large B-cell lymphoma, B-cell immunoblastic lymphoma, and small noncleaved cell lymphoma; Kaposi's sarcoma; and hepatitis B. virus-induced cancers, including hepatitis B virus (HBV), hepatitis C virus (HCV), and hepatocellular carcinoma; human lymphotropic virus type 1 (HTLV-1) and adult T-cell leukemia / lymphoma; and virus-induced cancers, including human papillomavirus (HPV)-induced cervical cancer; central nervous system (CNS) cancers, such as primary brain tumors, including gliomas (astrocytoma, anaplastic astrocytoma, or glioblastoma multiforme), oligodendroglioma, ependymoma, meningioma, lymphoma, schwannoma, and medulloblastoma;peripheral nervous system (PNS) cancers such as acoustic neuroma, and malignant peripheral nerve sheath tumors (MPNSTs), including neurofibromas and schwannomas, malignant fibrous cell tumors, malignant fibrous histiocytoma, malignant meningioma, malignant mesothelioma, and malignant mixed Mullerian tumor; oral cavity and oropharynx cancers, such as hypopharyngeal, laryngeal, nasopharyngeal, and oropharyngeal cancer; gastric cancers, such as lymphoma, gastric stromal tumor, and carcinoid tumor; testicular cancers, such as germ cell tumors (GCTs), including seminomas and non-seminomas, and gonadal stromal tumors, including Leydig cell tumors and Sertoli cell tumors; thymic carcinomas, such as thymoma, thymic carcinoma, Hodgkin's disease, non-Hodgkin's lymphoma carcinoid, or carcinoid tumor; rectal cancer; and colon cancer. In some cases, the diseases or conditions diagnosed by the methods of the invention include, but are not limited to, thyroid disorders such as, for example, follicular adenoma, Hurthle cell adenoma, lymphocytic thyroiditis, and thyroid hyperplasia. In some cases, the diseases or conditions diagnosed by the methods of the invention include, but are not limited to, malignant thyroid disorders such as, for example, follicular carcinoma, follicular papillary thyroid carcinoma, medullary carcinoma, and papillary carcinoma. In some cases, the methods of the invention provide a diagnosis of tissue as diseased or normal. In other cases, the methods of the invention provide a diagnosis of normal, benign, or malignant. In some cases, the methods of the invention provide a diagnosis of benign / normal or malignant. In some cases, the methods of the invention provide a diagnosis of one or more of the specific diseases or conditions provided herein;

[0068] II. Obtaining Biological Samples In some embodiments, the method of the present invention provides for obtaining a sample from a subject. As used herein, the term subject refers to any animal (e.g., mammal), including, but not limited to, humans, non-human primates, rodents, dogs, pigs, etc. Methods that may be provided herein include biopsy methods, including fine needle aspiration, core needle biopsy, aspiration biopsy, incisional biopsy, excision biopsy, punch biopsy, shave biopsy, or skin biopsy. Samples may be obtained from any of the tissues provided herein, including, but not limited to, skin, heart, lung, kidney, breast, pancreas, liver, muscle, smooth muscle, bladder, gallbladder, colon, intestine, brain, prostate, esophagus, or thyroid. Alternatively, samples may be obtained from other sources, including, but not limited to, blood, sweat, hair follicles, cheek tissue, tears, menses, stool, or saliva. In some embodiments of the present invention, a medical professional may obtain a biological sample for testing. In some cases, a medical professional may refer a subject to a testing facility or laboratory for depositing a biological sample. In other cases, the subject may provide the sample. In some cases, the molecular profiling business of the invention may obtain the sample.

[0069] The sample may be obtained by a method known in the art, such as a biopsy method provided herein, a swab, scrape, venisection, or any other method known in the art. In some cases, the sample may be obtained, stored, or transported using the components of the kit of the present invention. In some cases, multiple samples, such as multiple thyroid samples, may be obtained for diagnosis by the method of the present invention. In some cases, multiple samples, such as one or more samples from one tissue type (e.g., thyroid) and one or more samples from another tissue (e.g., cheek), may be obtained for diagnosis by the method of the present invention. In some cases, multiple samples, such as one or more samples from one tissue type (e.g., thyroid) and one or more samples from another tissue (e.g., cheek), may be obtained at the same time point or at different times. In some cases, the samples obtained at different times may be analyzed and / or stored by different methods. For example, a sample may be obtained and analyzed by cytological analysis (routine staining). In some cases, additional samples may be obtained from the subject based on the results of the cytological analysis. Diagnosing cancer may include an examination of a subject by a doctor, nurse, or other medical professional. The examination may be part of a routine examination, or the examination may be due to a specific complaint, including, but not limited to, one of the following: pain, illness, anticipation of illness, presence of a suspicious lump or mass, disease, or condition. The subject may or may not be aware of the disease or condition. The medical professional may obtain a biological sample for testing. In some cases, the medical professional may refer the subject to a testing facility or laboratory for deposit of a biological sample.

[0070] In some cases, the subject may be referred to a specialist, such as an oncologist, surgeon, or endocrinologist, for further diagnosis. The specialist may also obtain a biological sample for testing, or refer the individual to a testing facility or laboratory for depositing the biological sample. In either case, the biological sample may be obtained by a doctor, nurse, or other medical professional, such as a medical technician, endocrinologist, cytologist, phlebotomist, radiologist, or pulmonologist. The medical professional may indicate the appropriate test or assay to perform on the sample, or the molecular profiling business of the present disclosure may advise on which assay or test is most appropriately indicated. The molecular profiling business may charge the individual or the medical provider or insurance provider for the advisory business, sample acquisition and or storage, materials, or all products and services provided.

[0071] In some embodiments of the present invention, a medical professional may not be involved in the initial diagnosis or sample acquisition. Alternatively, an individual may obtain a sample through the use of an over-the-counter kit. The kit may contain a means for obtaining a sample as described herein, a means for storing the sample for testing, and instructions for proper use of the kit. In some cases, molecular profiling services may be included in the purchase price of the kit. In other cases, molecular profiling services are an additional charge.

[0072] The sample suitable for use by the molecular profiling business can be any material that contains the tissue, cell, nucleic acid, gene, gene fragment, expression product, gene expression product, or gene expression product fragment of the individual to be tested. Methods for determining the suitability and / or appropriateness of the sample are provided. Samples can include, but are not limited to, the tissue, cell, or biological material derived from or obtained from the cell of an individual. Samples can be a heterogeneous or homogeneous population of cells or tissues. Biological samples can be obtained using any method known in the art that can provide a sample suitable for the analysis methods described herein.

[0073] Samples may be obtained by non-invasive methods, including, but not limited to, skin or neck scraping, cheek swabs, saliva collection, urine collection, stool collection, menstrual, tear, or semen collection. In other cases, samples are obtained by invasive procedures, including, but not limited to, biopsy, alveolar or pulmonary lavage, needle aspiration, or venisection. Biopsy methods may further include incisional biopsy, excision biopsy, punch biopsy, shave biopsy, or skin biopsy. Needle aspiration methods may further include fine needle aspiration, core needle biopsy, aspiration biopsy, or large core biopsy. In some embodiments, multiple samples may be obtained by the methods herein to ensure a sufficient amount of biological material. Methods for obtaining suitable thyroid samples are known in the art and are further described in the ATA Guidelines for thryoid nodule management (Cooper et al. Thyroid Vol. 16 No. 2 2006), which is incorporated herein by reference in its entirety. General methods for obtaining biological samples are also known in the art and are further described, for example, in Ramzy, Ibrahim Clinical Cytopathology and Aspiration Biopsy 2001, which is incorporated herein by reference in its entirety. In one embodiment, the sample is a fine needle aspirate of a thyroid nodule or suspected thyroid tumor. In some cases, the fine needle aspirate may be obtained under guidance using ultrasound, x-ray, or other imaging devices.

[0074] In some embodiments of the invention, the molecular profiling business may obtain the biological sample directly from the subject, from a medical professional, from a third party, or from a kit provided by the molecular profiling business or a third party. In some cases, the biological sample may be obtained by the molecular profiling business after the subject, a medical professional, or a third party obtains the biological sample and sends it to the molecular profiling business. In some cases, the molecular profiling business may provide appropriate containers and excipients for storage and transport of the biological sample to the molecular profiling business.

[0075] III. Sample Storage In some embodiments, the methods of the invention provide for storing the sample for a period of time, such as seconds, minutes, hours, days, weeks, months, years, or more, after the sample is obtained and before the sample is analyzed by one or more methods of the invention. In some cases, the sample obtained from the subject is subdivided prior to storage or further analysis steps, such that different portions of the sample are subjected to different downstream methods or processes, including, but not limited to, storage, cytological analysis, validity testing, nucleic acid extraction, molecular profiling, or combinations thereof.

[0076] In some cases, a portion of the sample may be stored and another portion of the sample may be further manipulated. Such manipulations may include, but are not limited to, molecular profiling; cytological staining; extraction, detection, or quantification of nucleic acids (RNA or DNA); extraction, detection, or quantification of gene expression products (RNA or protein); fixation; and testing. The sample may be fixed before or during storage by any method known in the art, such as using glutaraldehyde, formaldehyde, or methanol. In other cases, the sample is obtained and stored, and after the storage step, different portions of the sample are subdivided for further analysis, such that they undergo different downstream methods or processes, including, but not limited to, storage, cytological analysis, validity testing, nucleic acid extraction, molecular profiling, or combinations thereof. In some cases, the sample is obtained and analyzed, for example, by cytological analysis, and the resulting sample material is further analyzed by one or more molecular profiling methods of the present invention. In such cases, the sample may be stored between the cytological analysis step and the molecular profiling step. After acquisition, the sample may be stored to facilitate transport or to await the results of other analyses, hi another embodiment, the sample may be stored while awaiting clarification from a physician or other medical professional.

[0077] The obtained sample may be placed in a suitable medium, excipient, solution, or container for short-term or long-term storage. Storage may require maintaining the sample in a refrigerated or frozen environment. The sample may be rapidly frozen before storage in a frozen environment. The frozen sample may be contacted with a suitable cryopreservation medium or compound, including but not limited to glycerol, ethylene glycol, sucrose, or glucose. Suitable medium, excipient, or solution includes but is not limited to Hank's salt solution, saline, cell growth medium, ammonium salt solution such as ammonium sulfate or ammonium phosphate, or water. Suitable concentrations of ammonium salts include solutions of about 0.1 g / ml, 0.2 g / ml, 0.3 g / ml, 0.4 g / ml, 0.5 g / ml, 0.6 g / ml, 0.7 g / ml, 0.8 g / ml, 0.9 g / ml, 1.0 g / ml, 1.1 g / ml, 1.2 g / ml, 1.3 g / ml, 1.4 g / ml, 1.5 g / ml, 1.6 g / ml, 1.7 g / ml, 1.8 g / ml, 1.9 g / ml, 2.0 g / ml, 2.2 g / ml, 2.3 g / ml, 2.5 g / ml, or more. The media, excipients, or solutions may or may not be sterile.

[0078] Samples may be stored at room temperature, or at low temperatures such as cold (e.g., about 20° C. to about 0° C.) or frozen temperatures, including, for example, 0° C., -1° C., -2° C., -3° C., -4° C., -5° C., -6° C., -7° C., -8° C., -9° C., -10° C., -12° C., -14° C., -15° C., -16° C., -20° C., -22° C., -25° C., -28° C., -30° C., -35° C., -40° C., -45° C., -50° C., -60° C., -70° C., -80° C., -100° C., -120° C., -140° C., -180° C., -190° C., or about -200° C. In some cases, samples may be stored in a refrigerator, on ice or frozen gel packs, in a freezer, in a cryogenic storage device, on dry ice, in liquid nitrogen, or in the vapor phase equilibrated with liquid nitrogen.

[0079] The medium, excipient, or solution may contain a preservative to maintain the sample in a proper condition for subsequent diagnosis or manipulation or to prevent coagulation. Preservatives may include citric acid, ethylenediaminetetraacetic acid, sodium azide, or thimerosal. The medium, excipient, or solution may contain appropriate buffers or salts, such as Tris or phosphate buffers, sodium salts (e.g., NaCl), calcium salts, magnesium salts, and the like. In some cases, the sample may be stored in a commercial preparation suitable for storage of cells for subsequent cytological analysis, such as, but not limited to, Cytyc ThinPrep, SurePath, or Monoprep.

[0080] The sample container may be any container suitable for storing and or transporting a biological sample, including, but not limited to, a cup, a cup with a lid, a tube, a sterile tube, a vacuum tube, a syringe, a bottle, a microscope slide, or other suitable container. The container may or may not be sterile.

[0081] IV. Sample transportation The methods of the present invention provide for the transportation of a sample. In some cases, the sample is transported from a clinic, hospital, doctor's office, or other location to a second location where the sample can be stored and / or analyzed, for example, by cytological analysis or molecular profiling. In some cases, the sample may be transported to a molecular profiling business to perform the analysis described herein. In other cases, the sample may be transported to a laboratory, such as a licensed laboratory, such as a Clinical Laboratory Improvement Amendments (CLIA) laboratory, or a laboratory capable of performing the methods of the present invention. The sample may be transported by the individual from whom the sample originates. Transportation by the individual may include the individual visiting a molecular profiling business or a designated sample receiving location and providing the sample. Providing the sample may include any of the techniques of sample acquisition described herein, or the sample may already be obtained and stored in a suitable container as described herein. In other cases, the sample may be transported to the molecular profiling business using courier, mail, shipping, or any method capable of transporting the sample in a suitable manner. In some cases, the sample may be provided to the molecular profiling business by a third party testing laboratory (e.g., a cytology laboratory). In other cases, the sample may be provided to the molecular profiling business by the subject's primary care physician, endocrinologist, or other medical professional. Shipping costs may be billed to the individual, medical provider, or insurance provider. Upon receipt, the molecular profiling business may begin analyzing the sample immediately or may store the sample in any manner described herein. The method of storage may or may not be the same as that chosen prior to the molecular profiling business receiving the sample.

[0082] The sample may be shipped in any medium or vehicle, including any medium or vehicle provided herein that is suitable for storage of the sample, such as cryopreservation medium or liquid based cytology preparation. In some cases, the sample may be shipped frozen or refrigerated at any of the suitable sample storage temperatures provided herein.

[0083] After receipt of the sample by the molecular profiling business, its agent or licensee, a medical professional, a researcher, or a third party laboratory or testing facility (e.g., a cytology laboratory), the sample may be assayed using a variety of routine analyses known in the art, such as cytological assays and genomic analysis. Such tests may indicate the presence of cancer, a type of cancer, other disease or condition, disease markers, or the absence of cancer, disease, condition, or disease markers. Testing may take the form of cytological investigations, including microscopic examination, as described below. Testing may include the use of one or more cytological stains. The biological material may be manipulated or prepared for testing prior to administration of the test by any suitable method known in the art for the preparation of biological samples. The particular assay to be performed may be determined by the molecular profiling company, the physician ordering the test, or a consulting medical professional, a cytology laboratory, the subject from whom the sample is derived, or a third party, such as an insurance provider. A particular assay may be selected based on the likelihood of obtaining a definitive diagnosis, the cost of the assay, the speed of the assay, or the suitability of the assay to the type of material provided.

