Pan-cancer platinum response predictors

Abstract

Comprehensive molecular profiling provides abundant data about the molecular state of patient samples.Such data can be compared with the patient's response to treatment to identify biomarker signatures that predict response or non-response to such treatment.This approach has been applied to identify biomarker signatures that correlate with the response of cancer patients to platinum-based chemotherapy.This paper describes data structures, data processing and machine learning models for predicting the probability of benefit from treatment for a disease or disorder of a subject with a specific set of biomarkers, and the exemplary application of such models to precision medicine, for example, to a method for selecting treatment, including platinum therapy, based on molecular profile. TIFF2023504270000037.tif122159

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of U.S. Provisional Patent Application Nos. 62 / 965,601, filed January 24, 2020; 63 / 005,105, filed April 3, 2020; and 63 / 012,740, filed April 20, 2020. In addition, this application is a continuation of and claims the benefit of International Patent Application No. PCT / US2019 / 064078, filed December 2, 2019, which claims the benefit of U.S. Provisional Patent Application Nos. 62 / 774,082, filed November 30, 2018; 62 / 788,689, filed January 4, 2019; and 62 / 789,495, filed January 7, 2019. The entire contents of the above applications are incorporated herein by reference.

[0002] Technical Field The present disclosure relates to the fields of data structures, data processing and machine learning and their use in precision medicine, e.g., the use of molecular profiling to guide personalized treatment recommendations for patients with various diseases and disorders, including cancer. [Background technology]

[0003] background Drug therapy for cancer patients has long been a challenge. Traditionally, when a patient was diagnosed with cancer, the treating physician typically selected from a predefined list of treatment options that conventionally corresponded to observable patient clinical factors, such as the type and stage of the cancer. As a result, cancer patients generally received the same treatment as other patients with the same type and stage of cancer. Because patients with the same type and stage of cancer often respond differently to the same treatment, the efficacy of such treatments is determined by trial and error. Moreover, when a patient does not respond immediately to any such "one-size-fits-all" treatment, or when a previously successful treatment stops working, the physician's treatment choice will often be based on anecdotal evidence at best.

[0004] Until the late 2000s, limited molecular testing was available to assist physicians in making more informed choices from a list of conventional treatments that corresponded to the patient's cancer type, also known as "cancer lineage." For example, a breast cancer patient's physician could have tested the patient's tumor for overexpression of the gene HER2 / neu if presented with a list of conventional treatment options that included Herceptin®. HER2 / neu was known at the time to be associated with breast cancer and responsiveness to Herceptin®. Approximately one-third of breast cancer patients whose tumors were known to overexpress the HER2 / neu gene showed an initial response to treatment with Herceptin®, but the majority of them began to progress within one year. See, e.g., Bartsch, R. et al., Trastuzumab in the management of early and advanced stage breast cancer, Biologies. 2007 Mar; 1(1): 19-31. This type of molecular test helped explain why known treatments for a particular type of cancer are more effective than others at treating some patients with that type of cancer, but the test did not identify or rule out any further treatment options for the patient.

[0005] Frustrated with a one-size-fits-all approach to treating cancer patients, and faced with the reality that many patients' tumors will progress and eventually exhaust all conventional therapies, oncologist Daniel Von Hoff sought to identify additional non-conventional treatment options for his patients. Recognizing the limitations of making treatment decisions based on clinical observations and the limitations of lineage-specific molecular testing, and believing that these limitations are causing effective treatment options to be overlooked, Von Hoff and his colleagues developed a system and method for determining personalized treatment regimens for cancer based on a comprehensive assessment of the tumor's molecular characteristics. Their approach to such "molecular profiling" used a variety of testing techniques to collect molecular information from a patient's tumor to create a unique molecular profile regardless of the type of cancer. Physicians could then use the results of that molecular profile to aid in the selection of candidate treatments for patients, regardless of the stage, anatomical location, or anatomical origin of the cancer cells. See Von Hoff DD, et al., Pilot study using molecular profiling of patients' tumors to find potential targets and select treatments for their refractory cancers. J Clin Oncol. 2010 Nov 20;28(33):4877-83. Such molecular profiling approaches can suggest potential benefits of treatments that would otherwise be overlooked by treating physicians, as well as non-promising benefits of certain treatments, thereby avoiding the time, expense, disease progression and side effects associated with ineffective treatments. Molecular profiling can be particularly beneficial in the "salvage therapy" setting when patients have failed to respond to or developed resistance to multiple treatment regimens. In addition, such approaches can also be used to guide decision-making for first-line and other standard treatment regimens.

[0006] Platinum therapy refers to treatment with chemotherapy agents that contain platinum coordination complexes. Such platinum-based drugs include cisplatin, carboplatin, oxaliplatin, and nedaplatin, FDA-approved drugs used to treat nearly half of the people undergoing chemotherapy for cancer. Additional platinum agents have been or are being studied, including but not limited to triplatin tetranitrate, phenanthriplatin, picoplatin, and satraplatin. Although such compounds can help some patients fight cancer, they can also have severe side effects, including but not limited to neurotoxicity (which can lead to neuropathy), nephrotoxicity, bone marrow suppression, anaphylaxis, cytopenias (e.g., leukopenia and neutropenia, thrombocytopenia, and anemia), hepatotoxicity, ototoxicity, cardiac toxicity, nausea and vomiting, diarrhea, mucositis, stomatitis, pain, hair loss, loss of appetite, cachexia, and asthenia. See, for example, Oun R et al. The side effects of platinum-based chemotherapy drugs: a review for chemists. Dalton Trans. 2018 May 15;47(19):6645-6653. doi: 10.1039 / c8dt00838h, the entire reference of which is incorporated herein by reference. Patients administered platinum compounds require constant monitoring and potential dose reductions, eventually developing resistance. Furthermore, patients may require additional non-chemotherapeutic drugs to treat such side effects. See ibid. Thus, there is a need to identify patients who are likely to benefit or not benefit from platinum therapy, at least for reasons that will help select beneficial treatments while avoiding the time, cost, and side effects of treatments that are unlikely or have reduced benefits.

[0007] A machine learning model can be configured to analyze labeled training data and draw inferences from the training data. When a machine learning model is trained, a set of unlabeled data can be provided as input to the machine learning model. The machine learning model can process the input data, e.g., molecular profiling data, and perform predictions about the input based on inferences learned during training. The present disclosure provides a "voting" methodology for combining multiple classifier models to achieve more accurate classification than can be achieved by using a single model.

[0008] Comprehensive molecular profiling provides a wealth of data on the molecular status of patient samples. We have performed such profiling on well over 100,000 tumor patients from virtually all cancer lines, and have followed up patient outcomes and responses to treatment in several thousand of these patients. For example, our molecular profiling data can be compared to patient benefit or reduced / lack of benefit to treatment and processed using machine learning algorithms, such as "voting" methodology, to identify additional biomarker signatures that predict the effectiveness of various treatments. Here, this "next-generation profiling" (NGP) approach is applied to identify biomarker signatures that predict the probability of benefit from platinum therapy in cancer patients. [Prior art documents] [Non-patent literature]

[0009] [Non-Patent Document 1] Bartsch, R. et al., Trastuzumab in the management of early and advanced stage breast cancer, Biologies. 2007 Mar; 1(1): 19-31 [Non-Patent Document 2] Von Hoff DD, et al., Pilot study using molecular profiling of patients' tumors to find potential targets and select treatments for their refractory cancers. J Clin Oncol. 2010 Nov 20;28(33):4877-83 [Non-Patent Document 3] Oun R et al. The side effects of platinum-based chemotherapy drugs: a review for chemists. Dalton Trans. 2018 May 15;47(19):6645-6653. doi: 10.1039 / c8dt00838h Summary of the Invention

[0010] overview Comprehensive molecular profiling provides a wealth of data regarding the molecular state of a patient sample. Such data can be compared to patient response to treatment to identify biomarker signatures that predict response or non-response to such treatment. This approach has been applied to identify biomarker signatures that correlate with increased benefit or decreased benefit of platinum-based chemotherapy, such as cisplatin, carboplatin, oxaliplatin, and / or nedaplatin, in cancer patients.

[0011] Described herein is a method of training a machine learning model to predict the efficacy of a treatment for a disease or disorder in a subject having a particular set of biomarkers.

[0012] Provided herein is a data processing apparatus for generating an input data structure for use in training a machine learning model to predict the efficacy of a treatment for a disease or disorder in a subject, the data processing apparatus including one or more processors and one or more storage devices that store instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, the operations including obtaining, by the data processing apparatus, one or more biomarker data structures and one or more outcome data structures; extracting, by the data processing apparatus, first data representing one or more biomarkers associated with the subject from the one or more biomarker data structures, extracting, by the data processing apparatus, second data representing the disease or disorder and the treatment from the one or more outcome data structures, and extracting, by the data processing apparatus, third data representing an outcome of the treatment for the disease or disorder. generating, by the data processing device, a data structure for input to the machine learning model based on first data representing the one or more biomarkers and second data representing the disease or disorder and the treatment; providing, by the data processing device, the generated data structure as input to the machine learning model; obtaining, by the data processing device, an output generated by the machine learning model based on processing of the machine learning model of the generated data structure; determining, by the data processing device, a difference between third data representing an outcome of the treatment for the disease or disorder and the output generated by the machine learning model; and adjusting, by the data processing device, one or more parameters of the machine learning model based on the difference between the third data representing the outcome of the treatment for the disease or disorder and the output generated by the machine learning model.

[0013] In some embodiments, the set of one or more biomarkers comprises one or more biomarkers set forth in any one of Tables 2-8. In some embodiments, the set of one or more biomarkers comprises each of the biomarkers in Tables 2-8. In some embodiments, the set of one or more biomarkers comprises at least one of the biomarkers in Tables 2-8; optionally, the set of one or more biomarkers comprises a biomarker in Table 5, Table 6, Table 7, Table 8, or any combination thereof; or optionally, the set of one or more biomarkers comprises substantially a whole exome, whole genome, and / or whole transcriptome.

[0014] Also provided herein is a data processing apparatus for generating an input data structure for use in training a machine learning model to predict a subject's therapeutic responsiveness to a particular treatment, the data processing apparatus comprising one or more processors and one or more storage devices storing instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, the operations including obtaining, by the data processing apparatus, from a first distributed data source, a first data structure that structures data representing a set of one or more biomarkers associated with a subject, the first data structure including a key-value that identifies the subject; storing, by the data processing apparatus, the first data structure in one or more memory devices; obtaining, by the data processing apparatus, from a second distributed data source, a second data structure that structures data representing outcome data for a subject having the one or more biomarkers, the outcome data including a key-value that identifies the subject. storing, by the data processing device, the second data structure in one or more memory devices; generating, by the data processing device and using the first data structure and the second data structure stored in the memory device, a labeled training data structure comprising (i) data representative of the set of one or more biomarkers, the disease or disorder, and the treatment, and (ii) a label providing an indication of the effectiveness of the treatment for the disease or disorder (the generating, by the data processing device and using the first data structure and the second data structure, includes correlating, by the data processing device, the first data structure structuring the data representative of the set of one or more biomarkers associated with the subject based on the key-values ​​identifying the subject, with the second data structure representing outcome data for the subject having the one or more biomarkers);and a step of training a machine learning model using the generated labeled training data structure by the data processing device (the step of training a machine learning model using the generated labeled training data structure includes providing, by the data processing device, the generated labeled training data structure to the machine learning model as an input to the machine learning model).

[0015] In some embodiments, the operations further include obtaining, by the data processing device, from the machine learning model an output generated by the machine learning model based on the machine learning model's processing of the generated labeled training data structure; and determining, by the data processing device, a difference between the output generated by the machine learning model and a label that provides an indication of the effectiveness of a treatment for the disease or disorder.

[0016] In some embodiments, the operations further include adjusting, by the data processing device, one or more parameters of the machine learning model based on the determined difference between the output generated by the machine learning model and the label that provides an indication of the effectiveness of a treatment for a disease or disorder.

[0017] In some embodiments, the set of one or more biomarkers comprises one or more biomarkers set forth in any one of Tables 2-8. In some embodiments, the set of one or more biomarkers comprises each of the biomarkers in Tables 2-8. In some embodiments, the set of one or more biomarkers comprises at least one of the biomarkers in Tables 2-8; optionally, the set of one or more biomarkers comprises the biomarkers in Table 5, Table 6, Table 7, Table 8, or any combination thereof; or optionally, the set of one or more biomarkers comprises substantially a whole exome, whole genome, and / or whole transcriptome.

[0018] Relatedly, provided herein is a method including steps corresponding to each of the operations of the data processing apparatus. Still further, provided herein is a system including one or more computers and one or more storage media storing instructions that, when executed by the one or more computers, cause the one or more computers to perform each of the operations described with reference to the data processing apparatus. Still further, provided herein is a non-transitory computer-readable medium storing software including instructions executable by one or more computers and, when so executed, cause the one or more computers to perform the operations described with reference to the data processing apparatus.

[0019] In another aspect, provided herein is a method for classification of an entity, the method comprising, for each particular machine learning model of a plurality of machine learning models, i) providing input data representative of a type of entity to be classified to the particular machine learning model trained to determine a prediction or classification; ii) obtaining output data representative of entity classifications of a plurality of candidate entity classes into initial entity classes generated by the particular machine learning model based on processing of the input data by the particular machine learning model; providing the obtained output data for each of the plurality of machine learning models to a voting unit, the provided output data including data representative of the initial entity classes determined by each of the plurality of machine learning models; and determining, by the voting unit, an actual entity class for the entity based on the provided output data.

[0020] In some embodiments, the actual entity class for an entity is determined by applying a majority voting rule to the provided output data.

[0021] In some embodiments, determining, by the voting unit, the actual entity class for the entity based on the output data provided includes: determining, by the voting unit, a number of occurrences of each initial entity class of a plurality of candidate entity classes; and selecting, by the voting unit, the initial entity class having the greatest number of occurrences among the plurality of candidate entity classes.

[0022] In some embodiments, each machine learning model of the plurality of machine learning models comprises a random forest classification algorithm, a support vector machine, a logistic regression, a k-nearest neighbor model, an artificial neural network, a naive Bayes model, a quadratic discriminant analysis, or a Gaussian process model.

[0023] In some embodiments, each machine learning model of the plurality of machine learning models comprises a random forest classification algorithm.

[0024] In some embodiments, the multiple machine learning models include multiple representations of the same type of classification algorithm.

[0025] In some embodiments, the input data represents (i) entity attributes and (ii) types of treatments for a disease or disorder.

[0026] In some embodiments, the plurality of candidate entity classes comprises a reactive class or a non-responsive class.

[0027] In some embodiments, the entity attributes include one or more biomarkers for the entity. In some embodiments, the one or more biomarkers include a panel of genes that are less than all known genes of the entity. In some embodiments, the one or more biomarkers include a panel of genes that includes all known genes for the entity. In some embodiments, the one or more biomarkers include one or more biomarkers listed in any one of Tables 2-8; optionally, the one or more biomarkers include a marker in Table 5, Table 6, Table 7, Table 8, or any combination thereof; or optionally, the one or more biomarkers include substantially the whole exome and / or the whole transcriptome.

[0028] In some embodiments, the input data further comprises data representative of a type of disease or disorder.

[0029] Relatedly, provided herein is a system including one or more computers and one or more storage media storing instructions that, when executed by the one or more computers, cause the one or more computers to perform each of the operations described with reference to the method for classification of entities. Still further, provided herein is a non-transitory computer-readable medium storing software including instructions executable by one or more computers and that, when so executed, cause the one or more computers to perform the operations described with reference to the method for classification of entities.

[0030] In yet another aspect, provided herein is a method comprising obtaining a biological sample comprising cells from a cancer in a subject; and performing an assay to evaluate at least one biomarker in the biological sample, wherein the biomarker is: (a) Group 1, including all 1, 2, 3, 4, 5, or 6 of MYC, EP300, U2AF1, ASXL1, MAML2, and CNTRL; (b) group 2, including all 1, 2, 3, 4, 5, 6, 7, or 8 of MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; (c) group 3, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, or 14 of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, HOXA11, AURKA, BIRC3, IKZF1, CASP8, and EP300; (d) group 4, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, or 13 of PBX1, BCL9, INHBA, PRRX1, YWHAE, GNAS, LHFPL6, FCRL4, AURKA, IKZF1, CASP8, PTEN, and EP300; (e) group 5, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 of BCL9, PBX1, PRRX1, INHBA, GNAS, YWHAE, LHFPL6, FCRL4, PTEN, HOXA11, AURKA, and BIRC3; (f) group 6, including all 1, 2, 3, 4, or 5 of BCL9, PBX1, PRRX1, INHBA, and YWHAE; (g) group 7, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, or 15 of BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; (h)BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CR Group 8, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 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 or 45 of EB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR; and (i) Group 9, including 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, or all 11 of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11; The method includes at least one of the following:

[0031] In some embodiments, the biological sample comprises formalin-fixed paraffin-embedded (FFPE) tissue, fixed tissue, core needle biopsy, fine needle aspirate, unstained slide, fresh frozen (FF) tissue, formalin sample, tissue contained in a solution that preserves nucleic acid or protein molecules, fresh sample, malignant fluid, bodily fluid, tumor sample, tissue sample, or any combination thereof.

[0032] In some embodiments, the biological sample comprises cells from a solid tumor.

[0033] In some embodiments, the biological sample comprises a bodily fluid.

[0034] In some embodiments, the bodily fluid comprises malignant fluid, pleural fluid, peritoneal fluid, or any combination thereof.

[0035] In some embodiments, the bodily fluid comprises peripheral blood, serum, plasma, ascites, urine, cerebrospinal fluid (CSF), sputum, saliva, bone marrow, synovial fluid, aqueous humor, amniotic fluid, earwax, breast milk, bronchoalveolar lavage fluid, semen, prostatic fluid, Cowper's fluid, pre-ejaculate fluid, female ejaculate, sweat, feces, tears, cyst fluid, pleural fluid, peritoneal fluid, pericardial fluid, lymph, chyme, chyle, bile, interstitial fluid, menstrual fluid, pus, sebum, vomit, vaginal fluid, mucosal secretions, stool water, pancreatic juice, nasal lavage fluid, bronchopulmonary aspirate, blastocyst fluid, or umbilical cord blood.

[0036] In some embodiments, the evaluation comprises determining the presence, level or status of a protein or nucleic acid for each biomarker, and optionally the nucleic acid comprises deoxyribonucleic acid (DNA), ribonucleic acid (RNA) or a combination thereof. In some embodiments, (a) the presence, level or status of a protein is determined using immunohistochemistry (IHC), flow cytometry, immunoassay, antibody or functional fragment thereof, aptamer or any combination thereof; and / or (b) the presence, level or status of a nucleic acid is determined using polymerase chain reaction (PCR), in situ hybridization, amplification, hybridization, microarray, nucleic acid sequencing, dye terminator sequencing, pyrosequencing, next generation sequencing (NGS; high throughput sequencing) or any combination thereof.

[0037] In some embodiments, the state of a nucleic acid comprises a sequence, a mutation, a polymorphism, a deletion, an insertion, a substitution, a translocation, a fusion, a truncation, a duplication, an amplification, a repeat, a copy number, a copy number variation (CNV; copy number alteration; CNA), or any combination thereof.

[0038] In some embodiments, the state of a nucleic acid comprises copy number.

[0039] In some embodiments, the method comprises performing an assay to determine the copy number of all members of Group 1 (i.e., MYC, EP300, U2AF1, ASXL1, MAML2 and CNTRL) or genomic regions adjacent thereto.

[0040] In some embodiments, the method includes performing an assay to determine the copy number of all members of Group 2 (i.e., MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN and CDX2) or genomic regions adjacent thereto.

[0041] In some embodiments, the method includes performing an assay to determine the copy number of all members of Group 3 (i.e., BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, HOXA11, AURKA, BIRC3, IKZF1, CASP8 and EP300) or genomic regions adjacent thereto.

[0042] In some embodiments, the method includes performing an assay to determine the copy number of all members of Group 4 (i.e., PBX1, BCL9, INHBA, PRRX1, YWHAE, GNAS, LHFPL6, FCRL4, AURKA, IKZF1, CASP8, PTEN and EP300) or genomic regions adjacent thereto.

[0043] In some embodiments, the method includes performing an assay to determine the copy number of all members of Group 5 (i.e., BCL9, PBX1, PRRX1, INHBA, GNAS, YWHAE, LHFPL6, FCRL4, PTEN, HOXA11, AURKA and BIRC3) or genomic regions adjacent thereto.

[0044] In some embodiments, the method comprises performing an assay to determine the copy number of all members of Group 6 (i.e., BCL9, PBX1, PRRX1, INHBA and YWHAE) or genomic regions adjacent thereto.

[0045] In some embodiments, the method includes performing an assay to determine the copy number of all members of Group 7 (i.e., BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1 and MNX1) or genomic regions adjacent thereto.

[0046] In some embodiments, the method comprises performing an assay to determine the copy number of all members of Group 8 (i.e., BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CREB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1 and EZR) or genomic regions adjacent thereto.

[0047] In some embodiments, the method includes performing an assay to determine the copy number of all members of Group 9 (i.e., BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA and HOXA11) or genomic regions adjacent thereto.