[0084] V. Testing for Validity After or during sample acquisition, for example, before or after the step of storing the sample, biological material can be collected and assessed for suitability, for example, to assess the suitability of the sample for use in the methods and compositions of the present invention. Assessment can be performed by the individual who obtains the sample, a molecular profiling business, an individual who uses the kit, or a third party such as a cytology laboratory, a pathologist, an endocrinologist, or a researcher. A sample can be determined to be suitable or unsuitable for further analysis due to many factors, including but not limited to: insufficient cells, insufficient genetic material, insufficient protein, DNA, or RNA, inappropriate cells for the required test, or inappropriate material for the required test, the age of the sample, the manner in which the sample was obtained, or the manner in which the sample was stored or transported. Suitability can be determined using a variety of methods known in the art, such as cell staining methods, measuring cell number or tissue volume, measuring total protein, measuring nucleic acid, visual inspection, microscopic inspection, or measuring temperature or pH. In one embodiment, sample suitability will be determined from the results of performing gene expression product level analysis experiments. In another embodiment, sample adequacy will be determined by measuring the content of markers of sample adequacy, including elements such as iodine, calcium, magnesium, phosphorus, carbon, nitrogen, sulfur, iron, etc.; proteins, such as, but not limited to, thyroglobulin; cell mass; and cellular components such as proteins, nucleic acids, lipids, or carbohydrates.

[0085] In some cases, iodine can be measured by chemical methods such as those described in U.S. Pat. No. 3,645,691, the entirety of which is incorporated herein by reference, or other chemical methods known in the art for measuring iodine content. Chemical methods for iodine measurement include, but are not limited to, methods based on the Sandell and Kolthoff reaction. The reaction proceeds according to the following equation: 2 Ce 4+ +As 3+ →2 Ce 3+ +As 5+ +I

[0086] Iodine has a catalytic effect in the course of the reaction, i.e. the more iodine present in the preparation being analyzed, the faster the reaction proceeds. The speed of the reaction is proportional to the iodine concentration. In some cases, the analysis can be carried out in the following manner:

[0087] A pre-measured amount of arsenic oxide As2O3 in concentrated sulfuric acid or concentrated nitric acid is added to the biological sample, and the temperature of the mixture is adjusted to the reaction temperature, i.e., typically 20°C to 60°C. A pre-measured amount of cerium (IV) sulfate in sulfuric acid or nitric acid is added thereto. The mixture is then reacted at the pre-measured temperature for a certain time. The reaction time is selected according to the order of magnitude of the amount of iodine to be measured and the respective reaction temperature selected. The reaction time is typically about 1 minute to about 40 minutes. The content of cerium (IV) ions in the test solution is then measured photometrically. The lower the concentration of cerium (IV) ions measured photometrically, the higher the reaction speed and therefore the amount of catalytic agent, i.e. iodine. In this way, the iodine in the sample can be measured directly and quantitatively.

[0088] In other cases, for example, 123 I, 124 I, 125 I, and 131The iodine content of a thyroid tissue sample can also be measured by detecting a specific isotope of iodine, such as I. In still other cases, the marker can be another radioisotope, such as an isotope of carbon, nitrogen, sulfur, oxygen, iron, phosphorus, or hydrogen. In some instances, the radioisotope can be administered prior to sample collection. Methods of administering a radioisotope suitable for validity testing are well known in the art and include injection into a vein or artery, or ingestion. A suitable period between administration of the isotope and taking of a thyroid nodule sample to allow a portion of the isotope to be absorbed into the thyroid tissue can include a period of about 1 minute to several days or about 1 week, including about 1 minute, 2 minutes, 5 minutes, 10 minutes, 15 minutes, 30 minutes, 1 hour, 8 hours, 12 hours, 24 hours, 48 ​​hours, 72 hours, or about 1 week, 1.5 weeks, or 2 weeks, and can be easily determined by one of ordinary skill in the art. Alternatively, the sample may be measured for natural levels of isotopes, such as radioactive isotopes of iodine, calcium, magnesium, carbon, nitrogen, sulfur, oxygen, iron, phosphorus, or hydrogen.

[0089] (i) Content validity testing of cells and / or tissues Methods for measuring the amount of tissue include, but are not limited to, weighing the sample or measuring the volume of the sample. Methods for measuring the amount of cells include, but are not limited to, cell counting, which in some cases may be performed after disaggregation, for example, by enzymes such as trypsin or collagenase, or by physical means, for example, by using a tissue homogenizer. Other methods for measuring the amount of cells recovered include, but are not limited to, quantifying dyes that bind to cellular material, or measuring the volume of the cell pellet obtained after centrifugation. Methods for determining that a reasonable number of specific types of cells are present include PCR, Q-PCR, RT-PCR, immunohistochemical analysis, cytological analysis, microscopic analysis, and or visual analysis.

[0090] (ii) Nucleic acid content validity test A variety of methods known in the art can be used to analyze samples by measuring nucleic acid content after extraction from biological samples. In some cases, nucleic acids such as RNA or mRNA are extracted from other nucleic acids before nucleic acid content analysis. Nucleic acid content can be extracted, purified, and measured by ultraviolet absorbance, including but not limited to absorbance at 260 nanometers, using a spectrophotometer. In other cases, nucleic acid content or integrity can be measured by fluorometer after contacting the sample with a stain. In still other cases, nucleic acid content or integrity can be measured after electrophoresis or using a device such as, for example, an Agilent bioanalyzer. It is understood that the method of the present invention is not limited to a particular method for measuring nucleic acid content and / or integrity.

[0091] In some embodiments, the amount or yield of RNA from a given sample is measured immediately after purification using a NanoDrop spectrophotometer in the nanogram to microgram range. In some embodiments, RNA quality is measured using an Agilent 2100 Bioanalyzer instrument and characterized by a calculated RNA Integrity Number (RIN, 1-10). NanoDrop is a cuvette-less spectrophotometer. It uses 1 microliter to measure 5ng-3,000ng per μl of sample. Key features of NanoDrop include low sample volume, no cuvette; large dynamic range of 5ng / μl-3,000ng / μl; and allows quantification of DNA, RNA, and protein. NanoDrop™ 2000c allows analysis of samples of 0.5μl-2.0μl without the need for cuvettes or capillaries.

[0092] RNA quality can be measured by a calculated RNA Integrity Number (RIN). The RNA Integrity Number (RIN) is an algorithm for assigning integrity values ​​to RNA measurements. RNA integrity is a major issue for gene expression studies and is traditionally assessed using the ratio of 28S rRNA to 18S rRNA, but this method has been shown to be inconsistent. The RIN algorithm is applied to RNA measurements by electrophoresis and is based on a combination of different features that contribute information about RNA integrity to provide a more robust universal measurement. In some embodiments, RNA quality is measured using an Agilent 2100 Bioanalyzer instrument. Protocols for measuring RNA quality are known and commercially available, for example, on the Agilent website. Briefly, in a first step, a researcher deposits a total RNA sample into an RNA Nano LabChip. In a second step, the LabChip is inserted into the Agilent Bioanalyzer, the analysis is performed, and a digital electropherogram is generated. Then, in a third step, the new RIN algorithm analyzes the entire electrophoretic record of the RNA sample, including the presence or absence of degradation products, to determine the integrity of the sample. The algorithm then assigns a RIN score of 1 to 10, with level 10 being perfectly intact RNA. The interpretation of the electropherogram is automatic and not subject to individual interpretation, allowing for universal and unbiased comparison of samples and improving experimental reproducibility. The RIN algorithm was developed using neural networks and adaptive learning, together with a large database of eukaryotic total RNA samples, primarily from human, rat, and mouse tissues. Advantages of RIN include:A numerical assessment of RNA integrity is obtained; for example, RNA samples before and after archival can be directly compared, and the integrity of the same tissue between different laboratories can be compared; experimental reproducibility is ensured, e.g., if the RIN is given and suitable for a microarray experiment, the same RIN value can always be used for similar experiments if the same organism / tissue / extraction method is used (Schroeder A, et al. BMC Molecular Biology 2006, 7:3 (2006)).

[0093] In some embodiments, RNA quality is measured on a scale of RIN 1-10, with 10 being the highest quality. In one aspect, the present invention provides a method for analyzing gene expression from samples with RNA RIN values ​​of 6.0 or less. In some embodiments, samples containing RNA with RIN numbers of 1.0, 2.0, 3.0, 4.0, 5.0, or 6.0 are analyzed for microarray gene expression using the methods and algorithms of the present invention. In some embodiments, the sample is a fine needle aspirate of thyroid tissue. The sample may be degraded with a RIN as low as 2.0.

[0094] Measuring gene expression in a given sample is a complex, dynamic, and costly process. RNA samples with RIN≦5.0 are typically not used for multigene microarray analysis, but instead can only be used for single-gene RT-PCR and / or TaqMan assays. This dichotomy of RNA utility by quality has traditionally limited sample utility and hindered research efforts. The present invention provides a method in which low-quality RNA can be used to obtain meaningful multigene expression results from samples containing low concentrations of RNA, such as thyroid FNA samples.

[0095] Furthermore, samples with low and / or non-measurable RNA concentrations by NanoDrop, which are usually considered unsuitable for multiplex gene expression profiling, can be measured and analyzed using the methods and algorithms of the present invention. Today, the most sensitive "state of the art" instrument used in laboratories to measure nucleic acid yield is the NanoDrop spectrophotometer. As with many quantification instruments of its kind, the accuracy of NanoDrop measurements is significantly reduced at very low RNA concentrations. The minimum amount of RNA required for input into microarray experiments also limits the usefulness of a given sample. In the present invention, samples containing very small amounts of nucleic acid can be estimated using a combination of measurements from both the NanoDrop and Bioanalyzer instruments, thereby optimizing the sample for multiplex gene expression assays and analysis.

[0096] (iii) Protein content adequacy test In some cases, the protein content in a biological sample can be measured using a variety of methods known in the art, including, but not limited to, ultraviolet absorbance at 280 nanometers, cell staining as described herein, or protein staining, for example, with Coomassie Blue or bicinchoninic acid. In some cases, proteins are extracted from the biological sample before measuring the sample. In some cases, multiple tests for the adequacy of the sample can be performed in parallel or one by one. In some cases, the sample can be divided into aliquots to perform multiple diagnostic tests before, during, or after the adequacy assessment. In some cases, the adequacy test is performed on a small amount of the sample that may or may not be suitable for further diagnostic tests. In other cases, the entire sample is assessed for adequacy. In any case, the cost of the test for adequacy can be charged to the subject, a medical provider, an insurance provider, or a government entity.

[0097] In some embodiments of the invention, samples may be tested for adequacy immediately or shortly after collection. In some cases, if sample adequacy testing does not indicate a sufficient quantity or sufficient quality of the sample, additional samples may be taken.

[0098] VI. Analysis of Samples In one aspect, the present invention provides a method for performing microarray gene expression analysis using polynucleotides such as DNA or RNA with low quantity and quality. In some embodiments, the present disclosure describes a method for diagnosing, characterizing, and / or monitoring cancer by analyzing gene expression using RNA with low quantity and quality. In one embodiment, the cancer is thyroid cancer. Thyroid RNA can be obtained from fine needle aspirates (FNA). In some embodiments, gene expression profiles are obtained from degraded samples with RNA RIN values ​​of 9.0, 8.0, 7.0, 6.0, 5.0, 4.0, 3.0, 2.0, 1.0, or less. In certain embodiments, gene expression profiles are obtained from samples with RINs of 6 or less, i.e., 6.0, 5.0, 4.0, 3.0, 2.0, 1.0, or less. The present invention provides a method in which low quality RNA can be used to obtain meaningful gene expression results from samples containing low concentrations of nucleic acids, such as thyroid FNA samples.

[0099] Another estimate of a sample's usefulness for gene expression assays is RNA yield, typically measured in nanogram to microgram amounts. The most sensitive "state of the art" instrument used in laboratories today to measure nucleic acid yield is the NanoDrop spectrophotometer. As with many quantification instruments of its type, the precision of NanoDrop measurements is significantly reduced at very low RNA concentrations. The minimum amount of RNA required for input to a microarray experiment also limits the usefulness of a given sample. In some aspects, the present invention solves the problem of low RNA concentrations by estimating sample input using a combination of measurements from both the NanoDrop and Bioanalyzer instruments. Because the quality of data obtained from gene expression studies depends on the amount of RNA, meaningful gene expression data can be generated from samples with low or unmeasurable RNA concentrations measured by the NanoDrop.

[0100] The methods and algorithms enable: (1) gene expression analysis of samples containing low amounts and / or low quality nucleic acid; (2) significant reduction in false positives and false negatives; (3) determination of the genetic, metabolic, or signaling pathways underlying the resulting pathology; (4) the ability to assign statistical probability to the accuracy of the diagnosis of a genetic disorder; (5) the ability to resolve ambiguous results; and (6) the ability to distinguish between cancer subtypes.

[0101] Cytological analysis Samples may be analyzed by cell staining combined with microscopic examination of cells in biological samples. Cell staining, or cytological examination, may be performed by a number of methods and suitable reagents known in the art, including but not limited to: EA staining, hematoxylin staining, cytostain, Papanicolaou staining, eosin, Nissl staining, toluidine blue, silver staining, azocarmine staining, neutral red, or Janus green. In some cases, cells are fixed and / or permeabilized, for example, with methanol, ethanol, glutaraldehyde, or formaldehyde, before or during the staining procedure. In some cases, cells are not fixed. In some cases, multiple stains are used in combination. In other cases, no stain is used at all. In some cases, measurement of nucleic acid content is performed using staining methods, for example, with ethidium bromide, hematoxylin, Nissl staining, or any nucleic acid stain known in the art.

[0102] In some embodiments of the present invention, cells may be smeared onto slides by standard methods known in the art for cytological investigation. In other cases, liquid based cytology (LBC) methods may be used. In some cases, LBC methods provide improved means of cytology slide preparation, more uniform samples, increased sensitivity and specificity, and improved sample handling efficiency. In liquid based cytology methods, biological samples are transferred from subjects to containers or vials containing liquid based cytology preparation solutions, such as Cytyc ThinPrep, SurePath, Monoprep, or other liquid based cytology preparation solutions known in the art. Furthermore, to ensure a substantially quantitative transfer of the sample, the sample may be rinsed from the collection device into the container or vial with liquid based cytology preparation solution. The solution containing the biological sample in the liquid based cytology preparation solution may then be stored and / or processed by a machine or by a person skilled in the art to create a layer of cells on a glass slide. The samples may further be stained and examined under a microscope in the same manner as conventional cytological preparations.

[0103] In some embodiments of the present invention, the sample may be analyzed by immunohistochemical staining. Immunohistochemical staining provides for the analysis of the presence, location, and distribution of specific molecules or antigens in a biological sample (e.g., cells or tissues) by the use of antibodies. Antigens can be small molecules, proteins, peptides, nucleic acids, or any other molecules that can be specifically recognized by antibodies. Samples may be analyzed by immunohistochemical methods that may or may not include a prior fixation and / or permeabilization step. In some cases, the antigen of interest may be detected by contacting the sample with an antigen-specific antibody, followed by one or more washes to remove non-specific binding. The specifically bound antibody may then be detected by an antibody detection reagent, such as, for example, a labeled secondary antibody, or labeled avidin / streptavidin. In some cases, the antigen-specific antibody may instead be directly labeled. Suitable labels for immunohistochemistry include fluorophores, such as fluorescein and rhodamine, enzymes, such as alkaline phosphatase and horseradish peroxidase, and 32 P and 125 Immunohistochemical staining includes, but is not limited to, radionuclides such as I. Gene product markers that can be detected by immunohistochemical staining include, but are not limited to, Her2 / Neu, Ras, Rho, EGFR, VEGFR, UbcH10, RET / PTC1, cytokeratin 20, calcitonin, GAL-3, thyroid peroxidase, and thyroglobulin.

[0104] VII. Assay Results The results of routine cytological or other assays may indicate a sample as negative (no cancer, disease, or condition), equivocal or suspicious (suggesting the presence of cancer, disease, or condition), diagnostic (positive diagnosis for cancer, disease, or condition), or non-diagnostic (providing insufficient information regarding the presence or absence of cancer, disease, or condition). The diagnostic result may be further classified as malignant or benign. The diagnostic result may also provide a score indicating, for example, the severity or aggressiveness of the cancer, or the likelihood of a correct diagnosis, for example, via a p-value, a corrected p-value, or a statistical confidence index. In some cases, the diagnostic result may indicate a particular type of cancer, disease, or condition, such as, for example, follicular adenoma, Hurthle cell adenoma, lymphocytic thyroiditis, hyperplasia, follicular carcinoma, follicular type papillary thyroid carcinoma, papillary carcinoma, or any of the diseases or conditions provided herein. In some cases, the diagnostic result may indicate a particular stage of the cancer, disease, or condition. The diagnostic results may inform a particular treatment or therapeutic intervention for the particular type or stage of cancer, disease, or condition diagnosed. In some embodiments, the results of the performed assays may be entered into a database. The molecular profiling company may bill the individual, insurance provider, health care provider, or government entity for one or more of the following: the performed assays, advisory services, reporting of results, database access, or data analysis. In some cases, all or some steps other than the molecular profiling are performed by a cytology laboratory or health care professional.

[0105] VIII. Molecular Profiling Cytological assays characterize the current diagnostic standard for many types of suspected tumors, including, for example, thyroid tumors or nodules. In some embodiments of the present invention, samples that are assayed as negative, indeterminate, diagnostic, or non-diagnostic may be subjected to subsequent assays to obtain further information. In the present invention, these subsequent assays include molecular profiling steps of genomic DNA, RNA, mRNA expression product levels, miRNA levels, gene expression product levels, or gene expression product alternative splicing. In some embodiments of the present invention, molecular profiling refers to the determination of the number (e.g., copy number) and / or type of genomic DNA in a biological sample. In some cases, the number and / or type may be further compared to a control sample or a sample considered normal. In some embodiments, the genomic DNA may be analyzed for copy number variations, such as copy number gains (amplifications) or losses, or variants such as insertions, deletions, truncations, etc. Molecular profiling may be performed on the same sample, a portion of the same sample, or a new sample, which may be obtained using any of the methods described herein. The molecular profiling company may request additional samples by contacting the individual directly or through an intermediary such as a physician, a third-party testing facility or laboratory, or a medical professional. In some cases, the sample is assayed using the methods and compositions of the molecular profiling business in combination with some or all of the cytological stains or other diagnostic methods. In other cases, the sample is assayed directly using the methods and compositions of the molecular profiling business without the prior use of routine cytological stains or other diagnostic methods. In some cases, the results of molecular profiling, alone or in combination with cytology or other assay methods, allow a person skilled in the art to diagnose and recommend treatment for the subject. In some cases, molecular profiling can be used alone or in combination with cytology to monitor a tumor or suspected tumor over time for malignant changes.

[0106] The molecular profiling method of the present invention provides for extracting and analyzing protein or nucleic acid (RNA or DNA) from one or more biological samples from a subject. In some cases, nucleic acid is extracted from the entire sample obtained. In other cases, nucleic acid is extracted from a portion of the sample obtained. In some cases, a portion of the sample that is not subjected to nucleic acid extraction may be analyzed by cytological examination or immunohistochemistry. Methods for extracting RNA or DNA from biological samples are well known in the art and include, for example, the use of commercially available kits such as Qiagen DNeasy Blood and Tissue Kit or Qiagen EZ1 RNA Universal Tissue Kit.

[0107] (i) Tissue fingerprinting In many cases, biological samples such as those provided by the methods of the present invention may contain several cell types or tissues, including, but not limited to, thyroid follicular cells, thyroid medullary cells, blood cells (RBC, WBC, platelets), smooth muscle cells, ducts, ductal cells, basement membrane, luminal, lobule, adipose tissue, skin cells, epithelial cells, and infiltrating macrophages and lymphocytes. In the case of a thyroid sample, the diagnostic classification of the biological sample may, for example, include primarily follicular cells (for cancers originating from follicular cells, such as papillary carcinoma, follicular carcinoma, and anaplastic thyroid carcinoma), and medullary cells (for medullary carcinoma). The diagnosis of an indeterminate biological sample from a thyroid biopsy, in some cases, concerns the distinction between follicular adenoma and follicular carcinoma. Thus, for example, the molecular profiling signal of follicular cells may be diluted and confused with other cell types present in the sample. Similarly, the diagnosis of biological samples from other tissues or organs often includes the diagnosis of one or more cell types among many cell types that may be present in the sample.

[0108] In some embodiments, the method of the present invention provides an upfront method to determine the cellular composition of a particular biological sample so that the resulting molecular profiling signature can be calibrated against the dilution effect of the presence of other cell and / or tissue types. In one aspect, the upfront method is an algorithm that uses a combination of known cell and / or tissue specific gene expression patterns as an upfront mini-classifier for each component of the sample. The algorithm uses this molecular fingerprint to pre-classify the sample according to composition and then apply correction / normalization factors. This data is then fed into a final classification algorithm that in some cases will incorporate that information to aid in the final diagnosis.

[0109] (ii) Genomic analysis In some embodiments, genome sequence analysis or genotyping can be carried out on sample.This genotyping can take the form of single nucleotide polymorphism (SNP) analysis, insertion deletion polymorphism (InDel) analysis, variable number of tandem repeat (VNTR) analysis, copy number variation (CNV) analysis, or mutation analysis such as partial genome sequencing or whole genome sequencing.The method of carrying out genome analysis is known in the art and can include high-throughput sequencing, for example, but not limited to the method described in U.S. Patent No. 7,335,762; No. 7,323,305; No. 7,264,929; No. 7,244,559; No. 7,211,390; No. 7,361,488; No. 7,300,788; and No. 7,280,922. The method of performing genomic analysis may also include microarray methods as described below. In some cases, genomic analysis may be performed in combination with any of the other methods described herein. For example, a sample can be obtained, tested for validity, and divided into aliquots. One or more aliquots can then be used for the cytological analysis of the present invention, one or more can be used for the RNA expression profiling method of the present invention, and one or more can be used for genomic analysis. It is further understood that the present invention anticipates that those skilled in the art may wish to perform other analyses on biological samples that are not expressly provided herein.

[0110] (iii) Expression product profiling Gene expression profiling is the measurement of the activity (expression) of thousands of genes at a time to create a complete picture of cell function. These profiles can, for example, distinguish between actively dividing cells or show how cells respond to a particular treatment. Many experiments of this kind measure the entire genome, i.e., all genes present in a particular cell, simultaneously. Microarray technology measures the relative activity of previously identified target genes. Sequence-based techniques such as serial analysis of gene expression (SAGE, SuperSAGE) are also used for gene expression profiling. SuperSAGE is particularly accurate and can measure any active gene, not just a predefined set. In RNA, mRNA, or gene expression profiling microarrays, the expression levels of thousands of genes are monitored simultaneously to study the effects of certain treatments, diseases, and developmental stages on gene expression. For example, microarray-based gene expression profiling can be used to characterize the genetic signatures of the genetic disorders disclosed herein, or different cancer types, cancer subtypes, and / or cancer stages.

[0111] Expression profiling experiments often involve measuring the relative amounts of gene expression products, such as mRNA, expressed in two or more experimental conditions. This is because altered levels of a particular sequence of gene expression products suggest changes in the need for the protein encoded by the gene expression product, possibly indicating a homeostatic response or a pathological condition. For example, if breast cancer cells express higher levels of mRNA associated with a particular transmembrane receptor than normal cells, this receptor may play a role in breast cancer. One aspect of the present invention involves profiling gene expression as part of an important diagnostic test for genetic disorders and cancer, particularly thyroid cancer.

[0112] In some embodiments, RNA samples with RIN≦5.0 are typically not used for multigene microarray analysis, but instead can only be used for single-gene RT-PCR assay and / or TaqMan assay.Microarray assay, RT-PCR assay and TaqMan assay are standard molecular techniques well known in related fields.TaqMan probe-based assay is widely used in real-time PCR, including gene expression assay, DNA quantification and SNP genotyping.

[0113] In one embodiment, gene expression products related to cancer known in the art are profiled.Such gene expression products are described and disclosed in U.S. Patent Nos. 7,358,061; 7,319,011; 5,965,360; 6,436,642; and U.S. Patent Application Nos. 2003 / 0186248, 2005 / 0042222, 2003 / 0190602, 2005 / 0048533, 2005 / 0266443, 2006 / 0035244, 2007 / 0035244, 2009 / 0046244, 2010 / 00266244, 2011 / 00266244, 2012 / 00266244, 2013 / 00266244, 2014 / 00266244, 2015 / 00266244, 2016 / 00266244, 2017 / 00266244, 2018 / 00266244, 2019 / 0035244, 2015 / 0026 ... This includes, but is not limited to, the gene expression products detailed in Nos. 2006 / 083744, 2006 / 0088851, 2006 / 0105360, 2006 / 0127907, 2007 / 0020657, 2007 / 0037186, 2007 / 0065833, 2007 / 0161004, 2007 / 0238119, and 2008 / 0044824.

[0114] It is further expected that other gene expression products associated with cancer may become known, and that the methods and compositions described herein may include such newly discovered gene expression products.

[0115] In some embodiments of the present invention, gene expression products are alternatively or additionally analyzed for features other than expression level. For example, gene products can be analyzed for alternative splicing. Alternative splicing, also called alternative exon usage, is an RNA splicing variation mechanism in which exons of the primary gene transcript, pre-mRNA, are separated and reconnected (i.e., spliced) to produce alternative mRNA molecules from the same gene. In some cases, these linear combinations are then subjected to the process of translation, where each of the alternative mRNA molecules from the same gene specifies a specific unique sequence of amino acids, resulting in a protein isoform. Alternative splicing can include the incorporation of different exons or different sets of exons, the retention of certain introns, or the use of alternative splice donor and acceptor sites.

[0116] In some cases, a marker or set of markers can be identified that shows diagnostic alternative splicing for benign, malignant or normal samples.In addition, alternative splicing markers can further provide diagnosis for certain types of thyroid cancer (e.g., papillary, follicular, medullary or anaplastic).Alternative splicing markers that are diagnostic for malignant lesions known in the art include those listed in U.S. Patent No. 6,436,642.

[0117] In some cases, the expression of non-protein-coding RNA expression products, such as miRNA and siRNA, can be assayed by the method of the present invention.The differential expression of these RNA expression products can indicate benign, malignant or normal samples.The differential expression of these RNA expression products can further indicate subtypes of benign (e.g., FA, NHP, LCT, BN, CN, HA) or malignant (e.g., FC, PTC, FVPTC, ATC, MTC) samples.In some cases, the differential expression of miRNA, siRNA, alternative splice RNA isoforms, mRNA, or any combination thereof can be assayed by the method of the present invention.

[0118] In some embodiments, the present invention provides a panel of 16 biomarkers each required to characterize, rule out, or diagnose a pathology within the thyroid gland. The 16 panels are as follows: 1 Normal thyroid (NML) 2 Lymphocytic autoimmune thyroiditis (LCT) 3 Nodular hyperplasia (NHP) 4. Follicular thyroid adenoma (FA) 5 Hersl cell thyroid adenoma (HC) 6 Parathyroid glands (non-thyroid tissue) 7 Anaplastic thyroid carcinoma (ATC) 8 Follicular Thyroid Carcinoma (FC) 9 Hersl cell thyroid cancer (HC) 10 Papillary thyroid carcinoma (PTC) 11 Follicular papillary carcinoma (FVPTC) 12 Medullary thyroid carcinoma (MTC) 13 Metastasis of renal cancer to the thyroid gland 14 Thyroid metastasis of melanoma 15 Thyroid metastasis of B-cell lymphoma 16 Breast cancer metastasis to the thyroid gland

[0119] Each panel contains a set of biomarkers required to characterize, rule out, and diagnose a given condition in the thyroid. Panels 1-6 represent benign conditions. Panels 7-16 represent malignant conditions.

[0120] The biology of the thyroid gland and each pathology found therein suggests that there will be overlap between the biomarkers in one panel and the biomarkers in another panel. Reflecting each pathology subtype, each diagnostic panel is heterogeneous and semi-overlapping with the biomarkers in other panels. The heterogeneity and overlap reflects the biology of the tissue collected in a given FNA and the gene expression differences that characterize each pathology subtype from one another.

[0121] In one aspect, the diagnostic value of the present invention lies in the comparison of (i) one or more markers of a panel with (ii) one or more markers of each additional panel. A potential application of the present invention is a higher diagnostic accuracy in FNA than is currently possible by other means.

[0122] In some embodiments, the biomarkers in each panel are interchangeable (modular). To accommodate the definition of new pathological subtypes (e.g., new case reports of metastases to the thyroid from other organs), the biomarkers of all panels can be replaced, increased, decreased, or improved. The present invention describes multiple markers that define each of the 16 heterogeneous, semi-overlapping, distinct pathologies found in the thyroid. All 16 panels are required to arrive at an accurate diagnosis, and a given panel alone does not have sufficient power to make a true diagnostic decision. In some embodiments, in the context of investigating multiple biomarkers that define all other pathological subtypes, the biomarkers of each panel are replaced with the appropriate combination of biomarkers, so that the multiple biomarkers of each panel still define a given pathological subtype.

[0123] The methods and compositions of the invention can have genes selected from 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, or 16 or more biomarker panels and can have 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50 or more gene expression products from each biomarker panel in any combination. In some embodiments, the set of combined genes provides a specificity or sensitivity of greater than 70%, 75%, 80%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 99.5%, or a positive or negative predictive value of at least 95%, 95.5%, 96%, 96.5%, 97%, 97.5%, 98%, 98.5%, 99%, 99.5% or more.

[0124] (1) In vitro methods for measuring expression product levels General methods for measuring gene expression product levels are known in the art and may include, but are not limited to, one or more of the following: additional cytological assays, assays for specific protein or enzyme activity, assays for specific expression products including proteins or RNA or specific RNA splice variants, in situ hybridization, whole or partial genome expression analysis, microarray hybridization assays, SAGE, enzyme-linked immunosorbent assays, mass spectrometry, immunohistochemistry, or blotting. Gene expression product levels may be normalized to an internal standard such as total mRNA, or the expression level of a specific gene, including, but not limited to, glyceraldehyde-3-phosphate dehydrogenase or tubulin.

[0125] In some embodiments of the present invention, gene expression product markers and alternative splicing markers can be determined by microarray analysis, for example, using Affymetrix arrays, cDNA microarrays, oligonucleotide microarrays, spotted microarrays, or other microarray products from Biorad, Agilent, or Eppendorf. Microarrays offer certain advantages because they can contain a large number of genes or alternative splice variants that can be assayed in a single experiment. In some cases, the microarray device can contain the entire human genome or transcriptome, or a substantial fraction thereof, allowing for comprehensive evaluation of gene expression patterns, genomic sequences, or alternative splicing. Markers can be found using standard molecular biology and microarray analysis techniques, such as those described in Sambrook Molecular Cloning a Laboratory Manual 2001 and Baldi, P., and Hatfield, WG, DNA Microarrays and Gene Expression 2002.