[0048] In some embodiments, the method comprises performing an assay to determine the copy number of (a) at least one or all members of Group 1 and Group 2, or genomic regions adjacent thereto; (b) at least one or all members of Group 3, or genomic regions adjacent thereto; or (c) at least one or all members of Group 2, Group 6, Group 7, Group 8, and Group 9, or genomic regions adjacent thereto.

[0049] In some embodiments, the method further comprises comparing the copy number of the biomarker to a reference copy number (e.g., diploid) to identify biomarkers having copy number variation (CNV).

[0050] In some embodiments, the method further comprises generating a molecular profile that identifies the gene or a region adjacent to the gene that has the CNV.

[0051] In some embodiments, the presence or level of PTEN protein is determined, and optionally, the presence or level of PTEN protein is determined using immunohistochemistry (IHC).

[0052] In some embodiments, the method further comprises determining the levels of proteins including TOPO1 and one or more mismatch repair proteins (e.g., MLH1, MSH2, MSH6 and PMS2), and optionally, the presence or level of PTEN protein is determined using immunohistochemistry (IHC).

[0053] In some embodiments, the method further comprises comparing the level of the one or more proteins to a reference level of that protein.

[0054] In some embodiments, the method further comprises generating a molecular profile that identifies proteins having levels that differ from the reference levels, e.g., significantly different from the reference levels.

[0055] In some embodiments, the method further comprises predicting an increase or decrease in benefit of the cancer treatment based on the evaluated biomarkers, and optionally, the treatment comprises platinum-based chemotherapy or a combination therapy comprising platinum-based chemotherapy, and optionally, the platinum-based chemotherapy comprises cisplatin, carboplatin, oxaliplatin and / or nedaplatin, and the combination therapy comprising platinum-based chemotherapy comprises FOLFOX, FOLFOXIRI and / or FOLFIRINOX.

[0056] In some embodiments, predicting the increased or decreased benefit of the treatment is based on (a) the copy number determined for the group; and / or (b) the molecular profile determined as described above.

[0057] In some embodiments, predicting an increased or decreased benefit of the treatment based on the copy numbers determined for the group comprises the use of a voting module.

[0058] In some aspects the voting module is as provided herein.

[0059] In some embodiments, the voting module includes the use of at least one random forest model.

[0060] In some embodiments, the use of the voting module comprises applying a machine learning classification model to the copy numbers obtained for each of group 2, group 6, group 7, group 8 and group 9 (see above), optionally where each machine learning classification model is a random forest model, optionally where the random forest model is one described in Table 10 below.

[0061] In some embodiments, the subject has not been previously treated with the therapy.

[0062] In some embodiments, the cancer comprises metastatic cancer, recurrent cancer, or a combination thereof.

[0063] In some embodiments, the subject has not previously undergone treatment for cancer.

[0064] In some embodiments, the method further comprises administering to the subject a treatment predicted to have an increased benefit.

[0065] In some embodiments, the method further comprises not administering to the subject a treatment that is predicted to have reduced benefit.

[0066] In some embodiments, progression free survival (PFS), disease free survival (DFS) or life span is extended by administration of said treatment.

[0067] In some embodiments, the cancer is acute lymphoblastic leukemia; acute myeloid leukemia; adrenocortical carcinoma; AIDS-related cancer; AIDS-related lymphoma; anal cancer; appendix cancer; astrocytoma; atypical teratoid / rhabdoid tumor; basal cell carcinoma; bladder cancer; brain stem glioma; brain tumor, brain stem glioma, central nervous system atypical teratoid / rhabdoid tumor, central nervous system embryonal tumor, astrocytoma, craniopharyngioma, ependymoblastoma, ependymoma, medulloblastoma, medulloepithelioma, intermediate pineal parenchymal tumor, supratentorial primitive neuroectodermal tumor and pineoblastoma; breast cancer; bronchoma tumors;Burkitt's lymphoma;carcinoid tumors;carcinoma of unknown primary (CUP);carcinoid tumors;carcinoma of unknown primary;central nervous system atypical teratoid / rhabdoid tumors;central nervous system embryonal tumors;cervical cancer;childhood cancer;chordoma;chronic lymphocytic leukemia;chronic myelogenous leukemia;chronic myeloproliferative disorders;colon cancer;colorectal cancer;craniopharyngioma;cutaneous T-cell lymphoma;endocrine islet cell tumors;endometrial cancer;ependymomas;ependymoma;esophageal cancer;nasal neuroblastoma;Ewing's sarcoma;extracranial germ cell tumors;extragonadal germ cell tumors;extrahepatic bile duct cancer;gallbladder cancer;gastric cancer (stomach) cancer);gastrointestinal carcinoid tumor;gastrointestinal stromal cell tumor;gastrointestinal stromal tumor (GIST);gestational trophoblastic tumor;glioma;hairy cell leukemia;head and neck cancer;cardiac cancer;Hodgkin lymphoma;hypopharyngeal cancer;intraocular melanoma;pancreatic islet tumor;Kaposi's sarcoma;kidney cancer;Langerhans cell histiocytosis;laryngeal cancer;lip cancer;liver cancer;malignant fibrous histiocytoma bone cancer;medulloblastoma;medulloepithelioma;melanoma;Merkel cell carcinoma;Merkel cell skin cancer;mesothelioma;metastatic squamous neck cancer of unknown primary;oral cancer;multiple endocrine neoplasia syndrome;multiple myeloma;multiple myeloma / plasma cell neoplasm;mycosis fungoides;myelodysplastic syndrome;myeloproliferative neoplasms;nasal cancer;nasopharyngeal carcinoma;neuroblastoma;non-Hodgkin lymphoma;non-melanoma skin cancer;non-small cell lung cancer;oral cancer cancer);oral cavity cancer;oropharyngeal cancer;osteosarcoma;other brain and spinal cord tumors;ovarian cancer;ovarian epithelial cancer;ovarian germ cell tumors;ovarian low malignant potential tumors;pancreatic cancer;papillomatosis;sinoasal cancer;parathyroid cancer;pelvic cancer;penile cancer;pharyngeal cancer;intermediate pineal parenchymal tumors;pineoblastoma;pituitary tumors;plasma cell neoplasms / multiple myeloma;pleuropulmonary blastoma;primary central nervous system (CNS) lymphoma;primary hepatocellular carcinoma;prostate cancer;rectal cancer;renal cancer;renal cell (kidney) cancer;renal cell carcinoma;airway cancer;retinoblastoma;rhabdomyosarcoma;salivary gland cancer;Sezary syndrome;small cell lung cancer;small intestine cancer;soft tissue sarcoma;squamous cell carcinoma;squamous cell carcinoma of the cervix;stomach (gastric) cancer;supratentorial primitive neuroectodermal tumor;T-cell lymphoma;testicular cancer;laryngopharyngeal cancer;thymic carcinoma;thymoma;thyroid cancer;transitional cell carcinoma;transitional cell carcinoma of the renal pelvis and ureter;trophoblastic tumor;ureteral cancer;urethral cancer;uterine cancer;uterine sarcoma;vaginal cancer;vulvar cancer;Waldenstrom macroglobulinemia;or Wilms' tumor.

[0068] In some embodiments, the cancer is acute myeloid leukemia (AML), breast cancer, cholangiocarcinoma, colorectal adenocarcinoma, extrahepatic bile duct adenocarcinoma, female genital malignancies, gastric adenocarcinoma, gastroesophageal adenocarcinoma, gastrointestinal stromal tumor (GIST), glioblastoma, head and neck squamous cell carcinoma, leukemia, hepatocellular carcinoma, low-grade glioma, lung bronchioloalveolar carcinoma (BAC), non-small cell lung cancer (NSCLC), small cell lung cancer (SCLC), lymphoma, male reproductive tract cancer, or malignant tumors. Includes organ malignancies, malignant solitary fibrous tumor of the pleura (MSFT), melanoma, multiple myeloma, neuroendocrine tumors, nodular diffuse large B-cell lymphoma, non-epithelial ovarian cancer (non-EOC), ovarian surface epithelial carcinoma, pancreatic adenocarcinoma, pituitary carcinoma, oligodendroglioma, prostate adenocarcinoma, retroperitoneal or peritoneal carcinoma, retroperitoneal or peritoneal sarcoma, small intestine malignancies, soft tissue tumors, thymic carcinoma, thyroid carcinoma or uveal melanoma.

[0069] In some embodiments, the cancer comprises colorectal cancer, ovarian cancer, esophageal cancer, gastroesophageal junction cancer, gastric cancer, head and neck cancer, bladder cancer, breast cancer, endometrial cancer, uterine cancer, cervical cancer, pancreatic cancer or lung cancer.

[0070] In some embodiments, the cancer comprises colorectal cancer. In some embodiments, the method further comprises determining a consensus molecular subtype (CMS) for colorectal cancer.

[0071] Further provided herein is a method of selecting a treatment for a subject having colorectal cancer, comprising obtaining a biological sample comprising cells from the colorectal cancer; performing next generation sequencing on genomic DNA from the biological sample to identify (a) group 2, including one, two, three, four, five, six, seven or all eight of MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; (b) group 3, including one, two, three, four or all eight of BCL9, PBX1, PRRX1, INHBA, and YWHAE; (c) group 6, including all 5 of the following: BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; (d) group 7, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, or 15 of the following: BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARC 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 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, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 109, 109, 109, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121 , 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 or 45 of group 8, and (e) determining the copy number for each of group 9, which includes one, two, three, four, five, six, seven, eight, nine, ten or all eleven of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11;The method includes: applying a machine learning classification model to the obtained copy number for each of group 2, group 6, group 7, group 8, and group 9 genes or genomic regions adjacent thereto (optionally, each machine learning classification model is a random forest model, and optionally, the random forest model is one described in Table 10); obtaining an indication from each machine learning classification model of whether the subject is predicted to have an increased probability of benefit from treatment with platinum-based chemotherapy or a decreased probability of said benefit, where optionally the platinum-based chemotherapy comprises cisplatin, carboplatin, oxaliplatin, and / or nedaplatin; and selecting a platinum-based chemotherapy or a combination therapy comprising platinum-based chemotherapy if the majority of the machine learning classification models predict that the subject will have an increased probability of benefit from platinum-based chemotherapy, and selecting an alternative treatment to platinum-based chemotherapy or an additional treatment to be used in combination with platinum-based chemotherapy if the majority of the machine learning classification models predict that the subject will have a decreased probability of benefit from platinum-based chemotherapy. In some embodiments, the method further includes administering the selected treatment to the subject;

[0072] Still further provided herein is a method for generating a molecular profiling report, comprising generating a report summarizing the results of carrying out the above-mentioned method.In some embodiments, the report includes (a) a prediction of the increase or decrease of the benefit of at least one treatment determined as disclosed above; or (b) a selected treatment determined as disclosed above.In some embodiments, the report is computer-generated; is a printed report or computer file; or is accessible via a web portal.

[0073] Relatedly, provided herein is a system for identifying a treatment for cancer in a subject, the system including: (a) at least one host server; (b) at least one user interface for accessing the at least one host server for accessing and inputting data; (c) at least one processor for processing the input data; (d) at least one memory coupled to the processor for storing the processed data and instructions for (1) accessing results of analyzing a biological sample as described above, and (2) predicting an increase or decrease in benefit of at least one treatment as described above or a selected treatment as described above; and (e) at least one display for displaying a treatment for the cancer, the treatment being a platinum-based chemotherapy, and optionally the platinum-based chemotherapy comprising cisplatin, carboplatin, oxaliplatin and / or nedaplatin.

[0074] In some embodiments, at least one display includes a report including the results of analyzing the biological sample and treatments having promising benefits in treating the cancer or treatments selected for treating the cancer.

[0075] Additionally provided herein is a method of providing recommendations for cancer treatment to provide longer progression-free survival, longer disease-free survival, longer overall survival or life extension, comprising the steps of: obtaining a biological sample comprising nucleic acids and / or proteins from an individual diagnosed with cancer; performing molecular testing on the biological sample to determine one or more molecular features selected from the group consisting of: nucleic acid sequences of a set of target genes or portions thereof; the presence of copy number variation of the set of target genes; the presence of gene fusions or other genomic alterations; one or more levels of a set of proteins and / or transcripts; and / or an epigenetic status of the set of target genes, e.g., as described herein, thereby generating a molecular profile for the cancer; comparing the molecular profile of the cancer to a reference molecular profile for that type of cancer; generating a list of molecular features that exhibit differences, e.g., significant differences, when compared to the reference molecular profile; and generating a list of one or more treatment recommendations for the individual based on the list of molecular features that exhibit differences when compared to the reference sequence profile of the target genes.

[0076] In some embodiments, the molecular test is at least one of next generation sequencing, Sanger sequencing, ISH, fragment analysis, PCR, IHC and immunoassay.

[0077] In some embodiments, the biological sample comprises a cell, a tissue sample, a blood sample, or a combination thereof.

[0078] In some embodiments, the molecular test detects at least one of a mutation, polymorphism, deletion, insertion, substitution, translocation, fusion, truncation, duplication, amplification, or repeat.

[0079] In some embodiments, the nucleic acid sequence comprises a deoxyribonucleic acid sequence.

[0080] In some embodiments, the nucleic acid sequence comprises a ribonucleic acid sequence.

[0081] Unless otherwise specified, all scientific and technical terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention belongs.Methods and materials for use in the present invention are described herein, but other suitable methods and materials known in the art can also be used.Materials, methods and examples are illustrative only and are not intended to be limiting.All publications, patent applications, patents, sequences, database entries and other references mentioned herein are incorporated by reference in their entirety.In case of conflict, the present specification, including definitions, will prevail.