[0126] Microarray analysis begins with extracting and purifying nucleic acid from biological samples (e.g., biopsy or fine needle aspirate) using methods known in the art.For expression and alternative splicing analysis, it may be advantageous to extract and / or purify RNA from DNA.It may be even more advantageous to extract and / or purify mRNA from other types of RNA, such as tRNA and rRNA.

[0127] The purified nucleic acid can be further labeled with a fluorescent, radionuclide, or chemical label such as biotin or digoxin, for example, by reverse transcription, PCR, ligation, chemical reaction, or other techniques. The labeling can be direct or indirect and may require a further coupling step. The coupling step can be performed before hybridization, for example, using aminoallyl-UTP and an NHS amino-reactive dye (such as a cyanine dye), or after, for example, using biotin and labeled streptavidin. Modified nucleotides (e.g., in a 1 aaUTP:4 TTP ratio) are enzymatically added at a lower rate compared to normal nucleotides, typically 1 base out of every 60 bases (as measured by a spectrophotometer). The aaDNA can then be purified, for example, by a column or diafiltration device. The aminoallyl group is an amine group on a long linker attached to the nucleic acid base, which reacts with a reactive label (e.g., a fluorescent dye).

[0128] The labeled sample can then be mixed with a hybridization solution that may contain SDS, SSC, dextran sulfate, a blocking agent (such as COT1 DNA, salmon sperm DNA, calf thymus DNA, PolyA, or PolyT), Denhardt's solution, formamide, or combinations thereof.

[0129] A hybridization probe is a DNA or RNA fragment of variable length that is used to detect the presence in a DNA or RNA sample of a nucleotide sequence (DNA target) that is complementary to the sequence in the probe. Thus, the probe hybridizes to a single-stranded nucleic acid (DNA or RNA) that has a base sequence that allows probe-target base pairing due to the complementarity between the probe and the target. The labeled probe is first denatured (by heating or under alkaline conditions) into a single DNA strand, and then hybridized to the target DNA.

[0130] To detect hybridization of the probe to a target sequence, the probe is tagged (or labeled) with a molecular marker. Commonly used markers include: 32 P, or digoxigenin, a non-radioactive antibody-based marker. The hybridized probe is then visualized via autoradiography or other imaging techniques to detect DNA sequences or RNA transcripts with moderate to high sequence similarity to the probe. Detection of sequences with moderate or high similarity depends on how stringent hybridization conditions are applied; high stringency, such as high hybridization temperature and low salt in the hybridization buffer, allows only hybridization between highly similar nucleic acid sequences, while low stringency, such as lower temperature and high salt, allows hybridization even with relatively low sequence similarity. The hybridization probe used in DNA microarrays refers to DNA that is covalently attached to an inert surface, such as a coated glass slide or a gene chip, and to which a mobile cDNA target is hybridized.

[0131] This mixture can then be denatured by heat or chemical means and added to the ports of the microarray. The holes can then be sealed and the microarray hybridized, for example, in a hybridization oven where the microarray is mixed by rotation, or in a mixer. After overnight hybridization, non-specific binding can be washed away (e.g., with SDS and SSC). The microarray can then be dried and scanned with a specialized machine where a laser excites the dye and a detector measures its emission. A template grid can be overlaid on the image and the intensity of features (several pixels make up one feature) can be quantified.

[0132] Various kits can be used for the amplification of nucleic acids and probe generation of the method of the present invention. Examples of kits that can be used in the present invention include, but are not limited to, Nugen WT-Ovation FFPE kit, cDNA amplification kit including Nugen Exon Module and Frag / Label module. NuGEN WT-Ovation™ FFPE System V2 is a whole transcriptome amplification system that allows for performing comprehensive gene expression analysis on vast archives of small degraded RNA derived from FFPE samples. This system is composed of the reagents and protocols required for amplification of as little as 50ng of total FFPE RNA. Protocols can be used for qPCR, sample archiving, fragmentation, and labeling. Amplified cDNA can be fragmented and labeled for GeneChip® 3' expression array analysis using NuGEN's FL-Ovation™ cDNA Biotin Module V2 in less than 2 hours. For analysis using Affymetrix GeneChip® Exon and Gene ST arrays, the amplified cDNA is used with the WT-Ovation Exon Module and then fragmented and labeled using the FL-Ovation® cDNA Biotin Module V2. For analysis on Agilent arrays, the amplified cDNA can be fragmented and labeled using NuGEN's FL-Ovation™ cDNA Fluorescent Module. More information on the Nugen WT-Ovation FFPE kit can be obtained at http: / / www.nugeninc.com / nugen / index.cfm / products / amplification-systems / wt-ovation-ffpe / .

[0133] In some embodiments, Ambion WT-expression kit can be used. Ambion WT-expression kit allows direct amplification of total RNA without separate ribosomal RNA (rRNA) depletion step. With Ambion® WT Expression Kit, as little as 50ng of total RNA sample can be analyzed on Affymetrix® GeneChip® Human Exon 1.0 ST Array, Mouse Exon 1.0 ST Array, and Rat Exon 1.0 ST Array, and Human Gene 1.0 ST Array, Mouse Gene 1.0 ST Array, and Rat Gene 1.0 ST Array. In addition to the lower input RNA requirement and the high concordance between Affymetrix® method and TaqMan® real-time PCR data, Ambion® WT Expression Kit provides a significant increase in sensitivity. For example, the Ambion® WT Expression Kit can result in a greater number of probe sets being detected above background at the exon level as a result of increased signal-to-noise ratio. The Ambion WT-expression kit can be used in combination with additional Affymetrix labeling kits.

[0134] In some embodiments, AmpTec Trinucleotide Nano mRNA Amplification Kit (6299-A15) can be used in the methods of the present invention. ExpressArt® TRinucleotide mRNA amplification Nano kit is suitable for a wide range of input total RNA from 1 ng to 700 ng. Depending on the amount of input total RNA and the required aRNA yield, it can be used in one round (input >300 ng total RNA) or two rounds (minimum input amount 1 ng total RNA), with aRNA yields in the range of >10 μg. AmpTec's proprietary TRinucleotide priming technology, in combination with selection against rRNA, results in preferential amplification of mRNA (unrelated to the universal eukaryotic 3' poly(A) sequence). More information regarding the AmpTec Trinucleotide Nano mRNA Amplification Kit can be obtained at http: / / www.amp-tec.com / products.htm. This kit can be used in combination with cDNA conversion kits and Affymetrix labeling kits.

[0135] The raw data can then be normalized, for example, by subtracting background intensity and then dividing the intensity to equalize the total intensity of the feature or the intensity of the reference gene on each channel.The t-value for all intensities can then be calculated.More advanced methods for Affymetrix chips include z-ratio regression, loess regression, and lowess regression, as well as RMA (robust multichip analysis).

[0136] (2) In vivo methods for measuring gene expression product levels It is further anticipated that the methods and compositions of the present invention can be used to measure gene expression product levels in an individual without first obtaining a sample. For example, gene expression product levels can be measured in vivo, i.e., in an individual. Methods for measuring gene expression product levels in vivo are known in the art and include imaging techniques such as CAT, MRI; NMR; PET; and optical, fluorescent, or biophotonic imaging of protein or RNA levels using antibodies or molecular beacons. Such methods are described in US 2008 / 0044824, US 2008 / 0131892, which are incorporated herein by reference. Additional methods for in vivo molecular profiling are contemplated to be within the scope of the present invention.

[0137] In some embodiments of the present invention, molecular profiling includes binding a sample or a portion of a sample to one or more probes of the present invention. Suitable probes are those that bind to the components of the sample to be measured, i.e., gene products, and include, but are not limited to, antibodies or antibody fragments, aptamers, nucleic acids, and oligonucleotides. Binding of a sample to a probe of the present invention represents the conversion of the substance that binds to the one or more probes between samples. A method of diagnosing cancer based on molecular profiling further includes detecting the gene expression products (i.e., mRNA or protein) and levels of a sample, comparing it with the amount in a normal control sample to determine the differential gene expression product level between the sample and the control; and classifying the test sample by inputting the one or more differential gene expression product levels into the trained algorithm of the present invention; validating the classification of the sample using the selection and classification algorithm of the present invention; and identifying the sample as positive for a genetic disorder or a type of cancer.

[0138] (i) Comparison of the sample with normal The results of molecular profiling carried out on the sample provided by an individual (test sample) can be compared with biological samples that are known to be normal or assumed to be normal. A normal sample is a sample that is expected to be free of or free of cancer, disease, or condition, or that is considered to test negative for cancer, disease, or condition in a molecular profiling assay. A normal sample can be derived from an individual different from the individual being tested, or from the same individual. In some cases, a normal sample is, for example, a sample obtained from a buccal swab of an individual such as the individual being tested. A normal sample can be assayed at the same time as a test sample, or at a different time point.

[0139] The results of the assay for the test sample can be compared with the results of the same assay for a normal sample. In some cases, the results of the assay for the normal sample are from a database or from a reference. In some cases, the results of the assay for the normal sample are known or are values ​​generally accepted by those skilled in the art. In some cases, the comparison is qualitative. In other cases, the comparison is quantitative. In some cases, the qualitative or quantitative comparison can include, but is not limited to, one or more of the following: fluorescence values, spot intensities, absorbance values, chemiluminescent signals, histograms, critical thresholds, statistical significance values, gene product expression levels, changes in gene product expression levels, alternative exon usage, changes in alternative exon usage, protein levels, DNA polymorphisms, copy number polymorphisms, indicators of the presence or absence of one or more DNA markers or regions, or nucleic acid sequence comparisons.

[0140] (ii) Evaluation of results In some embodiments, the molecular profiling results are evaluated using methods known in the art to correlate gene product expression levels or alternative exon usage with a particular phenotype, such as malignancy, malignant type (e.g., follicular carcinoma), benign, or normal (e.g., no disease or condition). In some cases, a specified statistical confidence level may be determined to provide a confidence level of the diagnosis. For example, a confidence level of greater than 90% may be determined to be a useful predictor of malignancy, malignant type, or benign. In other embodiments, a more stringent or less stringent confidence level may be selected. For example, a confidence level of approximately 70%, 75%, 80%, 85%, 90%, 95%, 97.5%, 99%, 99.5%, or 99.9% may be selected as a useful phenotype predictor. The provided confidence level may in some cases be related to the quality of the sample, the quality of the data, the quality of the analysis, the particular method used, and the number of gene expression products analyzed. The specified confidence level for providing a diagnosis can be selected based on the expected number and / or cost of false positives or false negatives. Methods for selecting parameters to achieve a specified confidence level or to identify markers with diagnostic power include, but are not limited to, receiver operator curve analysis (ROC), binormal ROC, principal component analysis, partial least squares analysis, singular value decomposition, least absolute shrinkage and selection operator (LASSO) analysis, least angle regression, and threshold gradient directed regularization.

[0141] (iii) Data Analysis Raw gene expression levels and alternative splicing data may in some cases be improved through the application of algorithms designed to normalize and / or improve the reliability of the data. In some embodiments of the present invention, data analysis requires a computer or other device, machine, or instrument for the application of the various algorithms described herein due to the large number of individual data points to be processed. "Machine learning algorithm" refers to a computation-based predictive methodology, also known to those skilled in the art as a "classifier", used to characterize gene expression profiles. For example, signals corresponding to certain expression levels obtained by microarray-based hybridization assays are typically subjected to an algorithm to classify the expression profile. Supervised learning generally involves "training" a classifier to recognize class distinctions and then "testing" the accuracy of the classifier on an independent test set. For new unknown samples, a classifier can be used to predict the class to which the sample belongs.

[0142] In some cases, the robust multi-array Average (RMA) method can be used to normalize the raw data. The RMA method starts by calculating background-corrected intensities for each matched cell on multiple microarrays. The background-corrected values ​​are restricted to positive values ​​as described by Irizarry et al. Biostatistics 2003 April 4 (2): 249-64. After background correction, the base 2 logarithm of the background-corrected intensity of each matched cell is then obtained. The background-corrected log-transformed matched intensities on each microarray are then normalized using the quantile normalization method, in which the array percentile probe value is replaced with the average value of all array percentile points for each input array and each probe expression value. This method is more fully described by Bolstad et al. Bioinformatics 2003. After quantile normalization, the normalized data can then be fitted to a linear model to obtain expression measurements for each probe on each microarray. Tukey's median polish algorithm (Tukey, JW, Exploratory Data Analysis. 1977) can then be used to measure log-scale expression levels for the normalized probe set data.

[0143] Data can be further filtered to remove data that may be considered suspicious.In some embodiments, data derived from microarray probes that have less than about 4, 5, 6, 7 or 8 guanosine + cytosine nucleotides can be considered unreliable due to abnormal hybridization tendency or secondary structure problem.Similarly, data derived from microarray probes that have more than about 12, 13, 14, 15, 16, 17, 18, 19, 20, 21 or 22 guanosine + cytosine nucleotides can be considered unreliable due to abnormal hybridization tendency or secondary structure problem.

[0144] In some cases, probe sets with low reliability can be selected for exclusion from data analysis by ranking the reliability of the probe sets against a set of reference data sets. For example, RefSeq or Ensembl (EMBL) are considered to be extremely high quality reference data sets. Data from probe sets that match RefSeq or Ensembl sequences can be specifically included in microarray analysis experiments in some cases due to their expected high reliability. Similarly, data from probe sets that match low reliability reference data sets can be excluded from further analysis or can be considered for inclusion or not individually. In some cases, Ensembl high throughput cDNA (HTC) and / or mRNA reference data sets can be used separately or together to determine the reliability of the probe sets. In other cases, the reliability of the probe sets can be ranked. For example, probes and / or probe sets that perfectly match all reference data sets, such as RefSeq, HTC, and mRNA, can be ranked as the most reliable (1). Furthermore, the probes and / or probe sets that match two of the three reference data sets can be ranked as the second most reliable (2), the probes and / or probe sets that match one of the three reference data sets can be ranked next (3), and the probes and / or probe sets that do not match any reference data set can be ranked last (4).The probes and / or probe sets can then be included or excluded from the analysis based on their ranking.For example, for further analysis, we can choose to include data from probe sets in categories 1, 2, 3, and 4; we can choose to include data from probe sets in categories 1, 2, and 3; we can choose to include data from probe sets in categories 1 and 2; or we can choose to include data from probe sets in category 1.In another example, the probe sets can be ranked by the number of base pair mismatches to the reference data set entries.Many methods for assessing the reliability of a given probe and / or probe set for molecular profiling are understood in the art, and it is understood that the methods of the present invention encompass any of these methods and combinations thereof.

[0145] In some embodiments of the invention, data from a probe set may be excluded from the analysis if it is not expressed or is expressed at undetectable levels (below background). A probe set is determined to be expressed above background for any group if: The integral of the standard normal distribution from T0 to infinity is < 0.01. Where: T0=Sqr(GroupSize)(TP) / Sqr(Pvar), GroupSize=number of CEL files in the group, T = mean probe score within the probe set; P = the mean of the background probe mean values ​​of GC content, and Pvar=total background probe variance / (number of probes in the probe set)^2.

[0146] This allows for the inclusion of probe sets whose mean within a group is greater than the mean expression of background probes of similar GC content to the probe set's probes as the central background for the probe set, and this allows for the derivation of probe set variance from the background probe set variance.

[0147] In some embodiments of the present invention, probe sets that show no or low variance can be excluded from further analysis.Low variance probe sets are excluded from analysis via chi-square test.A probe set is considered to be low variance if its transformed variance is on the left side of the 99 percent confidence interval of the chi-square distribution with (N-1) degrees of freedom. (N-1) *Probe set distribution / (gene probe set distribution) ~ Chi-Sq(N-1) where N is the number of input CEL files, (N-1) is the degrees of freedom for the chi-squared distribution, and "probeset variance for genes" is the average probeset variance across genes.

[0148] In some embodiments of the invention, a probe set for a given gene or transcript cluster may be eliminated from further analysis if it contains fewer than a minimum number of probes that pass the previously described filtering steps for GC content, reliability, variance, etc. For example, in some embodiments, a probe set for a given gene or transcript cluster may be eliminated from further analysis if it contains less than about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, or less than about 20 probes.

[0149] The methods of gene expression level or alternative splicing data analysis may further include the use of feature selection algorithms as provided herein. In some embodiments of the invention, feature selection is provided by use of the LIMMA software package (Smyth, GK (2005). Limma: linear models for microarray data. In: Bioinformatics and Computational Biology Solutions using R and Bioconductor, R. Gentleman, V. Carey, S. Dudoit, R. Irizarry, W. Huber (eds.), Springer, New York, pages 397-420).

[0150] The method of gene expression level and / or alternative splicing data analysis may further include the use of a pre-classifier algorithm. For example, the algorithm may use cell-specific molecular fingerprints to pre-classify samples according to composition and then apply correction / normalization factors. This data / information may then be fed into a final classification algorithm that will incorporate the information to aid in the final diagnosis.

[0151] The method of gene expression level and / or alternative splicing data analysis may further comprise the use of classification algorithms as provided herein. In some embodiments of the present invention, a support vector machine (SVM) algorithm, a random forest algorithm, or a combination thereof is provided for classification of microarray data. In some embodiments, the markers identified that distinguish samples (e.g., benign or malignant, normal or malignant) or subtypes (e.g., PTC or FVPTC) are selected based on statistical significance. In some cases, statistical significance selection is performed after applying the Benjamini Hochberg correction for false discovery rate (FDR).

[0152] In some cases, the classification algorithm may be supplemented with a meta-analysis approach, such as that described by Fishel and Kaufman et al. 2007 Bioinformatics 23(13): 1599-606. In some cases, the classification algorithm may be supplemented with a meta-analysis approach, such as a reproducibility analysis. In some cases, the reproducibility analysis selects markers that appear in at least one predictive expression product marker set.

[0153] In some cases, the results of feature selection and classification can be ranked using Bayesian post-analysis methods. For example, microarray data can be extracted, normalized, and summarized using methods known in the art, such as those provided herein. The data can then be subjected to a feature selection step, such as any feature selection method known in the art, such as those provided herein, including but not limited to the feature selection method provided in LIMMA. The data can then be subjected to a classification step, such as any classification method known in the art, such as using any of the algorithms or methods provided herein, including but not limited to using the SVM algorithm or the Random Forest algorithm. The results of the classification algorithm can then be ranked according to a posteriori probability function. For example, posterior probability functions can be derived from surveys of known molecular profiling results, such as published results, to derive prior probabilities from the rates of type I and type II errors of assigning markers to categories (e.g., benign, malignant, normal, ATC, PTC, MTC, FC, FN, FA, FVPTC, CN, HA, HC, LCT, NHP, etc.). These error rates can be calculated based on the reported sample size for each study, using estimated fold change values ​​(e.g., 1.1-fold, 1.2-fold, 1.3-fold, 1.4-fold, 1.5-fold, 1.6-fold, 1.7-fold, 1.8-fold, 1.9-fold, 2-fold, 2.2-fold, 2.4-fold, 2.5-fold, 3-fold, 4-fold, 5-fold, 6-fold, 7-fold, 8-fold, 9-fold, 10-fold, or more). These prior probabilities can then be combined with the molecular profiling datasets of the present invention to estimate the posterior probabilities of differential gene expression. Finally, the posterior probability estimates can be combined with the second dataset of the invention to formulate a final posterior probability of differential expression. Additional methods for deriving and applying posterior probabilities to the analysis of microarray data are known in the art and are described, for example, in Smyth, GK 2004 Stat. Appl. Genet. Mol. Biol. 3: Article 3.In some cases, the posterior probability may be used to rank the markers provided by the classification algorithm. In some cases, the markers may be ranked according to the posterior probability, and those that pass the selected threshold may be selected as markers with differential expression that are indicative of or diagnostic for samples that are, for example, benign, malignant, normal, ATC, PTC, MTC, FC, FN, FA, FVPTC, CN, HA, HC, LCT, or NHP. Exemplary thresholds include prior probabilities of 0.7, 0.75, 0.8, 0.85, 0.9, 0.925, 0.95, 0.975, 0.98, 0.985, 0.99, 0.995, or higher.

[0154] Statistical evaluation of molecular profiling results may provide quantitative values ​​that indicate one or more of the following: likelihood of diagnostic accuracy, likelihood of cancer, disease, or condition, likelihood of a particular cancer, disease, or condition, likelihood of success of a particular therapeutic intervention. Thus, physicians who are likely not trained in genetics or molecular biology do not need to understand the raw data. Rather, the data is presented directly to the physician in the most useful form to guide patient care. Molecular profiling results may be statistically evaluated using a number of methods known in the art, including, but not limited to, the following: Student's t-test, two-tailed t-test, Pearson's rank sum test, hidden Markov model analysis, analysis of qq plots, principal component analysis, one-way ANOVA, two-way ANOVA, LIMMA, etc.

[0155] In some embodiments of the invention, the use of molecular profiling, alone or in combination with cytological analysis, may provide about 85% to about 99% or about 100% accurate diagnosis. In some cases, the molecular profiling business may provide about 85%, 86%, 87%, 88%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 97.5%, 98%, 98.5%, 99%, 99.5%, 99.75%, 99.8%, 99.85%, or 99.9% accurate diagnosis of malignant, benign, or normal through the use of molecular profiling and / or cytology.

[0156] In some cases, accuracy can be determined by following subjects over time to determine the accuracy of initial diagnosis.In other cases, accuracy can be established deterministically or using statistical methods.For example, receiver operating characteristic (ROC) analysis can be used to determine the optimal assay parameters to achieve a certain level of accuracy, specificity, positive predictive value, negative predictive value, and / or false detection rate.Methods for using ROC analysis in cancer diagnosis are known in the art and described, for example, in US Patent Application No. 2006 / 019615, the entirety of which is incorporated herein by reference.

[0157] In some embodiments of the present invention, gene expression products and the compositions of nucleotides encoding such products that are determined to show the greatest difference in expression levels or the greatest difference in alternative splicing between benign and normal, benign and malignant, or malignant and normal can be selected for use as molecular profiling reagents of the present invention. Such gene expression products can be particularly useful by providing a wider dynamic range, a larger signal-to-noise ratio, improved diagnostic power, a lower probability of false positives or false negatives, or a larger statistical confidence level than other methods known or used in the art.

[0158] In other embodiments of the present invention, the use of molecular profiling, alone or in combination with cytological analysis, can reduce the number of samples that are scored as non-diagnostic by about 100%, 99%, 95%, 90%, 80%, 75%, 70%, 65%, or about 60% compared to the use of standard cytological techniques known in the art.In some cases, the method of the present invention can reduce the number of samples that are scored as intermediate or suspicious by about 100%, 99%, 98%, 97%, 95%, 90%, 85%, 80%, 75%, 70%, 65%, or about 60% compared to standard cytological methods used in the art.

[0159] In some cases, the results of the molecular profiling assay are entered into a database for access by an agent or distributor of the molecular profiling business, an individual, a health care provider, or an insurance provider. In some cases, the assay results include an interpretation or diagnosis by an agent, distributor, or consultant of the business, such as a health care professional. In other cases, computer or algorithmic analysis of the data is provided automatically. In some cases, the molecular profiling business may bill an individual, an insurance provider, a health care provider, a researcher, or a government entity for one or more of the following: the molecular profiling assay performed, advisory services, data analysis, reporting of results, or database access.

[0160] In some embodiments of the present invention, the results of molecular profiling are presented as a report on a computer screen or as a paper record.In some cases, the report may include information such as, but not limited to, one or more of the following: the number of differentially expressed genes, suitability of the initial sample, the number of genes that show differential alternative splicing, diagnosis, statistical confidence of diagnosis, possibility of cancer or malignancy, and required treatment.

[0161] (iv) Categorization of samples based on molecular profiling results The results of molecular profiling can be classified as one of the following: benign (absence of cancer, disease, or condition), malignant (positive diagnosis for cancer, disease, or condition), or non-diagnostic (providing insufficient information regarding the presence or absence of cancer, disease, or condition). In some cases, the diagnostic result may further classify the type of cancer, disease, or condition. In other cases, the diagnostic result may indicate certain molecular pathways involved in the cancer, disease, or condition, or a certain grade or stage of a particular cancer, disease, or condition. In still other cases, the diagnostic result may inform appropriate therapeutic intervention, such as a particular drug regimen, such as a kinase inhibitor like Gleevec or any drug known in the art, or a surgical intervention, such as a thyroidectomy or hemithyroidectomy.

[0162] In some embodiments of the present invention, the results are classified using a trained algorithm. The trained algorithm of the present invention includes algorithms developed using a reference set of known malignant, benign and normal samples, including but not limited to the samples listed in Figure 1. Algorithms suitable for categorizing samples include but are not limited to k-nearest neighbor algorithms, concept vector algorithms, naive Bayesian algorithms, neural network algorithms, hidden Markov model algorithms, genetic algorithms, and mutual information feature selection algorithms, or any combination thereof. In some cases, the trained algorithm of the present invention may incorporate data other than DNA polymorphism data, such as but not limited to gene expression or alternative splicing data, sequencing data, scoring or diagnosis by the cytologist or pathologist of the present invention, information provided by the preliminary classification algorithm of the present invention, or information about the medical history of the subject of the present invention.

[0163] (v) Monitoring of a subject or therapeutic intervention via molecular profiling In some embodiments, subjects can be monitored using the methods and compositions of the present invention. For example, subjects can be diagnosed with cancer or genetic disorders. This initial diagnosis may or may not include the use of molecular profiling. Subjects may be prescribed therapeutic intervention, such as thyroidectomy for subjects suspected of having thyroid cancer. The results of therapeutic intervention can be continuously monitored by molecular profiling to detect efficacy of therapeutic intervention. In another example, subjects can be diagnosed with benign tumors or precancerous lesions or nodules, and tumors, nodules, or lesions can be continuously monitored by molecular profiling to detect changes in the state of tumors or lesions.

[0164] Molecular profiling may be used to confirm the potential efficacy of a particular therapeutic intervention before administering it to a subject. For example, the subject may be diagnosed with cancer. Molecular profiling may show upregulation of gene expression products known to be involved in cancer malignancy, such as RAS oncogenes. Tumor samples can be obtained and cultured in vitro using methods known in the art. Then, the application of various inhibitors of aberrantly activated or aberrantly regulated pathways, or drugs known to inhibit the activity of pathways, can be tested on tumor cell lines for growth inhibition. Molecular profiling may be used, for example, to monitor the effect of these inhibitors on downstream targets of related pathways.

[0165] (vi) Molecular profiling as a research tool In some embodiments, molecular profiling may be used as a research tool to identify new markers for the diagnosis of suspected tumors; to monitor the effects of drugs or drug candidates on biological samples such as tumor cells, cell lines, tissues, or organisms; or to discover new pathways for carcinogenesis and / or tumor suppression.

[0166] (vii) Molecular profiling-based biomarkers Thyroid genes are described according to the following groups: (1) benign or malignant, (2) alternative gene splicing, (3) KEGG pathways, (4) normal thyroid, (5) thyroid subtypes, (6) gene ontology, and (7) biomarkers of metastasis from nonthyroid organs to the thyroid. The methods and compositions of the invention can have genes selected from one or more of the groups listed above, and / or 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50 or more subgroups (e.g., one or more different KEGG pathways) from any of the groups listed above, and can have 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50 or more gene expression products from each group, in any combination. In some embodiments, the set of combined genes provides a specificity or sensitivity of greater than 70%, 75%, 80%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 99.5%, or a positive or negative predictive value of at least 95%, 95.5%, 96%, 96.5%, 97%, 97.5%, 98%, 98.5%, 99%, 99.5% or more.

[0167] In some embodiments, extracellular matrix genes, adhesion genes, focal adhesion genes, and tight junction genes are used as biomarkers for thyroid cancer. In some embodiments, the signal transduction pathway is selected from one of the following three pathways: adhesion pathway, focal adhesion pathway, and tight junction pathway. In some embodiments, at least one gene is selected from one of the three pathways. In some embodiments, at least one gene is selected from each of the three pathways. In some embodiments, at least one gene is selected from two of the three pathways. In some embodiments, at least one gene involved in all three pathways is selected. In one example, a set of genes involved in adhesion pathway, focal adhesion pathway, and tight junction pathway is selected as a marker for the diagnosis of cancer, such as thyroid cancer.

[0168] The follicular cells lining thyroid follicles are highly polarized and structurally organized, requiring distinct roles for luminal and apical cell membranes. In some embodiments, cytoskeleton genes, plasma membrane genes, and extracellular space genes are used as biomarkers for thyroid cancer. In some embodiments, genes overlapping all four pathways, namely, ECM pathway, focal adhesion pathway, adhesion pathway, and tight junction pathway, are used as biomarkers for thyroid cancer. In one example, the present invention provides a benign vs. malignant group (n=948) as a thyroid classification gene list. This list was categorized according to alternative splicing, KEGG pathway, and gene ontology. KEGG pathways are further described in Table 1.

[0169] In some embodiments, the present invention provides methods of diagnosing cancer, including gene expression products from one or more signaling pathways, including, but not limited to, the following signaling pathways: acute myeloid leukemia signaling, type 2 somatostatin receptor signaling, cAMP-mediated signaling, cell cycle and DNA damage checkpoint signaling, G protein-coupled receptor signaling, integrin signaling, melanoma cell signaling, relaxin signaling, and thyroid cancer signaling. The methods and compositions of the present invention can have genes selected from 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50, or more signaling pathways, and can have 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50, or more gene expression products from each signaling pathway in any combination. In some embodiments, the set of combined genes provides a specificity or sensitivity of greater than 70%, 75%, 80%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 99.5%, or a positive or negative predictive value of at least 95%, 95.5%, 96%, 96.5%, 97%, 97.5%, 98%, 98.5%, 99%, 99.5% or more.

[0170] In some embodiments, the present invention provides methods of diagnosing cancer comprising gene expression products from one or more ontology groups, including, but not limited to, the following ontology groups: cell senescence, cell cortex, cell cycle, cell death / apoptosis, cell differentiation, cell division, cell junctions, cell migration, cell morphogenesis, cell motility, cell processes, cell proliferation, cell recognition, cell body, cell surface, cell surface-bound receptor signaling, cell adhesion, transcription, immune response, or inflammation. The methods and compositions of the invention can have genes selected from 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50, or more ontology groups and can have 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50, or more gene expression products from each ontology group in any combination. In some embodiments, the set of combined genes provides a specificity or sensitivity of greater than 70%, 75%, 80%, 85%, 86%, 87%, 88%, 89%, 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, 99%, or 99.5%, or a positive or negative predictive value of at least 95%, 95.5%, 96%, 96.5%, 97%, 97.5%, 98%, 98.5%, 99%, 99.5% or more.