[0082] [The present invention 1001] 1. A data processing apparatus for generating an input data structure for use in training a machine learning model for predicting the efficacy of a treatment for a disease or disorder of interest, comprising: the data processing apparatus includes one or more processors and one or more storage devices that store instructions that, when executed by the one or more processors, cause the one or more processors to perform operations; The operation is obtaining, by said data processing device, one or more biomarker data structures and one or more outcome data structures; extracting, by the data processing device, first data representing one or more biomarkers associated with a subject from the one or more biomarker data structures, second data representing a disease or disorder and a treatment from the one or more outcome data structures, and third data representing an outcome of a treatment for the disease or disorder; generating, by the data processing device, a data structure for input into a machine learning model based on the first data representative of the one or more biomarkers and the second data representative of the disease or disorder and treatment; providing, by the data processing device, the generated data structure as an input to the machine learning model; obtaining, by the data processing device, an output generated by the machine learning model based on processing of the generated data structure by the machine learning model; determining, by the data processing device, a difference between third data representative of an outcome of a treatment for the disease or disorder and an output generated by the machine learning model; and adjusting, by the data processing device, one or more parameters of the machine learning model based on a difference between third data representative of an outcome of a treatment for the disease or disorder and an output generated by the machine learning model. The data processing device. [The present invention 1002] 1001. The data processing apparatus of the present invention, wherein the set of one or more biomarkers comprises one or more biomarkers listed in any one of Tables 2-8. [The present invention 1003] A data processing apparatus of the present invention 1001, wherein the set of one or more biomarkers includes each of the biomarkers of the present invention 1002. [The present invention 1004] The data processing device of the present invention 1001, wherein the set of one or more biomarkers comprises at least one of the biomarkers of the present invention 1002; optionally, the set of one or more biomarkers comprises the markers in Table 5, Table 6, Table 7, Table 8 or any combination thereof; or optionally, the set of one or more biomarkers comprises substantially the whole exome, whole genome and / or whole transcriptome. [The present invention 1005] 1. A data processing apparatus for generating an input data structure for use in training a machine learning model for predicting a subject's therapeutic responsiveness to a particular treatment, comprising: the data processing apparatus includes one or more processors and one or more storage devices that store instructions that, when executed by the one or more processors, cause the one or more processors to perform operations; The operation is obtaining, by the data processing device, from a first distributed data source, a first data structure structuring data representing a set of one or more biomarkers associated with a subject, the first data structure including a key-value that identifies the subject; storing, by the data processing apparatus, the first data structure in one or more memory devices; obtaining, by the data processing device, from a second distributed data source, a second data structure structuring data representing outcome data for subjects having the one or more biomarkers, the outcome data including data identifying a disease or disorder, a treatment, and an indication of efficacy of the treatment, and the second data structure also including key-values ​​identifying the subjects; storing, by the data processing apparatus, the second data structure in one or more memory devices; generating, by the data processing apparatus and by using the first data structure and the second data structure stored in the memory device, (i) a labeled training data structure comprising data representative of a set of one or more biomarkers, the disease or disorder, and a treatment, and (ii) a label providing an indication of the effectiveness of a treatment for the disease or disorder, wherein the generating, by the data processing apparatus and by using the first data structure and the second data structure, comprises correlating, by the data processing apparatus, a first data structure that structures data representative of the set of one or more biomarkers associated with the subject based on a key-value that identifies the subject, with a second data structure that represents outcome data for subjects having the one or more biomarkers; and training, by the data processing device, a machine learning model using the generated labeled training data structure, wherein training, by the data processing device, a machine learning model using the generated labeled training data structure comprises providing, by the data processing device, the generated labeled training data structure to the machine learning model as an input to the machine learning model. The data processing device. [The present invention 1006] The operation is obtaining, by the data processing device, from the machine learning model, an output generated by the machine learning model based on the machine learning model's processing of the generated labeled training data structure; and determining, by the data processing device, a difference between an output generated by the machine learning model and a label that provides an indication of the effectiveness of a treatment for a disease or disorder. The data processing device of the present invention 1005 further comprises: [The present invention 1007] The operation is adjusting, by the data processing device, one or more parameters of the machine learning model based on the determined difference between the output generated by the machine learning model and the label that provides an indication of the effectiveness of the treatment for the disease or disorder. The data processing device of the present invention 1006 further comprising: [The present invention 1008] The data processing device of the present invention 1005, wherein the set of one or more biomarkers comprises one or more biomarkers set forth in any one of Tables 2-8; optionally, the set of one or more biomarkers comprises the markers in Table 5, Table 6, Table 7, Table 8, or any combination thereof; or optionally, the set of one or more biomarkers comprises substantially the whole exome, whole genome, and / or whole transcriptome. [The present invention 1009] A data processing apparatus of the present invention 1005, wherein the set of one or more biomarkers includes each of the biomarkers of the present invention 1008. [The present invention 1010] A data processing apparatus of the present invention 1005, wherein the set of one or more biomarkers includes one of the biomarkers of the present invention 1008. [The present invention 1011] A method including steps corresponding to any of the operations 1001 to 1010 of the present invention. [The present invention 1012] A system including one or more computers and one or more data storage media storing instructions that, when executed by the one or more computers, cause the one or more computers to perform each of the operations of the present invention 1001 to 1010. [The present invention 1013] Instructions executable by one or more computers, which when so executed cause the one or more computers to perform any of the operations of the present invention 1001-1010 A non-transitory computer readable medium storing software comprising: [The present invention 1014] 1. A method for classification of entities, comprising: With respect to each particular machine learning model of the plurality of machine learning models, providing input data representing the type of entity to be classified to a particular machine learning model that has been trained to determine a prediction or classification; obtaining output data representative of entity classifications of a plurality of candidate entity classes into initial entity classes generated by the particular machine learning model based on processing of input data by the particular machine learning model; providing output data obtained for each of the plurality of machine learning models to a voting unit, the provided output data including data representative of an initial entity class determined by each of the plurality of machine learning models; and determining, by the voting unit, a real entity class for the entity based on the provided output data. The method comprising: [The present invention 1015] The method of the present invention 1014, wherein the actual entity class for the entity is determined by applying a majority voting rule to the provided output data. [The present invention 1016] determining, by the voting unit, a real entity class for the entity based on the output data provided, determining, by the voting unit, a number of occurrences of each initial entity class of a plurality of candidate entity classes; and selecting, by the voting unit, an initial entity class having a maximum number of occurrences from among the plurality of candidate entity classes; The method of the present invention 1014 or 1015, comprising: [The present invention 1017] Any of the methods of claims 1014 to 1016, wherein each machine learning model of the plurality of machine learning models comprises a random forest classification algorithm, a support vector machine, a logistic regression, a k-nearest neighbor model, an artificial neural network, a naive Bayes model, a quadratic discriminant analysis, or a Gaussian process model. [The present invention 1018] The method of any of claims 1014 to 1016, wherein each machine learning model of the plurality of machine learning models comprises a random forest classification algorithm. [The present invention 1019] The method of any of claims 1014-1018, wherein the multiple machine learning models include multiple representations of the same type of classification algorithm. [The present invention 1020] The method of any of claims 1014-1018, wherein the input data represents (i) entity attributes and (ii) types of treatment for a disease or disorder. [The present invention 1021] The method of the present invention 1020, wherein the plurality of candidate entity classes includes a reactive class or a non-responsive class. [The present invention 1022] The method of any one of claims 1020 to 1021, wherein the entity attributes include one or more biomarkers for the entity. [The present invention 1023] The method of claim 1022, wherein the one or more biomarkers comprise a panel of genes that is less than all known genes of the entity. [The present invention 1024] The method of claim 1022, wherein the one or more biomarkers comprises a panel of genes including all known genes for the entity. [The present invention 1025] The method of the present invention 1022, wherein the one or more biomarkers comprise one or more biomarkers set forth in any one of Tables 2-8; optionally, the one or more biomarkers comprise a marker in Table 5, Table 6, Table 7, Table 8, or any combination thereof; or optionally, the one or more biomarkers comprise substantially the whole exome and / or the whole transcriptome. [The present invention 1026] The method of any one of claims 1020 to 1025, wherein the input data further comprises data representative of a type of disease or disorder. [The present invention 1027] A system including one or more computers and one or more data storage media storing instructions that, when executed by the one or more computers, cause the one or more computers to perform each of the operations of the present invention 1014-1026. [The present invention 1028] Instructions executable by one or more computers, which when so executed cause the one or more computers to perform any of the operations of the present invention 1014-1026 A non-transitory computer readable medium storing software comprising: [The present invention 1029] Obtaining a biological sample containing cells derived from a cancer in a subject; and performing an assay to evaluate at least one biomarker in said biological sample. A method comprising: The biomarker is (a) Group 1, including all 1, 2, 3, 4, 5, or 6 of MYC, EP300, U2AF1, ASXL1, MAML2, and CNTRL; (b) group 2, including all 1, 2, 3, 4, 5, 6, 7, or 8 of MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; (c) group 3, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, or 14 of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, HOXA11, AURKA, BIRC3, IKZF1, CASP8, and EP300; (d) group 4, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, or 13 of PBX1, BCL9, INHBA, PRRX1, YWHAE, GNAS, LHFPL6, FCRL4, AURKA, IKZF1, CASP8, PTEN, and EP300; (e) group 5, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 of BCL9, PBX1, PRRX1, INHBA, GNAS, YWHAE, LHFPL6, FCRL4, PTEN, HOXA11, AURKA, and BIRC3; (f) group 6, including all 1, 2, 3, 4, or 5 of BCL9, PBX1, PRRX1, INHBA, and YWHAE; (g) group 7, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, or 15 of BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; (h)BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CR Group 8, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 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 or 45 of EB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR; and (i) Group 9, including 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, or all 11 of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11; The method includes at least one of the following: [The present invention 1030] The method of the present invention 1029, wherein the biological sample comprises formalin-fixed paraffin-embedded (FFPE) tissue, fixed tissue, core needle biopsy, fine needle aspirate, unstained slide, fresh frozen (FF) tissue, formalin sample, tissue contained in a solution that preserves nucleic acid or protein molecules, fresh sample, malignant fluid, body fluid, tumor sample, tissue sample, or any combination thereof. [The present invention 1031] The method of any one of claims 1029 to 1030, wherein the biological sample comprises cells from a solid tumor. [The present invention 1032] The method of any one of claims 1029 to 1030, wherein the biological sample comprises a body fluid. [The present invention 1033] The method of any of claims 1029 to 1032, wherein the body fluid comprises malignant fluid, pleural fluid, peritoneal fluid, or any combination thereof. [The present invention 1034] 46. ​​The method of any of claims 1029 to 1033, wherein the body fluid comprises peripheral blood, serum, plasma, ascites, urine, cerebrospinal fluid (CSF), sputum, saliva, bone marrow, synovial fluid, aqueous humor, amniotic fluid, earwax, breast milk, bronchoalveolar lavage fluid, semen, prostatic fluid, Cowper's fluid, pre-ejaculate fluid, female ejaculate, sweat, feces, tears, cyst fluid, pleural fluid, peritoneal fluid, pericardial fluid, lymph, chyme, chyle, bile, interstitial fluid, menstrual secretions, pus, sebum, vomit, vaginal fluid, mucosal secretions, stool water, pancreatic juice, nasal washings, bronchopulmonary aspirate, blastocytic cavity fluid or umbilical cord blood. [The present invention 1035] Any of the methods of claims 1029 to 1034, wherein the evaluation comprises determining the presence, level or status of a protein or nucleic acid for each biomarker, and optionally, the nucleic acid comprises deoxyribonucleic acid (DNA), ribonucleic acid (RNA), or a combination thereof. [The present invention 1036] (a) the presence, level or status of a protein is determined using immunohistochemistry (IHC), flow cytometry, immunoassays, antibodies or functional fragments thereof, aptamers, or any combination thereof; and / or (b) the presence, level or state of nucleic acid is determined using polymerase chain reaction (PCR), in situ hybridization, amplification, hybridization, microarray, nucleic acid sequencing, dye terminator sequencing, pyrosequencing, next generation sequencing (NGS; high throughput sequencing), or any combination thereof; The method of the present invention 1035. [The present invention 1037] The method of the present invention 1036, wherein the nucleic acid state comprises a sequence, a mutation, a polymorphism, a deletion, an insertion, a substitution, a translocation, a fusion, a truncation, a duplication, an amplification, a repeat, a copy number, a copy number variation (CNV; copy number alteration; CNA), or any combination thereof. [The present invention 1038] The method of any one of claims 10 to 37, wherein the state of the nucleic acid comprises copy number. [The present invention 1039] The method of claim 1038, comprising the step of performing an assay to determine the copy number of all members of group 1 (i.e., MYC, EP300, U2AF1, ASXL1, MAML2 and CNTRL) or genomic regions adjacent thereto. [The present invention 1040] The method of the present invention 1038, comprising the step of performing an assay to determine the copy number of all members of group 2 (i.e., MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN and CDX2) or genomic regions adjacent thereto. [The present invention 1041] The method of the present invention 1038, comprising the step of performing an assay to determine the copy number of all members of group 3 (i.e., BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, HOXA11, AURKA, BIRC3, IKZF1, CASP8 and EP300) or genomic regions adjacent thereto. [The present invention 1042] The method of the present invention 1038, comprising the step of performing an assay to determine the copy number of all members of group 4 (i.e., PBX1, BCL9, INHBA, PRRX1, YWHAE, GNAS, LHFPL6, FCRL4, AURKA, IKZF1, CASP8, PTEN and EP300) or genomic regions adjacent thereto. [The present invention 1043] The method of the present invention 1038, comprising the step of performing an assay to determine the copy number of all members of group 5 (i.e., BCL9, PBX1, PRRX1, INHBA, GNAS, YWHAE, LHFPL6, FCRL4, PTEN, HOXA11, AURKA and BIRC3) or genomic regions adjacent thereto. [The present invention 1044] The method of claim 1038, comprising the step of performing an assay to determine the copy number of all members of Group 6 (i.e., BCL9, PBX1, PRRX1, INHBA and YWHAE) or genomic regions adjacent thereto. [The present invention 1045] The method of the present invention 1038, comprising the step of performing an assay to determine the copy number of all members of Group 7 (i.e., BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1 and MNX1) or genomic regions adjacent thereto. [The present invention 1046] The method of the present invention 1038, comprising the step of performing an assay to determine the copy number of all members of Group 8 (i.e., BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CREB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1 and EZR) or genomic regions adjacent thereto. [The present invention 1047] The method of the present invention 1038, comprising the step of performing an assay to determine the copy number of all members of group 9 (i.e., BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA and HOXA11) or genomic regions adjacent thereto. [The present invention 1048] (a) at least one or all members of Group 1 and Group 2 or genomic regions adjacent thereto; (b) at least one or all members of Group 3 or genomic regions adjacent thereto; or (c) at least one or all of the members of Group 2, Group 6, Group 7, Group 8, and Group 9, or genomic regions adjacent thereto The method of any one of claims 10 to 15, further comprising the step of performing an assay to determine the copy number of said gene. [The present invention 1049] The method of any of claims 1038 to 1048, further comprising the step of comparing the copy number of the biomarker to a reference copy number (e.g., diploid) to identify biomarkers having copy number variation (CNV). [The present invention 1050] The method of claim 1049, further comprising the step of generating a molecular profile that identifies the gene or a region adjacent thereto that has the CNV. [The present invention 1051] The method of any of claims 1029 to 1050, wherein the presence or level of PTEN protein is determined, and optionally, the presence or level of PTEN protein is determined using immunohistochemistry (IHC). [The present invention 1052] Any of the methods of claims 1029 to 1051, further comprising determining the level of proteins including TOPO1 and one or more mismatch repair proteins (e.g., MLH1, MSH2, MSH6 and PMS2), and optionally, the presence or level of the PTEN protein is determined using immunohistochemistry (IHC). [The present invention 1053] The method of any one of claims 1051 to 1052, further comprising the step of comparing the level of the protein or proteins to a reference level of each of said protein or proteins. [The present invention 1054] The method of any one of claims 10 to 15, further comprising the step of generating a molecular profile that identifies proteins having levels that differ from a reference level, e.g., that differ significantly from said reference level. [The present invention 1055] Any of the methods of claims 1029 to 1054, further comprising a step of predicting an increase or decrease in benefit of the cancer treatment based on the evaluated biomarkers, and optionally, the treatment comprises platinum-based chemotherapy or a combination therapy comprising platinum-based chemotherapy, optionally, the platinum-based chemotherapy comprises cisplatin, carboplatin, oxaliplatin, and / or nedaplatin, and the combination therapy comprising platinum-based chemotherapy comprises FOLFOX, FOLFOXIRI, and / or FOLFIRINOX. [The present invention 1056] predicting an increase or decrease in the benefit of the treatment, (a) the determined copy number of any of 1038 to 1048 of the present invention; and / or (b) Molecular profile of 1050 or 1054 of the present invention The method of the present invention 1055 is based on the above. [The present invention 1057] The method of claim 1056, wherein the step of predicting an increase or decrease in the benefit of treatment based on the determined copy number of any of claims 1038 to 1048 comprises the use of a voting module. [The present invention 1058] The method of the present invention 1057, wherein the voting module is any of the methods of the present invention 1014 to 1026. [The present invention 1059] The method of any one of claims 1057 to 1058, wherein the voting module includes the use of at least one random forest model. [The present invention 1060] Any of the methods of claims 1057 to 1059, wherein using the voting module comprises applying a machine learning classification model to the copy numbers obtained for each of group 2, group 6, group 7, group 8 and group 9, optionally wherein each machine learning classification model is a random forest model, and optionally wherein the random forest model is one described in Table 10. [The present invention 1061] The method of any of claims 1055 to 1060, wherein the subject has not been previously treated with said treatment. [The present invention 1062] The method of any one of claims 1029 to 1061, wherein the cancer comprises metastatic cancer, recurrent cancer, or a combination thereof. [The present invention 1063] The method of any of claims 1029 to 1062, wherein the subject has not previously undergone treatment for cancer. [The present invention 1064] The method of any one of claims 1055 to 1063, further comprising administering to the subject a treatment predicted to have increased benefit. [The present invention 1065] Any of the methods of claims 1055 to 1064, further comprising the step of not administering to the subject a treatment that is predicted to have reduced benefit. [The present invention 1066] The method of any one of claims 1064 to 1065, wherein progression free survival (PFS), disease free survival (DFS) or life span is extended by administration of said treatment. [The present invention 1067] Cancers include acute lymphoblastic leukemia; acute myeloid leukemia; adrenal cortical carcinoma; AIDS-related cancer; AIDS-related lymphoma; anal cancer; appendix cancer; astrocytoma; atypical teratoid / rhabdoid tumor; basal cell carcinoma; bladder cancer; brain stem glioma; brain tumor, brain stem glioma, central nervous system atypical teratoid / rhabdoid tumor, central nervous system embryonal tumor, astrocytoma, craniopharyngioma, ependymoblastoma, ependymoma, medulloblastoma, medulloepithelioma, intermediate pineal parenchymal tumor, supratentorial primitive neuroectodermal tumor and pineoblastoma; breast cancer; bronchial tumor; Burkitt's disease Lymphoma;Carcinoma of unknown primary (CUP);Carcinoid tumor;Carcinoma of unknown primary;Central nervous system atypical teratoid / rhabdoid tumor;Central nervous system embryonal tumor;Cervical cancer;Childhood cancer;Chordoma;Chronic lymphocytic leukemia;Chronic myelogenous leukemia;Chronic myeloproliferative disorder;Colon cancer;Colorectal cancer;Craniopharyngioma;Cutaneous T-cell lymphoma;Endocrine pancreatic islet cell tumor;Endometrial cancer;Ependymoblastoma;Ependymoma;Esophageal cancer;Nasal neuroblastoma;Ewing's sarcoma;Extracranial germ cell tumor;Extragonadal germ cell tumor;Extrahepatic bile duct cancer;Gallbladder cancer;Gastric cancer (stomach) cancer);gastrointestinal carcinoid tumor;gastrointestinal stromal cell tumor;gastrointestinal stromal tumor (GIST);gestational trophoblastic tumor;glioma;hairy cell leukemia;head and neck cancer;cardiac cancer;Hodgkin lymphoma;hypopharyngeal cancer;intraocular melanoma;pancreatic islet tumor;Kaposi's sarcoma;kidney cancer;Langerhans cell histiocytosis;laryngeal cancer;lip cancer;liver cancer;malignant fibrous histiocytoma bone cancer;medulloblastoma;medulloepithelioma;melanoma;Merkel cell carcinoma;Merkel cell skin cancer;mesothelioma;metastatic squamous neck cancer of unknown primary;oral cancer;multiple endocrine neoplasia syndrome;multiple myeloma;multiple myeloma / plasma cell neoplasm;mycosis fungoides;myelodysplastic syndrome;myeloproliferative neoplasms;nasal cancer;nasopharyngeal carcinoma;neuroblastoma;non-Hodgkin lymphoma;non-melanoma skin cancer;non-small cell lung cancer;oral cancer cancer);oral cavity cancer;oropharyngeal cancer;osteosarcoma;other brain and spinal cord tumors;ovarian cancer;ovarian epithelial cancer;ovarian germ cell tumors;ovarian low malignant potential tumors;pancreatic cancer;papillomatosis;sinoasal cancer;parathyroid cancer;pelvic cancer;penile cancer;pharyngeal cancer;intermediate pineal parenchymal tumors;pineoblastoma;pituitary tumors;plasma cell neoplasms / multiple myeloma;pleuropulmonary blastoma;primary central nervous system (CNS) lymphoma;primary hepatocellular carcinoma;The method of any of claims 1029 to 1066, comprising the treatment of: prostate cancer; rectal cancer; kidney cancer; renal cell (kidney) cancer; renal cell carcinoma; airway cancer; retinoblastoma; rhabdomyosarcoma; salivary gland cancer; Sézary syndrome; small cell lung cancer; small intestine cancer; soft tissue sarcoma; squamous cell carcinoma; cervical squamous cell carcinoma; stomach (gastric) cancer; supratentorial primitive neuroectodermal tumor; T-cell lymphoma; testicular cancer; laryngeal cancer; thymic cancer; thymoma; thyroid cancer; transitional cell carcinoma; transitional cell carcinoma of the renal pelvis and ureter; trophoblastic tumor; ureteral cancer; urethral cancer; uterine cancer; uterine sarcoma; vaginal cancer; vulvar cancer; Waldenstrom macroglobulinemia; or Wilms' tumor. [The present invention 1068] Cancers include acute myeloid leukemia (AML), breast cancer, cholangiocarcinoma, colorectal adenocarcinoma, extrahepatic bile duct adenocarcinoma, female genital malignancies, gastric adenocarcinoma, gastroesophageal adenocarcinoma, gastrointestinal stromal tumor (GIST), glioblastoma, head and neck squamous cell carcinoma, leukemia, hepatocellular carcinoma, low-grade glioma, pulmonary bronchioloalveolar carcinoma (BAC), non-small cell lung cancer (NSCLC), small cell lung cancer (SCLC), lymphoma, male genital malignancies, and malignant solitary fibrotic pleural carcinoma. The method of any of claims 1029 to 1066, comprising treating a patient with a tumor of the ovarian gland, a thymic carcinoma, a thyroid carcinoma, a thyroid cancer, a thyroid tumor, a thyroid cancer ... [The present invention 1069] The method of any of claims 1029 to 1066, wherein the cancer comprises colorectal cancer, ovarian cancer, esophageal cancer, esophagogastric junction cancer, gastric cancer, head and neck cancer, bladder cancer, breast cancer, endometrial cancer, uterine cancer, cervical cancer, pancreatic cancer, or lung cancer. [The present invention 1070] The method of claim 1069, further comprising determining a consensus molecular subtype (CMS) for the cancer, wherein the cancer comprises colorectal cancer. [The present invention 1071] 1. A method of selecting a treatment for a subject having cancer, comprising: Obtaining a biological sample containing cancer-derived cells; Next generation sequencing is performed on genomic DNA from the biological sample, (a) Group 2, including 1, 2, 3, 4, 5, 6, 7, or all 8 of MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; (b) group 6, including one, two, three, four or all five of BCL9, PBX1, PRRX1, INHBA and YWHAE; (c) group 7, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, or 15 of BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; (d)BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CR Group 8, including all 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 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, or 45 of EB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR, (e) Group 9, including 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, or all 11 of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11. determining the copy number for each of the genes or their adjacent genomic regions; applying a machine learning classification model to the copy numbers obtained for each of group 2, group 6, group 7, group 8 and group 9, optionally wherein each machine learning classification model is a random forest model, optionally wherein the random forest model is one described in Table 10; Obtaining an indication from each machine learning classification model of whether the subject is predicted to have an increased probability or a decreased probability of benefit from treatment with platinum-based chemotherapy; and selecting a platinum-based chemotherapy or a combination therapy including platinum-based chemotherapy if the majority of the machine learning classification models predict that the subject will have an increased probability of benefit from platinum-based chemotherapy, and selecting an alternative treatment to platinum-based chemotherapy or an additional treatment in combination with platinum-based chemotherapy if the majority of the machine learning classification models predict that the subject will have a decreased probability of benefit from platinum-based chemotherapy. The method comprising: [The present invention 1072] The method of claim 1071 further comprising the step of administering the selected treatment to the subject. [The present invention 1073] A method for generating a molecular profiling report, comprising the step of generating a report summarizing the results of carrying out any one of the methods of the present invention 1029 to 1072. [The present invention 1074] The report, (a) a prediction of an increase or decrease in the benefit of at least one of the treatments of the present invention 1055 to 1060; or (b) the selected treatment of invention 1071 or 1072 The method of the present invention 1073, comprising: [The present invention 1075] The method of any one of claims 1073 to 1074, wherein the report is computer generated; a printed report or a computer file; or accessible via a web portal. [The present invention 1076] 1. A system for identifying a treatment for cancer in a subject, comprising: (a) at least one host server; (b) at least one user interface for accessing the at least one host server for accessing and inputting data; (c) at least one processor for processing input data; (d) The data processed; (1) Access the results of analyzing any one of the biological samples of the present invention 1029 to 1072; and (2) Predicting an increase or decrease in the benefit of at least one treatment of any one of the inventions 1055 to 1060 or a selected treatment of the inventions 1071 or 1072 With instructions for at least one memory coupled to the processor for storing (e) at least one display for displaying a treatment for cancer, the treatment being platinum therapy; The system comprising: [The present invention 1077] The system of the present invention 1076, wherein at least one display includes a report including the results of analyzing the biological sample and treatments having promising benefits for treating the cancer or treatments selected for treating the cancer. Other features and advantages of the invention will become apparent from the following detailed description and drawings, and from the appended claims. [Brief description of the drawings]

[0083] [Figure 1A] FIG. 1 is a block diagram of an example of a prior art system for training a machine learning model. [Figure 1B] FIG. 1 is a block diagram of a system for generating a training data structure for training a machine learning model to predict the efficacy of a treatment for a disease or disorder in a subject having a particular set of biomarkers. [Figure 1C] FIG. 1 is a block diagram of a system for using a trained machine learning model to predict the efficacy of a treatment for a disease or disorder in a subject having a particular set of biomarkers. [Figure 1D] 1 is a flow chart of a process for generating training data for training a machine learning model to predict the efficacy of a treatment for a disease or disorder in a subject having a particular set of biomarkers. [Figure 1E]1 is a flow chart of a process of using a trained machine learning model to predict the efficacy of a treatment for a disease or disorder in a subject having a particular set of biomarkers. [Figure 1F] FIG. 1 is a block diagram of a system for predicting efficacy of a treatment for a disease or disorder in a subject having a particular set of biomarkers by interpreting outputs generated by multiple machine learning models using a voting unit. [Figure 1G] FIG. 6 is a block diagram of system components that can be used to implement the systems of FIGS. 2-5. [Figure 1H] FIG. 1 shows a block diagram of an exemplary embodiment of a system for determining personalized medical interventions for cancer utilizing molecular profiling of a patient's biological specimen. [Figure 2A] 1 is a flow chart of an exemplary embodiment of a method for determining personalized medical intervention for cancer utilizing molecular profiling of a patient biological specimen. [Figure 2B] 1 is a flow chart of an exemplary embodiment of a method for identifying a signature or molecular profile that can be used to predict benefit from a therapy. [Figure 2C] 2C is a flow chart of an exemplary embodiment of an alternative version of FIG. 2B. [Figure 3A] Pairwise hazard ratio graphs showing model performance using CNV profiling of eight markers in the setting of treatment with FOLFOX. CNA = copy number alteration. The eight markers were MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN and CDX2. [Figure 3B] Pairwise hazard ratio graphs showing model performance using CNV profiling of eight markers in the setting of treatment with FOLFIRI. CNA = copy number alteration. The eight markers were MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN and CDX2. [Figure 3C]Paired hazard ratio graphs showing model performance using CNV profiling of six markers with treatment with FOLFOX. The six markers were MYC, EP300, U2AF1, ASXL1, MAML2 and CNTRL. [Figure 3D] 1 is a paired hazard ratio graph showing model performance using CNV profiling of six markers in the setting of treatment with FOLFIRI. The six markers were MYC, EP300, U2AF1, ASXL1, MAML2, and CNTRL. [Figure 3E] FIG. 1 shows an exemplary random forest decision tree for the eight marker signatures shown in FIGS. 3A-B. [Figure 4A] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4B] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4C] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4D] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4E] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4F] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4G] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4H] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4I]We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4J] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4K] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4L] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4M] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4N] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 4O] We present the development of a biosignature to predict benefit of FOLFOX regimens in patients with metastatic colorectal cancer. [Figure 5A] We present the development of a biosignature to predict benefit of FOLFOX regimen in patients with colorectal cancer. [Figure 5B] We present the development of a biosignature to predict benefit of FOLFOX regimen in patients with colorectal cancer. [Figure 5C] We present the development of a biosignature to predict benefit of FOLFOX regimen in patients with colorectal cancer. [Figure 6A]We show consort diagrams of patient selection for training and cross-validating the FOLFOXai model (Figure 6A), real-world evidence ("RWE") laboratory data (Figure 6B), and TRIBE2 randomized Phase III cases (Figure 6C). Outcomes in Figure 6A were recently refreshed in September 2018, and outcomes in Figure 6B were recently refreshed in March 2020. We locked the FOLFOXai algorithm for 9 months in December 2018 and subsequently accepted the outcomes associated with Figure 6B. [Figure 6B] See legend to Figure 6A. [Figure 6C] See legend to Figure 6A. [Figure 6D] Blinded RWE cohorts receiving first-line FOLFOX (Figures 6D-6E) or first-line FOLFIRI (Figures 6F-6G) are shown for both time to next treatment (TTNT) and overall survival (OS). Figure 6D) Median TTNT in the benefit-increasing cohort was 3.1 months (37%) longer than median TTNT in the benefit-decreasing cohort (HR=0.508, 95% CI:0.390-0.665, log-rank p<0.001). Figure 6E) Median OS in the benefit-increasing cohort was 11.1 months (51%) longer than median OS in the benefit-decreasing cohort (HR=0.489, 95% CI:0.339-0.703, log-rank p<0.001). Figure 6F) Median TTNT in the increased benefit cohort was 4.4 months (35%) shorter than the median TTNT in the decreased benefit cohort (HR=1.757, 95% CI: 0.918-3.364, log-rank p=0.087). Figure 6G) Median OS in the increased benefit cohort was 11.6 months (38%) shorter than the median OS in the decreased benefit cohort (HR=2.829, 95% CI: 1.047-7.645, log-rank p=0.032). [Figure 6E] See legend to Figure 6D. [Figure 6F] See legend to Figure 6D. [Figure 6G] See legend to Figure 6D. [Figure 6H] The density of predictions in the RWE test cohort and the TRIBE2 study is shown. As shown in the plot, each figure shows a prediction of decreased benefit or increased benefit of the indicated treatment in the indicated cohort. The grey area in between represents a "no call" providing a buffer around the 50% IB probability threshold. Figure 6H) Metastatic CRC (mCRC) with FOLFOX in the RWE cohort. Figure 6I) mCRC with FOLFIRI in the RWE cohort. Figure 6J) mCRC with FOLFOX in the TRIBE2 cohort. Figure 6K) mCRC with FOLFOXIRI in the TRIBE2 cohort. Figure 6L) Pancreatic ductal adenocarcinoma (PDAC) with gemcitabine / nab-paclitaxel in the RWE cohort. Figure 6M) PDAC with FOLFIRINOX in the RWE cohort. Figure 6N) Advanced esophageal / esophagogastric junction cancer (EC / GEJC) with oxaliplatin in the RWE cohort. [Figure 6I] See legend to Figure 6H. [Figure 6J] See legend to Figure 6H. [Figure 6K] See legend to Figure 6H. [Figure 6L] See legend to Figure 6H. [Figure 6M] See legend to Figure 6H. [Figure 6N] See legend to Figure 6H. [Figure 6O]We show the performance of the algorithm in patients in arm A (FOLFOX, Figures 6O-6P) and arm B (FOLFOXIRI, Figures 6Q-6R) of the TRIBE2 trial. Figure 6O) The median progression-free survival (PFS) in the benefit-increasing cohort was 0.9 months (10%) longer than the median PFS in the benefit-decreasing cohort (HR=0.757, 95% CI:0.505-1.135, log-rank p=0.18). Figure 6P) The median OS in the benefit-increasing cohort was 6.0 months (32%) longer than the median OS in the benefit-decreasing cohort (HR=0.629, 95% CI:0.403-0.981, log-rank p=0.04). Figure 6Q) Median PFS in the increased benefit cohort was 6.2 months (82%) longer than median PFS in the decreased benefit cohort (HR=0.684, 95% CI:0.396-1.181, log-rank p=0.169). Figure 6R) Median OS in the increased benefit cohort was 14.1 months (89%) longer than median OS in the decreased benefit cohort (HR=0.483, 95% CI:0.270-0.864, log-rank p=0.02). [Figure 6P] See legend to Figure 6O. [Figure 6Q] See legend to Figure 6O. [Figure 6R] See legend to Figure 6O. [Figure 6S] Correlation between TTNT and PFS of cases in the TRIBE2 study is shown. FOLFOX is shown by gray circles and FOLFIRI is shown by black circles. [Figure 6T] Distribution of mCRC RWE TNT compared with PFS from the FOLFOX and FOLFOXIRI arms of the TRIBE2 trial is shown. Panel i) mCRC FOLFOX RWE; Panel ii) mCRC FOLFIRI RWE; Panel iii) mCRC FOLFOX TRIBE2; Panel iv) mCRC FOLFOXIRI TRIBE2. [Figure 6U]Cross-validated performance of training cases is shown. The median TTNT in the increased-benefit cohort was 3.9 months (52%) longer than the median TTNT in the decreased-benefit group (HR=0.398, 95% CI:0.244-0.647, log-rank p<0.001). [Figure 6V] We show the relative risks for TTNT (Figure 6V) and OS (Figure 6W) as a function of the probability of diminished benefit for patients, which we determine for both the FOLFOX and FOLFIRI RWE cohorts from a Cox proportional hazards model using the probability of diminished benefit and treatment as covariates. The crossover of the curves indicates that DB probability has a different risk effect depending on the treatment received, as evidenced by the inverse signals observed in the FOLFOX (see Figures 6D-6E) and FOLFIRI (see Figures 6F-6G) cohorts. [Figure 6W] See legend to Figure 6V. [Figure 6X] Kaplan-Meier plots of the probability of overall survival in patients with advanced PDAC receiving first-line FOLFIRINOX (Figure 6X) or nab-paclitaxel / gemcitabine (Figure 6Y) or GC / EGJC receiving oxaliplatin-containing regimens (Figure 6Z). Figure 6X) Median OS in the benefit-increasing cohort was 10.1 months (90%) longer than median OS in the benefit-decreasing cohort (HR=0.478, 95% CI:0.289-0.792, log-rank p=0.003). Figure 6Y) Median OS in the benefit-increasing cohort was 1 month (11%) longer than median OS in the benefit-decreasing cohort (HR=0.958, 95% CI:0.658-1.395, log-rank p=0.823). Figure 6Z) Median OS in the increased-benefit cohort was 5.1 months (58%) longer than median OS in the decreased-benefit cohort (HR=0.437, 95% CI:0.250-0.763, log-rank p=0.003). [Figure 6Y] See legend to Figure 6X. [Figure 6Z]See legend to Figure 6X. [Figure 7A] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7B] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7C] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7D] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7E] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7F] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7G] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7H] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7I] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7J] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7K]5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7L] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7M] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7N] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7O] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7P] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7Q] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7R] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7S] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7T] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7U] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7V]5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7W] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7X] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7Y] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7Z] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7AA] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7AB] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7AC] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7A-D] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7AE] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7AF] 5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. [Figure 7A-G]5A-C and 6A-Z to predict the benefit of platinum-based therapy in other settings. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0084] Detailed Description Described herein are methods and systems for identifying therapeutic agents for use in personalized-based treatment by using molecular profiling, including systems, methods, devices and computer programs for training machine learning models and then using the trained machine learning models to predict the effectiveness of treatment for a subject's disease or disorder.In some embodiments, the system can include one or more computer programs on one or more computers at one or more locations, configured for use in, for example, the methods described herein.