[0171] Table 1. Genes involved in KEGG pathways TIFF0007602319000001.tif142166

[0172] The top biomarkers for benign versus malignant thyroid disease (n=948) are listed in List 1 below.

[0173] Listing 1 TIFF0007602319000002.tif235167TIFF0007602319000003.tif249167TIFF0007602319000004.tif142167

[0174] Genes that undergo alternative splicing (n=283) are listed in List 2 below.

[0175] Listing 2 TIFF0007602319000005.tif189167

[0176] The genes involved in the KEGG pathways are listed below in Table 6. There are 18 pathways, containing a total of n=109 unique genes.

[0177] (Table 6) TIFF0007602319000006.tif120163TIFF0007602319000007.tif248164TIFF0007602319000008.tif128166

[0178] The top genes (n=55) separating benign and malignant thyroid (combined) from normal thyroid are listed in List 3 below.

[0179] Listing 3 TIFF0007602319000009.tif41165

[0180] Thyroid surgical pathology subtypes (n=873) are listed below.

[0181] (i) List 4: FA subtype (n=243) TIFF0007602319000010.tif161166

[0182] (ii) List 5: FC subtype (n=102) TIFF0007602319000011.tif75165

[0183] (iii) List 6: LCT subtype (n=140) TIFF0007602319000012.tif94166

[0184] (iv) List 7: FVPTC subtype (n=182) TIFF0007602319000013.tif127167

[0185] (v) List 8: PTC subtype (n=604) TIFF0007602319000014.tif61166TIFF0007602319000015.tif248169TIFF0007602319000016.tif88167

[0186] (vi) List 9: NHP subtypes (n=653) TIFF0007602319000017.tif157167TIFF0007602319000018.tif251166TIFF0007602319000019.tif27166

[0187] (vii) List 10: MTC subtype (n=48) TIFF0007602319000020.tif36164

[0188] (viii) List 11: HC subtype (n=65) TIFF0007602319000021.tif47162

[0189] (ix) List 12: HA subtype (n=24) TIFF0007602319000022.tif19165

[0190] (x) List 13: ATC subtype (n=12) TIFF0007602319000023.tif14164

[0191] The top 948 dominant gene ontologies for thyroid biomarkers are listed below.

[0192] List 14: Angiogenesis (n=23) TIFF0007602319000024.tif20165

[0193] List 15: Apoptosis (n=43) TIFF0007602319000025.tif35165

[0194] List 16: Cell cycle, transcription factors (n=184) TIFF0007602319000026.tif115167

[0195] List 17: Cell membrane (n=410) TIFF0007602319000027.tif255161

[0196] List 18: Rare membrane components (n=55) TIFF0007602319000028.tif36167

[0197] List 19: Cell-cell adhesion (n=85) TIFF0007602319000029.tif55163

[0198] List 20: Cell apical membrane (n=15) TIFF0007602319000030.tif15165

[0199] List 21: Cell basolateral membrane, cell lateral membrane (n=28) TIFF0007602319000031.tif21164

[0200] List 22: Integrins (n=14) TIFF0007602319000032.tif14155

[0201] List 23: Cell-cell junctions (n=40) TIFF0007602319000033.tif28163

[0202] List 24: Cell surface (n=17) TIFF0007602319000034.tif14164

[0203] List 25: Extracellular space (n=156) TIFF0007602319000035.tif94167

[0204] List 26: Cytoskeleton (n=94) TIFF0007602319000036.tif62165

[0205] In some embodiments, the present invention provides a method of classifying cancer comprising the steps of: obtaining a biological sample comprising gene expression products; measuring expression levels for one or more gene expression products of the biological sample; and identifying the biological sample as cancerous if the gene expression levels indicate the presence of thyroid cancer in the biological sample. This can be done by correlating the gene expression levels with the presence of thyroid cancer in the biological sample. In one embodiment, the gene expression products are selected from one or more genes listed in Table 2. In some embodiments, the method further comprises identifying the biological sample as positive for cancer metastasizing to the thyroid from a non-thyroid organ if there is a difference in gene expression levels between the biological sample and a control sample at a specified confidence level.

[0206] Biomarkers involved in metastasis from non-thyroid organs to thyroid are provided. Such metastatic cancers that metastasize to thyroid and can be diagnosed using the method of the present invention include, but are not limited to, metastatic parathyroid carcinoma, metastatic melanoma, metastatic renal carcinoma, metastatic breast carcinoma, and metastatic B-cell lymphoma. Exemplary biomarkers that can be used by the method of the present invention to diagnose metastasis to thyroid are listed in Table 2.

[0207] (Table 2) Biomarkers associated with thyroid metastasis TIFF0007602319000037.tif101170TIFF0007602319000038.tif242170TIFF0007602319000039.tif240170TIFF0007602319000040.tif242170

[0208] (viii) Classification error rate In some embodiments, the top thyroid biomarkers (948 genes) are subdivided into bins (50 TCIDs each) to establish the minimum number of genes required to achieve an overall classification error rate of less than 4% (Figure 1). The initial TCIDs used for classification correspond to Affymetrix Human Exon 1.0ST microarray chips, each of which may map to multiple genes or no genes (Affymetrix annotation file: HuEx-1_0-st-v2.na29.hg18.transcript.csv). If a gene does not map to a TCID, the biomarker is designated as TCID-######.

[0209] List 27: Error rate range 1 (TCID1-50 (n=50), gene symbol, n=58) TIFF0007602319000041.tif41166

[0210] List 28: Error rate range 2 (TCID51-100 (n=50), gene symbol, n=59) TIFF0007602319000042.tif41161

[0211] List 29: Error rate range 3 (TCID 101-150 (n=50), gene symbol, n=52) TIFF0007602319000043.tif35169

[0212] List 30: Error rate range 4 (TCID 151-200 (n=50), gene symbol, n=58) TIFF0007602319000044.tif40161

[0213] List 31: Error rate range 5 (TCID 201-250 (n=50), gene symbol, n=53) TIFF0007602319000045.tif34160

[0214] IX. Composition (i) Gene Expression Products and Splice Variants of the Invention Molecular profiling can also include assays of the present disclosure, including, but not limited to, assays for one or more of the following: protein, protein expression product, DNA, DNA polymorphisms, RNA, RNA expression product, RNA expression product levels, or RNA expression product splice variants of the genes provided in Figures 2-6, 9-13, 16, or 17. In some cases, the methods of the present invention can include assays for about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50, 60, 70, 80, 90, 100, 120, 140, 160, 180, 200, 240, 280, 300, 350, 40 ... Molecular profiling of 0, 450, 500, 600, 700, 800, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 5000 or more DNA polymorphisms, expression product markers, and / or alternative splice variant markers provides improved cancer diagnosis.

[0215] In one embodiment, the molecular profiling comprises microarray hybridization performed to measure gene expression product levels for one or more genes selected from Figures 2, 6, 9-13, 16, or 17. In some cases, the gene expression product levels of one or more genes from one group are compared to the gene expression product levels of one or more genes from another group. By way of example only and not by way of limitation, the expression level of the gene TPO may be compared to the expression level of the gene GAPDH. In another embodiment, gene expression levels are measured for one or more genes involved in one or more of the following metabolic or signaling pathways: thyroid hormone production and / or release, protein kinase signaling pathway, lipid kinase signaling pathway, and cyclin. In some cases, the methods of the invention provide for the analysis of gene expression product levels and / or alternative exon usage of at least one gene of 1, 2, 3, 4, 5, 6, 7, 9, 10, 11, 12, 13, 14, or 15 or more different metabolic or signaling pathways.

[0216] (ii) Composition of the Present Invention Compositions of the disclosure are also provided that include one or more of the following: nucleotides (e.g., DNA or RNA) corresponding to a gene or a portion of a gene provided in Figures 2-6, 9-13, 16, or 17, and nucleotides (e.g., DNA or RNA) corresponding to a gene complement or a portion of a gene complement provided in Figures 2-6, 9-13, 16, or 17. Nucleotides of the invention can be at least about 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 100, 150, 200, 250, 300, 350, or about 400 or 500 nucleotides in length. In some embodiments of the invention, the nucleotides may be natural or artificial derivatives of ribonucleic acid or deoxyribonucleic acid, including, but not limited to, peptide nucleic acid, pyranosyl RNA, nucleosides, methylated nucleic acid, PEGylated nucleic acid, cyclic nucleotides, and chemically modified nucleotides. In some of the compositions of the invention, the nucleotides of the invention are chemically modified to include a detectable label. In some embodiments of the invention, the biological sample is chemically modified to include a label.

[0217] Further compositions of the disclosure include oligonucleotides for detecting (i.e., measuring) the expression products of the genes and their complements provided in Figures 2-6, 9-13, 16, or 17. Further compositions of the disclosure include oligonucleotides for detecting (i.e., measuring) the expression products of polymorphic alleles of the genes and their complements provided in Figures 2-6, 9-13, 16, or 17. Such polymorphic alleles include, but are not limited to, splice site variants, single nucleotide polymorphisms, repeat number polymorphisms, insertions, deletions, and homologs. In some cases, the variant alleles are about 99.9% to about 70% identical to the genes listed in Figure 6, e.g., about 99.75%, 99.5%, 99.25%, 99%, 97.5%, 95%, 92.5%, 90%, 85%, 80%, 75%, about 70% identical. In some cases, the variant allele differs from the gene provided in Figures 2-6, 9-13, 16, or 17 by about 1 to about 500 nucleotides, e.g., by about 1, 2, 3, 5, 7, 10, 15, 20, 25, 30, 35, 50, 75, 100, 150, 200, 250, 300, or about 400 nucleotides.

[0218] In some embodiments, the composition of the present invention can be specifically selected from the top gene products that are differentially expressed between benign and malignant samples, or the top gene products that are differentially spliced ​​between benign and malignant samples, or the top gene products that are differentially expressed between normal and benign or malignant samples, or the top gene products that are differentially spliced ​​between normal and benign or malignant samples.In some cases, the top gene products that are differentially expressed can be selected from Figure 2 and / or Figure 4.In some cases, the top gene products that are differentially spliced ​​can be selected from Figure 3 and / or Figure 5.

[0219] IX. Business Methods As described herein, the term customer or potential customer refers to an individual or entity that may use the methods or services of a molecular profiling business. Potential customers of the molecular profiling methods and services described herein include, for example, patients, subjects, physicians, cytology laboratories, health care providers, researchers, insurance companies, government entities such as Medicaid, employers, or any other entity interested in achieving a more economical or efficient system for the diagnosis, monitoring, and treatment of cancer.

[0220] Such individuals can use the molecular profiling results, for example, to selectively require expensive drugs or therapeutic interventions to those patients most likely to benefit from the drug or intervention, or to identify individuals who may not benefit, or may be harmed, by unnecessary use of a drug or other therapeutic intervention.

[0221] (i) Method of Sale The services of the molecular profiling business of the present invention may be sold to individuals concerned about their health, physicians, or other medical professionals, for example, as a way to enhance diagnosis and care; to cytology laboratories, for example, as a service to provide enhanced diagnosis to their clients; and to health care providers, insurance companies, and government entities, for example, as a way to lower costs by eliminating unwarranted therapeutic interventions. Methods of selling to prospective clients further include the sale of database access for researchers and physicians seeking to find new correlations between gene expression products and diseases or conditions.

[0222] Methods of marketing may include the use of print, radio, television, or internet-based advertising to potential customers. It is also possible to market to potential customers through a specific medium, for example, endocrinologists can market by placing advertisements in trade and medical journals, including but not limited to Journal of the American Medical Association, Physicians Practice, American Medical News, Consultant, Medical Economics, Physician's Money Digest, American Family Physician, Monthly Prescribing Reference, Physicians' Travel and Meeting Guide, Patient Care, Cortlandt Forum, Internal Medicine News, Hospital Physician, Family Practice Management, Internal Medicine World Report, Women's Health in Primary Care, Family Practice News, Physician's Weekly, Health Monitor, The Endocrinologist, Journal of Endocrinology, The Open Endocrinology Journal, and The Journal of Molecular Endocrinology. Sales can also take the form of collaboration with medical professionals to conduct experiments using the methods and services of the invention and, in some cases, publish results or seek funding for further research. In some cases, methods of sales may include the use of databases of physicians or medical professionals, such as, for example, the American Medical Association (AMA) database, to determine contact information.

[0223] In one embodiment, the method of sale involves collaboration with cytology testing laboratories to provide molecular profiling services to customers whose samples cannot be definitively diagnosed using routine methods.

[0224] (ii) Computer-based business methods A molecular profiling business may use one or more computers in the methods of the invention, such as computer 800 as illustrated in FIG. 22. Computer 800 may be used for managing customer and sample information, such as sample or customer tracking, database management, analysis of molecular profiling data, analysis of cytological data, data storage, billing, sales, reporting of results, or storage of results. The computer may include a monitor 807 or other graphical interface for displaying data, results, billing information, sales information (e.g., demographics), customer information, or sample information. The computer may also include means 816, 815 for inputting data or information. The computer may include a processing unit 801, and fixed media 803, or removable media 811, or a combination thereof. A user physically close to the computer may access the computer, for example, via a keyboard and / or mouse, or a user 822 not necessarily close to the physical computer may access the computer through a communication medium 805, such as a modem, an Internet connection, a telephone connection, or a wired or wireless communication carrier wave. In some cases, the computer may be connected to a server 809 or other communication device for relaying information from the user to the computer or from the computer to the user. In some cases, the user may store data or information obtained from the computer through the communication medium 805 on a medium such as a removable medium 812. It is envisioned that data related to the present invention may be transmitted through such a network or connection for receipt and / or viewing by a person. The recipient may be, but is not limited to, an individual, a health care provider, or a health care administrator. In one embodiment, the computer readable medium includes a medium suitable for transmission of results of an analysis of a biological sample, such as an exosome biosignature. The medium may include results related to an exosome biosignature of a subject derived using the methods described herein.

[0225] A molecular profiling business may enter sample information into a database for one or more of the following purposes: inventory tracking, assay result tracking, order tracking, customer management, customer service, billing, and sales. Sample information may include, but is not limited to: customer name, unique customer ID, medical professional associated with the customer, assay method required, assay results, validation status, validation tests required, individual medical history, preliminary diagnosis, suspected diagnosis, sample history, insurance provider, medical provider, third party testing facility, or any information suitable for storage in a database. Sample history may include, but is not limited to: sample age, sample type, method of acquisition, method of storage, or method of transport.

[0226] The database may be accessible by customers, health care professionals, insurance providers, third parties, or any individual or entity to which the molecular profiling business grants access. Database access may take the form of electronic communication, such as computer or telephone. The database may be accessed through an intermediary, such as a customer service representative, business representative, consultant, independent testing facility, or health care professional. The availability or extent of database access or sample information, such as assay results, may change upon payment for products and services given or to be given. The extent of database access or sample information may be limited according to accepted or legal requirements for patient or customer confidentiality. The molecular profiling business may charge an individual, insurance provider, health care provider, or government entity for one or more of the following: sample receipt, sample storage, sample preparation, cytological testing, molecular profiling, entering and updating sample information into the database, or database access.

[0227] (iii) Business Flow FIG. 18a is a flow chart illustrating one manner in which a sample may be processed by a molecular profiling business. For example, a sample of thyroid cells may be obtained by an endocrinologist, perhaps via fine needle aspiration 100. The sample is subjected to routine cytological staining 125. The routine cytological staining provides four different possible preliminary diagnoses: non-diagnostic 105, benign 110, uncertain or suspicious 115, or malignant 120. The molecular profiling business may then analyze gene expression product levels as described herein 130. Analysis of gene expression product levels, molecular profiling, may result in a definitive diagnosis of malignant 140 or benign 135. In some cases, only a subset of samples are analyzed by molecular profiling, such as samples that provide an uncertain non-diagnostic result in routine cytological investigation. Alternative embodiments in which samples may be processed by the methods of the present invention are provided in FIGS. 18b and 21.

[0228] In some cases, the results of molecular profiling confirm the results of routine cytological testing. In other cases, the results of molecular profiling are different. In such cases, the sample may be further tested, the data may be re-examined, or the results of molecular profiling or the results of cytological assays may be taken as the correct diagnosis. Benign diagnosis may also include diseases or conditions that are not malignant cancers but may require further monitoring or treatment. Similarly, malignant diagnosis may further include the diagnosis of a specific type of cancer, or a specific metabolic or signaling pathway involved in a disease or condition. The diagnosis may require treatment or therapeutic intervention such as radioactive iodine ablation, surgery, thyroidectomy; or further monitoring.

[0229] XI. Kits A molecular profiling business may provide a kit for obtaining a suitable sample. A kit 203 as illustrated in FIG. 19 may include a container 202, a means for obtaining a sample 200, a reagent for storing the sample 205, and instructions for use of the kit. In another embodiment, the kit further includes reagents and materials for performing a molecular profiling analysis. In some cases, the reagents and materials include a computer program for analyzing data generated by the molecular profiling method. In yet other cases, the kit contains a means by which the biological sample is stored and transported to a testing institution, such as the molecular profiling business or a third-party testing facility.

[0230] Molecular profiling businesses can also provide kits for performing molecular profiling. The kits can include means for extracting proteins or nucleic acids, including all necessary buffers and reagents; and means for analyzing protein or nucleic acid levels, including controls and reagents. The kits can also include software, or a license to obtain and use software, for analysis of data provided using the methods and compositions of the invention. EXAMPLES

[0231] Example 1: Gene Expression Product Analysis of Thyroid Samples Seventy-five thyroid samples were investigated for gene expression analysis using Affymetrix Human Exon 10ST arrays according to the manufacturer's instructions to identify genes that show significant differential expression and / or alternative splicing between malignant, benign, and normal samples. Three groups were compared and classified according to the pathological surgical diagnosis of the tissue: benign (n=29), malignant (n=37), and normal (n=9). Samples were prepared from surgical thyroid tissue, snap frozen, and then RNA was prepared by standard methods. The designations and pathological classifications of the 75 samples are shown in Figure 1.

[0232] Microarray analysis was performed with XRAY version 2.69 (Biotique Systems Inc.). Input files were normalized by full quantile normalization (Irizarry et al. Biostatistics 2003 April 4 (2): 249-64). For each input array and each probe expression value, the array-ith percentile probe value was replaced with the average value of all array-ith percentile points. A total of 6,553,590 probes were processed in the analysis. Probes with GC numbers less than 6 and probes with GC numbers greater than 17 were excluded from the analysis. Expression scores for each probe set were derived through application of a median polish (exon RMA) to the probe scores in all input hybridizations, and probe sets with less than 3 probes (passing all of the tests defined above) were excluded from further analysis. Only "core" probe sets were analyzed, which correspond to probe sets that matched entries in the high-quality databases RefSeq and Ensembl. Non-expressed and invariant probes were also removed from both gene-level and probe set-level analyses. One-way ANOVA analysis was used to examine gene expression at the probe set level between malignant and benign groups.

[0233] The top 100 differentially expressed genes (i.e., genes that showed the greatest differential expression) by gene-level analysis were obtained from the dataset in which benign, malignant, and normal thyroid samples were compared. Markers were selected based on statistical significance after Benjamini and Hochberg correction for false discovery rate (FDR). An FDR filter value of p<0.01 was used, followed by ranking by the absolute fold change (>1.9) calculated for each marker as the highest differential gene expression value in any group (benign, malignant, or normal) divided by the lowest differential expression in the remaining two groups. The results of this analysis are shown in Figure 2. This table lists three sets of calculated fold changes for a given marker that allow comparison between groups. The malignant / benign, malignant / normal, and benign / normal fold changes were all calculated by dividing the expression of one group by the expression of another group.

[0234] The top 100 genes undergoing alternative splicing were obtained from a dataset in which benign, malignant and normal thyroid samples were compared. Markers were selected based on statistical significance after Benjamini and Hochberg correction for false discovery rate (FDR). An FDR filter value of p<0.01 was used and markers were ranked starting from the lowest p-value. The threshold for listing the values ​​by the software used was p<1.0E-301, and numbers with smaller p-values ​​were automatically assigned a value of 0.00E+00. The results of this analysis are shown in Figure 3. All of the markers shown are highly significant for alternative exon splicing.

[0235] By probe set level analysis, the top 100 differentially expressed genes in thyroid samples from Figure 1 were obtained from the dataset in which benign and malignant samples were analyzed. Markers were selected based on significance after Benjamini and Hochberg correction for false discovery rate (FDR). Markers were selected based on significance after Benjamini and Hochberg correction for false discovery rate (FDR). An FDR filter value of p<0.01 was used, followed by ranking by absolute fold change (>2.0) calculated for each marker as malignant expression divided by benign expression. The results of this analysis are shown in Figure 4.

[0236] The top 100 statistically significant diagnostic markers as measured by gene-level analysis of thyroid samples shown in Figure 1 were also compiled. Data from comparisons of benign, malignant, and normal datasets, and between benign and malignant datasets were used. Markers were selected based on significance after Benjamini and Hochberg correction for false discovery rate (FDR). An FDR filter value of p<0.01 was used, followed by ranking by absolute fold change (>1.6) calculated for each marker as the highest differential expression value in any group (benign, malignant, or normal) divided by the lowest differential expression in the remaining two groups. Fold changes for malignant / benign, malignant / normal, and benign / normal were all calculated similarly by dividing the expression in one group by the expression in another group. The results of this analysis are shown in Figure 5.

[0237] A full list of 4918 genes identified as statistically significantly differentially expressed, differentially spliced, or both between benign and malignant samples, between benign and normal samples, or between malignant and normal samples, either at the probe set level or gene level, was also compiled. Markers were selected based on statistical significance after Benjamini and Hochberg correction for false discovery rate (FDR), using an FDR filter value of p<0.01. The results are shown in Figure 6.

[0238] Example 2: Gene Expression Product Analysis of Thyroid Tissue Samples To identify genes whose RNA expression levels differ significantly between benign and malignant samples, a total of 205 thyroid tissue samples (Figure 7) are investigated using the Affymetrix HumanExon10ST array chip. Samples are classified according to postoperative thyroid pathology. Samples showing follicular adenoma (FA), lymphocytic thyroiditis (LCT), or nodular hyperplasia (NHP) are classified as benign; samples showing Hurthle cell carcinoma (HC), follicular carcinoma (FC), follicular type papillary thyroid carcinoma (FVPTC), papillary thyroid carcinoma (PTC), medullary thyroid carcinoma (MTC), or anaplastic carcinoma (ATC) are classified as malignant.

[0239] Affymetrix software is used to extract, normalize, and summarize the intensity data from approximately 6.5 million probes. Approximately 280,000 core probe sets are then used in feature selection and classification. The models used are LIMMA for feature selection, and Random Forest and Support Vector Machine (SVM) for classification. Random data subsets are used to iteratively train, classify, and cross-validate. Top features are identified in two separate analyses (malignant vs. benign, and MTC vs. others) using the above classification engine.

[0240] Markers are selected based on significance after Benjamini and Hochberg correction for false detection rate (FDR). An FDR filter of p<0.05 is used.

[0241] Comparison of malignant and benign thyroid tissue samples finds 413 markers that are diagnostic for thyroid diseases or conditions. The top 100 markers are listed in FIG.

[0242] Comparison of MTC thyroid tissue samples with other (i.e., non-MTC) thyroid tissue samples finds 671 markers that are diagnostic for thyroid diseases or conditions. The top 100 markers are listed in FIG.

[0243] Example 3: Meta-analysis of gene expression product data from thyroid samples Surgical thyroid tissue samples (Figure 7) and thyroid samples obtained via fine needle aspiration (Figure 8) are identified as benign or malignant by pathological examination and then examined by hybridization to Affymetrix HumanExon10ST arrays. A meta-analysis approach is used that allows the identification of genes with reproducible features in each classification. Affymetrix software is used to extract, normalize, and summarize intensity data from approximately 6.5 million probes. Approximately 280,000 probe sets are used for feature selection and classification. LIMMA is used for feature selection. Classification is performed by random forest and SVM methods. Markers that repeatedly appear in training, classification, and cross-validation iterations of surgical and fine needle aspirate samples are identified and ranked. The top ranked features that appear for both surgical and fine needle aspirate data are used to create a joint set of core features. Markers with non-zero reproducibility scores are selected as significant. A total of 102 markers were found to be significant and are listed in FIG.

[0244] Example 4: Bayesian analysis of gene expression product data from thyroid samples To identify genes that distinguish benign from malignant nodules in the human thyroid gland, two groups of well-characterized samples are compared. Samples are obtained from surgical thyroid tissue (tissue; n=205, FIG. 7) or fine needle aspirates (FNA; n=74, FIG. 8) and interrogated by hybridization to HumanExon10ST microarrays. The pathology label for each distinct thyroid subtype is coded as either benign (B) or malignant (M). A total of 499 markers are identified that show distinct differential expression between benign and malignant samples.

[0245] Affymetrix software is used to extract, normalize, and summarize intensity data from approximately 6.5 million probes. Approximately 280,000 core probe sets are then used in feature selection and classification of approximately 22,000 genes. The models used are LIMMA (for feature selection) and SVM (for classification), respectively.

[0246] Next, we previously published that molecular profiling studies are examined to derive the rates of type I and type II errors of assigning genes to "benign" or "malignant" categories. Error rates are calculated based on the sample size reported in each specific published study, with an estimated fold change of 2. Finally, these prior probabilities are combined with the output of the tissue dataset to estimate the posterior probability of differential gene expression, and then with the FNA dataset to formulate the final posterior probability of differential expression (Smyth 2004). These posterior probabilities are used to rank genes, and those that exceed a posterior probability threshold of 0.9 are selected. A total of 499 markers are identified as significant, and the top 100 are listed in Figure 12.

[0247] Example 5: Subtype analysis of gene expression product data from thyroid samples To distinguish benign nodules from those with distinct pathologies in the human thyroid gland, well-characterized samples are examined. Hybridization of 205 species to HumanExon10ST microarrays is examined. The pathology labels for each distinct thyroid subtype are used to systematically compare one group with another. A total of 250 mRNA markers are identified that separate the thyroid gland into a wide range of pathological subtypes.

[0248] To identify genes whose mRNA expression differs significantly between distinct thyroid pathology subtypes, a total of 205 thyroid tissue samples are interrogated using the Affymetrix HumanExon10ST array chip (Figure 7). Samples classified according to postoperative thyroid pathology as follicular adenoma (FA, n=22), lymphocytic thyroiditis (LCT, n=39), nodular hyperplasia (NHP, n=24) are all collectively classified as benign (n=85). In contrast, samples classified as Hurthle cell carcinoma (HC, n=27), follicular carcinoma (FC, n=19), follicular type papillary thyroid carcinoma (FVPTC, n=21), papillary thyroid carcinoma (PTC, n=26), medullary thyroid carcinoma (MTC, n=22), and anaplastic carcinoma (ATC, n=5) are all collectively classified as malignant (n=120).

[0249] Affymetrix software is used to extract, normalize, and summarize intensity data from approximately 6.5 million probes. Approximately 280,000 core probe sets are then used in feature selection and classification. A given set of benign subtypes (e.g., NHP) is compared against a pool of all other malignant subtypes (e.g., NHP vs. M), and then the benign subset is compared again against each set of malignant subtypes (NHP vs. FC, NHP vs. PTC, etc.). The model used in the classification engine is LIMMA (for feature selection), and random forest and SVM are used for classification. Training, classification, and cross-validation iterations are performed using random data subsets. A joint core set of genes that separate distinct thyroid subtypes is created.

[0250] Markers are selected based on the set of genes that optimize the classifier after pairwise classification. A total of 251 markers mapping to 250 distinct genes allow the separation of 1 to 3 distinct thyroid subtypes (Figure 13).

[0251] Example 6: Differentially expressed miRNAs identified via Agilent vs microRNA arrays Thyroid samples are hybridized to the Agilent Human v2 microRNA (miRNA) array. This array contains probes for 723 human miRNAs and 76 viral miRNAs, which are targeted using approximately 15,000 probe sets. To identify significant differentially expressed miRNAs, a comparison is performed between benign thyroid samples (B) and malignant thyroid samples (M). All samples are obtained from clinical fine needle aspirates (n=89, FIG. 14).

[0252] Array intensity data is extracted, normalized, and summarized, followed by modeling using a classification engine. Briefly, the model used is LIMMA (for feature selection), and Random Forest and Support Vector Machine (SVM) are used for classification. Training, classification, and cross-validation iterations are performed using random data subsets. Although several miRNAs are differentially expressed in malignant compared to benign (Figure 16), no classifiers that function independently were identified by this approach.

[0253] Example 7 Differentially expressed miRNAs that are diagnostic for thyroid disease Thyroid nodule samples are hybridized to Illumina Human v2 miRNA array. This array contains probes for 1146 human miRNAs. Comparison between benign and malignant thyroid samples is performed to identify significant differentially expressed miRNAs. All samples are obtained from clinical FNA (n=24, FIG. 15).

[0254] Array intensity data is extracted, normalized, and summarized, followed by modeling using a classification engine. Briefly, the models used are LIMMA (for feature selection), and random forest and support vector machine (SVM) for classification. An additional "hot probe" method, which in part incorporates a meta-analysis approach to the algorithm, is added to the classification engine. Training, classification, and cross-validation iterations are performed using random data subsets. The "hot probe" method identifies probes that appear in all loops of the cross-validation, thereby creating a robust reproducible set of features. Markers are selected based on the p-value (P) of the comparison between malignant and benign samples. A total of 145 miRNAs with differential expression identified as diagnostic for benign or malignant thyroid conditions are identified (Figure 17).

[0255] Example 8: Exemplary Device for Molecular Profiling The molecular profiling business of the present invention compiles a list of 4918 genes in FIG. 6 that are differentially expressed, differentially spliced, or both, between benign and malignant samples, between benign and normal samples, or between malignant and normal samples, either at the probe set level or gene level. A subset of the 4918 genes is selected by the molecular profiling business for use in the diagnosis of biological samples. A composition of short (i.e., 12-25 nucleotides in length) oligonucleotides complementary to the subset of 4918 genes selected for use by the molecular profiling business is synthesized by standard methods known in the art and immobilized to a solid support, such as nitrocellulose, glass, polymer, or chip, at a known location on the solid support.