[0085] An aspect of the present disclosure relates to a system that generates a set of one or more training data structures that can be used to train a machine learning model to provide various classifications, such as characterizing the phenotype of a biological sample. The characterization of the phenotype can include providing a diagnosis, prognosis, theranosis, or other related classification. For example, the classification can be a classification that predicts the disease state of a subject having a particular set of biomarkers or the effectiveness of a treatment for a disease or disorder. When trained, the trained machine learning model can be used to process input data provided by the system and make a prediction based on the processed input data. The input data can include a set of features related to the subject, such data representing one or more subject biomarkers, and the data representing a disease or disorder. In some embodiments, the input data can further include features representing a proposed type of treatment, and make a prediction that describes the likely response of the subject to the treatment. The prediction can include data output by the machine learning model based on the processing of the particular set of features provided to the machine learning model as input. The data may optionally include data representing one or more subject biomarkers, data representing a disease or disorder, and data representing a proposed type of treatment.

[0086] An innovative aspect of the present disclosure includes the extraction of specific data from the incoming data stream for use in generating a training data structure. Of great importance is the selection of a specific set of one or more biomarkers for inclusion in the training data structure. The reason is that the presence, absence, or status of a specific biomarker may indicate a desired classification. For example, a specific biomarker may be selected to determine whether a treatment for a disease or disorder is effective or ineffective. By way of illustration, in the present disclosure, applicants present a set of specific biomarkers that, when used in training a machine learning model, gives rise to a trained model that can predict treatment efficiency more accurately than if a different set of biomarkers were used. See Examples 2-4.

[0087] The system is configured to obtain output data generated by the trained machine learning model based on the processing of the data by the machine learning model. In various aspects, the data includes biological data representing one or more biomarkers, data representing a disease or disorder, and data representing a treatment type. The system can then predict the effectiveness of a treatment for a subject having a particular set of biomarkers. In some embodiments, the disease or disorder can include a type of cancer, and the treatment for the subject can include one or more therapeutic agents, such as small molecule drugs, biologics, and various combinations thereof. In this setting, the output of the trained machine learning model generated based on the processing of the input data including a set of biomarkers, a disease or disorder, and a treatment type includes data representing the level of responsiveness that the subject exhibits to a treatment for the disease or disorder.

[0088] In some embodiments, the output data generated by the trained machine learning model may include a probability of a desired classification. By way of illustration, such a probability may be the probability that a subject will respond favorably to a treatment for a disease or disorder. In other embodiments, the output data may include any output data generated by the trained machine learning model based on processing the trained machine learning model of input data. In some aspects, the input data includes a set of biomarkers, data representing a disease or disorder, and data representing a treatment type.