[0256] Example 9: Molecular profiling of biological samples Biological samples are obtained by fine needle aspiration and stored in two aliquots (one for molecular profiling and one for cytological analysis). The aliquot of biological sample for molecular profiling is added to lysis buffer and triturated so that the cells of the biological sample are lysed. The lysis buffer is prepared as follows: for 1 ml of cDNA lysis buffer, the following are mixed on ice: 0.2 ml Moloney Murine Leukemia Virus (MMLV) Reverse Transcriptase 5x (Gibco-BRL), 0.76 ml H20 (RNAse·DNAse free, Specialty Media), 5 μl Nonidet P40 (USB), 10 μl Prime RNase inhibitor (3'5' Incorporated), 10 μl RNAguard (Pharmacia), and 20 μl of a freshly made 24-fold dilution of the stock primer mix. The stock primer mix, kept in aliquots at -20°C, contains 10 μl each of 100 mM dATP, dCTP, dGTP, dTTP solution (12.5 mM final) (Boehringer); 10 μl of 50 OD / ml pd(T)19-24 (Pharmacia); and 30 μl of H20.

[0257] The cellular RNA is then primed with an oligo-dT primer. Reverse transcription is then performed using reverse transcriptase under time- and reagent-limited conditions to facilitate incomplete extension and prepare short cDNAs of about 500 bp to about 1000 bp. PolyA (dATP) and terminal transferase are then used to add a tail containing multiple dATPs to the 5' end of the cDNA.

[0258] The cDNA is then amplified using PCR reagents using a 60 residue long primer with 24(dT) at the 3' end. PCR cycling is performed at 94°C for 1 minute, then 42°C for 2 minutes, then 72°C for 6 minutes, with a 10 second extension time in each cycle. 10 cycles are performed. Additional Taq polymerase is then added and an additional 25 cycles are performed.

[0259] The cDNA is phenol-chloroform extracted and ethanol precipitated, and then half of the sample is frozen at -80°C as a stock to avoid thawing and freezing the entire amount of cDNA during analysis.

[0260] 5 μg of PCR product is combined with 15.5 μl of EF sln (Tris in Qiagen kit PCR purification), 4 μl of Promega's lox One-Phor-All buffer, and 0.5 units of DNaseI. The entire volume is then held at 37° C. for 14 minutes, then at 99° C. for 15 minutes, and then placed on ice for 5 minutes to fragment the PCR product into segments of about 50 bp to about 100 bp in length. The fragments are then end-labeled by combining the entire volume with 1 μl of Biotin-N6-ddATP ("NEN") and 1.5 μl of TdT (Terminal Transferase) (15 units / μl). The entire volume is then held at 37° C. for 1 hour, then at 99° C. for 15 minutes, and then on ice for 5 minutes.

[0261] The labeled, fragmented cDNA is hybridized with the probe set of the invention in 200 microliters of hybridization solution containing 5-10 micrograms of labeled target in 1×MES buffer (0.1 M MES, 1.0 M NaCl, 0.01% Triton X-100, pH 6.7) and 0.1 mg / ml herring sperm DNA. The array used is the Affymetrix Human Exon 10ST array. The array is placed on a rotisserie and rotated at 60 rpm at 45° C. for 16 hours. After hybridization, the array is washed with 6xSSPE-T (0.9M NaCl, 60mM NaH2P04, 6mM EDTA, 0.005% Triton X-100, pH 7.6) on a fluidics station (Affymetrix) at 22°C for 10x2 cycles, then washed with 0.1 MES at 45°C for 30 minutes. The array is then stained with streptavidin-phycoerythrin conjugate (Molecular Probes), followed by washing again with 6xSSPE-T on a fluidics station for 10x2 cycles. To enhance the signal, the array is further stained with anti-streptavidin antibody for 30 minutes, followed by staining again with streptavidin-phycoerythrin conjugate for 15 minutes. After washing with 6xSSPE-T on a fluidics station for 10x2 cycles, the arrays are scanned at 3 micrometer resolution using a modified confocal scanner to measure raw fluorescence intensity values ​​at each location of the array, which correspond to the gene expression level for the sequence at that array location.

[0262] The raw fluorescence intensity values ​​are converted to gene expression product levels, normalized via the RMA method, filtered to remove data that may be considered suspicious, and input to a preliminary classification algorithm that corrects the gene expression product levels for the cell type composition of the biological sample. The corrected gene expression product levels are input to a trained algorithm to classify the biological sample as benign, malignant, or normal. The trained algorithm provides a record of its output, including a diagnosis and a confidence level.

[0263] Example 10: Molecular profiling of thyroid nodules An individual notices a lump in the thyroid gland. The individual visits their family physician. The family physician decides to obtain a sample from the lump and submit it for molecular profiling analysis. The physician uses a kit from a molecular profiling business to obtain the sample via fine needle aspiration, perform validity testing, store the sample in a liquid based cytology solution, and send it to a molecular profiling business. The molecular profiling business splits the sample, one portion for cytological analysis and the other portion for mRNA extraction from the sample, analyzes the quality and suitability of the extracted mRNA sample, and analyzes the expression levels and alternative exon usage of a subset of the genes listed in FIG. 5. In this case, the specific gene expression products to be profiled are determined by the sample type, the physician's preliminary diagnosis, and the molecular profiling company.

[0264] The molecular profiling business analyzes the data and provides a resulting diagnosis to the individual's physician, as illustrated in Figure 20. The results provide: (1) a list of the gene expression products profiled, (2) the results of the profiling (e.g., expression levels normalized to an internal standard such as total mRNA or expression of a well-characterized gene product such as tubulin), (3) predicted gene product expression levels for normal tissue of the matching type, and (4) a diagnosis and recommended treatment for Bob based on the gene product expression levels. The molecular profiling business bills the individual's insurance provider for the products and services rendered.

[0265] Example 11: Molecular profiling as an adjunct to cytological investigations An individual notices a suspicious lump in the thyroid gland. The individual sees their primary care physician, who investigates the individual and refers them to an endocrinologist. The endocrinologist obtains a sample via fine needle aspiration and sends the sample to a cytology laboratory. The cytology laboratory performs routine cytology testing on a portion of the fine needle aspirate, and the results are equivocal (i.e., inconclusive). The cytology laboratory suggests to the endocrinologist that the remaining sample may be suitable for molecular profiling, and the endocrinologist agrees.

[0266] The remaining sample is analyzed using the methods and compositions herein. The results of the molecular profiling analysis suggest a high probability of early stage follicular cell carcinoma. The results further suggest that the molecular profiling analysis combined with the patient's data, including the patient's age and the size of the lump or nodule, requires a thyroidectomy followed by radioactive iodine ablation. An endocrinologist reviews the results and prescribes the recommended treatment.

[0267] The cytology testing laboratory bills the endocrinologist for routine cytology testing and molecular profiling. The endocrinologist passes payment to the cytology testing laboratory and bills the individual's insurance provider for all products and services rendered. The cytology testing laboratory passes payment for molecular profiling to the molecular profiling business, minus a small difference.

[0268] Example 12: Molecular profiling performed by a third party An individual complains to a physician about a suspicious lump in the cervix. The physician investigates the individual, pending results, and prescribes molecular profiling testing and follow-up. The individual visits a clinical trial laboratory, also known as a CLIA laboratory. The CLIA laboratory is licensed to perform the molecular profiling of the present invention. The individual provides a sample at the CLIA laboratory via fine needle aspiration, which is analyzed using the molecular profiling methods and compositions herein. The results of the molecular profiling are communicated electronically to the individual's physician, who is contacted to schedule a follow-up. The physician presents the results of the molecular profiling to the individual and prescribes treatment.

[0269] Example 13: Overlapping genes using different analytical methods The results described in Example 2 were obtained by investigating surgical thyroid nodule tissue samples and comparing gene expression in malignant versus benign (the "Malignant vs. Benign" dataset). This analysis identified 412 genes that were differentially expressed (FDR p<0.05). In a previous study described in Example 1, the inventors described 4918 genes that can distinguish between malignant and benign thyroid nodules ("4918") using (i) a different cohort of samples and (ii) a different analytical method. The "Malignant vs. Benign" tissue discovery dataset shares 231 / 412 genes with the "4918" discovery dataset, and 181 / 412 genes were newly discovered.

[0270] A similar comparison between medullary thyroid carcinoma (MTC) and "other" thyroid subtypes using tissue cohorts pointed out 668 significant genes that were differentially expressed between these two groups (Figure 10). When cross-checked against the previous "4918" gene list, we note that 305 / 668 genes were previously described and 363 / 668 genes were newly discovered.

[0271] We then combined the surgical tissue dataset with the fine needle aspirate (FNA) dataset and again compared malignant versus benign using (i) a "hot probe" approach and (ii) a "Bayesian" approach. Each analysis identified 102 and 498 significant genes, respectively (Tables 11 and 12).

[0272] A total of 1343 significant genes had been identified up to this point. However, a subset analysis was then performed aimed at identifying genes that separate distinct disease subtypes from each other. This analysis used the surgical tissue cohort and yielded 250 significant genes (Figure 13).

[0273] In summary, the five comparisons described herein yield 1437 significant genes. Of these, 636 / 1437 genes are described for the first time to distinguish benign from malignant thyroid pathologies. At present, 568 / 636 have not yet been described as diagnostic markers for thyroid cancer in published scientific literature or patent applications.

[0274] Example 14: Clinical Thyroid FNA method To identify genes whose mRNA expression was significantly different between benign and malignant samples, prospective clinical thyroid FNA samples were interrogated using Affymetrix Human Exon 1.0ST microarrays.

[0275] Affymetrix software was used to extract, normalize, and summarize intensity data from approximately 6.5 million probes. A core probe set of approximately 280,000 was then used in feature selection and classification. The model used was LIMMA (for feature selection), and Random Forest and SVM were used for classification (Smyth 2004; Diaz-Uriarte and Alvarez de Andres 2006). Training, classification, and cross-validation iterations were performed using random data subsets. Top features were identified in three separate analyses using the classification engine described above.

[0276] The annotation of genes and their mapping to transcript cluster IDs (TCIDs) is constantly evolving, but the nucleotide sequences of the probe sets that make up the TCIDs do not change. Furthermore, many significant TCIDs do not map to known genes, but are equally important biomarkers in the classification of thyroid malignancies. Results are described using both the TCIDs and the genes currently mapped to each (Affymetrix annotation file: HuEx-1_0-st-v2.na29.hg18.transcript.csv).

[0277] result The study of differential gene expression in prospectively collected clinical thyroid FNAs required numerous statistical subanalyses. These subanalyses alone led to the discovery of genes that were informative in the classification of thyroid nodules of unknown pathology. However, the confluence of the data sets led to the novel characterization of a thyroid gene panel that can accurately classify thyroid FNAs with improved accuracy over current cytopathology and molecular profiling methods.

[0278] Table 3. Top benign vs. malignant analysis This analysis yielded 175 unique TCIDs that have now been mapped to 198 genes. TIFF0007602319000046.tif56128TIFF0007602319000047.tif246128TIFF0007602319000048.tif246125TIFF0007602319000049.tif247124 TIFF0007602319000050.tif246124TIFF0007602319000051.tif246125TIFF0007602319000052.tif246124TIFF0007602319000053.tif109132

[0279] (Table 4) Top subtype analysis This analysis yielded 599 unique TCIDs that have now been mapped to 681 genes. TIFF0007602319000054.tif121162TIFF0007602319000055.tif246163TIFF0007602319000056.tif24816 3TIFF0007602319000057.tif248163TIFF0007602319000058.tif246161TIFF0007602319000059.tif24716 3TIFF0007602319000060.tif246163TIFF0007602319000061.tif247162TIFF0007602319000062.tif24616 2TIFF0007602319000063.tif246163TIFF0007602319000064.tif247164TIFF0007602319000065.tif24616 4TIFF0007602319000066.tif246162TIFF0007602319000067.tif246163TIFF0007602319000068.tif24616 1TIFF0007602319000069.tif246162TIFF0007602319000070.tif246162TIFF0007602319000071.tif24716 2TIFF0007602319000072.tif247163TIFF0007602319000073.tif246162TIFF0007602319000074.tif24716 4TIFF0007602319000075.tif248163TIFF0007602319000076.tif247162TIFF0007602319000077.tif69162

[0280] Table 5. Trident analysis This benign vs. malignant analysis yielded 210 unique TCIDs that currently map to 237 genes. These genes represent the association of three statistically significant subanalyses (reproducibility, Bayesian, and tissue analyses) using a single dataset. TIFF0007602319000078.tif155167

[0281] TIFF0007602319000079.tif247166TIFF0007602319000080.tif247165TIFF0007602319000081.tif246165TIFF0007602319000082.tif24616 5TIFF0007602319000083.tif246166TIFF0007602319000084.tif246166TIFF0007602319000085.tif247166TIFF0007602319000086.tif85165

[0282] While preferred embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions will occur to those skilled in the art without departing from the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be used in the practice of the invention. The following claims define the scope of the invention, and it is intended that methods and structures falling within the scope of these claims, and their equivalents, be embraced therein.

Claims

1. A method for analyzing a sample of thyroid tissue from a subject having, or suspected of having, metastatic cancer from a non-thyroid organ to the thyroid gland, the method comprising: (a) determining the expression level of one or more gene expression products in said sample of thyroid tissue, said sample being obtained by fine needle aspirate (FNA), said one or more gene expression products corresponding to one or more genes set forth in Table 2, said one or more genes being associated with metastasis to the thyroid from said non-thyroid organ, and said thyroid tissue being identified as indeterminate by a cytological assay; (b) applying a trained algorithm to the expression levels determined in step (a) to classify the sample as benign, suspicious, or positive for the metastatic cancer, wherein the trained algorithm has been trained on a training set of samples that includes known benign thyroid samples, known normal thyroid samples, and known malignant thyroid samples, and wherein the training set of samples includes fine needle aspiration (FNA) samples; A method comprising:

2. 2. The method of claim 1, wherein the training set of samples includes training samples having a condition that is metastatic cancer from a non-thyroid organ to the thyroid gland.

3. 3. The method of claim 2, wherein the condition is metastatic melanoma to the thyroid gland, metastatic renal carcinoma to the thyroid gland, metastatic breast cancer to the thyroid gland, or metastatic B-cell lymphoma to the thyroid gland.

4. 3. The method of claim 2, wherein the condition is metastatic B-cell lymphoma to the thyroid gland.

5. 2. The method of claim 1, wherein the training set of samples comprises at least 500 training samples.

6. 2. The method of claim 1, wherein the training set of samples comprises training samples selected from the group consisting of normal thyroid samples, follicular adenoma samples, parathyroid samples, follicular carcinoma samples, lymphocytic thyroiditis samples, follicular papillary thyroid carcinoma samples, papillary thyroid carcinoma samples, nodular hyperplasia samples, medullary thyroid carcinoma samples, Hurthle cell carcinoma samples, Hurthle cell adenoma samples, and anaplastic thyroid carcinoma samples.

7. 2. The method of claim 1, wherein step (a) comprises determining the expression levels of ten or more gene expression products in the sample of thyroid tissue.

8. The method of claim 1 , wherein the one or more gene expression products comprise ribonucleic acid (RNA).

9. The method of claim 1, wherein the one or more gene expression products comprise a protein.

10. 2. The method of claim 1, wherein the one or more gene expression products comprise microRNA (miRNA), ribosomal RNA (rRNA), transfer ribonucleic acid (tRNA), small interfering RNA (siRNA), or messenger RNA (mRNA).

11. The method of claim 10, wherein the one or more gene expression products include messenger RNA (mRNA).

12. The method of claim 1, wherein the expression level of the one or more gene expression products is measured by microarray, serial analysis of gene expression (SAGE), blotting, RT-PCR or quantitative PCR, ELISA, mass spectrometry, proteomic techniques, or immunohistochemistry.

13. The method of claim 12, wherein the expression level of the one or more gene expression products is measured by RT-PCR.

Images