[0089] In some embodiments, the training data structure generated by the present disclosure may include a plurality of training data structures, each of which includes a field representing a feature vector corresponding to a particular training sample. The feature vector includes a set of features derived from and representing the training sample. The training sample may include, for example, one or more biomarkers of the subject, a disease or disorder of the subject, and a proposed treatment for the disease or disorder. The training data structure is flexible, since each training data structure may be assigned a weight representing each feature of the feature vector. Thus, each training data structure of the plurality of training data structures may be specially configured to allow a particular inference to be made by the machine learning model during training.

[0090] Consider a non-limiting example in which a model is trained to make predictions of the likely benefit of a particular treatment for a disease or disorder. As a result, the novel training data structure generated in accordance with the present specification is designed to improve the performance of a machine learning model, because it can be used to train the machine learning model to predict the effectiveness of a treatment for a disease or disorder of a subject having a particular set of biomarkers. By way of illustration, a machine learning model that was not able to perform predictions regarding the effectiveness of a treatment for a disease or disorder of a subject having a particular set of biomarkers before being trained using the training data structure, system, and operation described by the present disclosure can learn to make predictions regarding the effectiveness of a treatment for the disease or disorder of the subject by being trained using the training data structure, system, and operation described by the present disclosure. Thus, this process takes an otherwise general-purpose machine learning model and transforms it into a specialized computer for performing the specific task of predicting the effectiveness of a treatment for a disease or disorder of a subject having a particular set of biomarkers.

[0091] FIG. 1A is a block diagram of an example of a prior art system 100 for training a machine learning model 110. In some embodiments, the machine learning model may be, for example, a support vector machine. Alternatively, the machine learning model may include a neural network model, a linear regression model, a random forest model, a logistic regression model, a naive Bayes model, a quadratic discriminant analysis model, a k-nearest neighbor model, a support vector machine, etc. The machine learning model training system 100 may be implemented as a computer program on one or more computers at one or more locations on which the systems, components, and techniques described below may be implemented. The machine learning model training system 100 trains the machine learning model 110 using training data items from a database (or dataset) 120 of training data items. The training data items may include multiple feature vectors. Each training vector may include multiple values, each corresponding to a particular feature of the training sample that the training vector represents. The training features may also be referred to as independent variables. In addition, the system 100 maintains a respective weight for each feature included in the feature vector.

[0092] The machine learning model 110 is configured to receive input training data items 122 and process the input training data items 122 to generate output 118. The input training data items may include multiple features (or independent variables "X") and training labels (or dependent variables "Y"). The machine learning model may be trained using the training items and, when trained, is capable of predicting X=f(Y).

[0093] To enable the machine learning model 110 to generate accurate outputs for received data items, the machine learning model training system 100 may train the machine learning model 110 to adjust values ​​of parameters of the machine learning model 110, e.g., to determine trained values ​​of the parameters from initial values. These parameters derived from the training process may include weights that can be used during a prediction stage using the fully trained machine learning model 110.

[0094] When training the machine learning model 110, the machine learning model training system 100 uses training data items stored in a database (dataset) 120 of labeled training data items. The database 120 stores a set of training data items, with each training data item in the set of training data items associated with a respective label. In general, the label for the training data item identifies the correct classification (or prediction) for the training data item, i.e., the classification that should be identified as the classification of the training data item by the output value generated by the machine learning model 110. Referring to FIG. 1A, a training data item 122 may be associated with a training label 122a.

[0095] The machine learning model training system 100 trains the machine learning model 110 to optimize an objective function. Optimizing the objective function may include, for example, minimizing a loss function 130. In general, the loss function 130 is a function that depends on (i) the output 118 generated by the machine learning model 110 by processing a given training data item 122, and (ii) the label 122a for the training data item 122, i.e., the target output that the machine learning model 110 should have generated by processing the training data item 122.

[0096] A conventional machine learning model training system 100 may train the machine learning model 110 to minimize a (cumulative) loss function 130 by performing multiple iterations of conventional machine learning model training techniques, such as hinge loss, stochastic gradient methods, stochastic gradient descent with backpropagation, etc., on training data items from a database 120 to iteratively adjust the values ​​of the parameters of the machine learning model 110. The fully trained machine learning model 110 may then be deployed as a predictive model that can be used to make predictions based on unlabeled input data.

[0097] FIG. 1B is a block diagram of a system 200 for generating a training data structure for training a machine learning model to predict the efficacy of a treatment for a disease or disorder in a subject having a particular set of biomarkers.

[0098] The system 200 includes two or more distributed computers 210, 310, a network 230, and an application server 240. The application server 240 includes an extraction unit 242, a memory unit 244, a vector generation unit 250, and a machine learning model 270. The machine learning model 270 may include one or more of a vector support machine, a neural network model, a linear regression model, a random forest model, a logistic regression model, a naive Bayes model, a quadratic discriminant analysis model, a k-nearest neighbor model, a support vector machine, and the like. Each distributed computer 210, 310 may include a smartphone, a tablet computer, a laptop computer, or a desktop computer, and the like. Alternatively, the distributed computers 210, 310 may include a server computer that receives data entered by one or more terminals 205, 305, respectively. The terminal computers 205, 305 may include any user device, including a smartphone, a tablet computer, a laptop computer, a desktop computer, and the like. The network 230 may include one or more networks 230, such as a LAN, a WAN, a wired Ethernet network, a wireless network, a cellular network, the Internet, or any combination thereof.

[0099] The application server 240 is configured to obtain or otherwise receive data records 220, 222, 224, 320 provided by one or more distributed computers, such as the first distributed computer 210 and the second distributed computer 310, using the network 230. In some embodiments, each distributed computer 210, 310 may provide different types of data records 220, 222, 224, 320. For example, the first distributed computer 210 may provide biomarker data records 220, 222, 224 representing subject biomarkers, and the second distributed computer 310 may provide outcome data 320 representing subject outcome data obtained from the outcome database 312.

[0100] The biomarker data records 220, 222, 224 may include any type of biomarker data describing a biometric attribute of a subject. Illustratively, the example of FIG. 1B shows the biomarker data records as including data records representing DNA biomarkers 220, protein biomarkers 222, and RNA data biomarkers 224. Each of these biomarker data records may include a data structure having fields that structure information 220a, 222a, 224a describing a biomarker of a subject, such as a DNA biomarker 220a, a protein biomarker 222a, or an RNA biomarker 224a of a subject. However, the disclosure need not be so limited. For example, the biomarker data records 220, 222, 224 may include next generation sequencing data, such as DNA alterations. Such next generation sequencing data may include single variants, insertions and deletions, substitutions, translocations, fusions, truncations, duplications, amplifications, losses, copy numbers, repeats, total gene mutation dose, microsatellite instability, and the like. Next generation sequencing data may be for a particular set of genes or any other desired genomic loci, or may include whole exome or whole genome data. Alternatively or additionally, biomarker data records 220, 222, 224 may also include in situ hybridization data, such as DNA copies. Such in situ hybridization data may include gene copies, gene translocations, and the like. Alternatively or additionally, biomarker data records 220, 222, 224 may include RNA data, such as gene expression or gene fusions, including whole transcriptome sequencing. Alternatively or additionally, biomarker data records 220, 222, 224 may include protein expression data, such as obtained using immunohistochemistry (IHC). Alternatively or additionally, biomarker data records 220, 222, 224 may include ADAPT data, such as complex numbers.

[0101] In some embodiments, the set of one or more biomarkers includes one or more biomarkers set forth in any one of Tables 2-8. However, the disclosure need not be so limited, and other types of biomarkers may be used instead. For example, biomarker data may be obtained by whole exome sequencing, whole genome sequencing, whole transcriptome sequencing, or any useful combination thereof.

[0102] The outcome data record 320 may describe the outcome of a treatment for a subject. For example, the outcome data record 320 obtained from the outcome database 312 may include one or more data structures with fields structuring data attributes of the subject, such as the disease or disorder 320a, the treatment the subject received for the disease or disorder 320a, the treatment result 320a, or a combination of both. In addition, the outcome data record 320 may also include fields structuring data attributes describing the details of the treatment and the subject's response to the treatment. An example of a disease or disorder may include, for example, certain cancers. The type of treatment may include, for example, the type of drug, biologic, or other treatment the subject received for the disease or disorder included in the outcome data record 320. The treatment result may include data representing the subject's outcome of the treatment regimen, such as benefit, moderate benefit, or no benefit. In some embodiments, the treatment result may include the type of cancerous tumor at the end of the treatment, such as the amount the tumor shrank, the overall size of the tumor after treatment, etc. Alternatively or additionally, the treatment outcome may include counts or ratios of white blood cells, red blood cells, etc. Treatment details may include dosage, e.g., amount of medication taken, drug regimen, number of missed doses, etc. Thus, while the example of FIG. 1B shows that the outcome data may include disease or disorder, treatment, and treatment outcome, the outcome data may include other types of information as described herein. Moreover, the outcome data need not be limited to human "patients." Instead, the outcome data records 220, 222, 224 and biometric data records 320 may be associated with any desired subject, including any non-human organism.

[0103] In some embodiments, each of the data records 220, 222, 224, 320 may include keyed data that allows the data records from each distributed computer to be correlated by the application server 240. The keyed data may include, for example, data representing a subject identifier. The subject identifier may include any form of data that identifies a subject that can associate the subject's biomarkers with the subject's outcome data.

[0104] The first distributed computer 210 may provide 208 the biomarker data records 220, 222, 224 to the application server 240. The second distributed computer 310 may provide 210 the outcome data records 320 to the application server 240. The application server 240 may provide the biomarker data records 220 and the outcome data records 220, 222, 224 to the extraction unit 242.

[0105] The extraction unit 242 may process the received biomarker data 220, 222, 224 and outcome data records 320 to extract data 220a-1, 222a-1, 224a-1, 320a-1, 320a-2, 320a-3 that may be used to train a machine learning model. For example, the extraction unit 242 may obtain data structured by fields of the data structure of the biometric data records 220, 222, 224, data structured by fields of the data structure of the outcome data record 320, or a combination thereof. The extraction unit 242 may perform one or more information extraction algorithms, such as keyed data extraction, pattern matching, natural language processing, etc., to identify and obtain the data 220a-1, 222a-1, 224a-1, 320a-1, 320a-2, 320a-3 from the biometric data records 220, 222, 224 and outcome data records 320, respectively. The extraction unit 242 may provide the extracted data to a memory unit 244. The extracted data may be stored in the memory unit 244, such as a flash memory (as opposed to a hard disk), to improve data access time and reduce latency in accessing the extracted data to improve system performance. In some embodiments, the extracted data may be stored in the memory unit 244 as an in-memory data grid.

[0106] More specifically, the extraction unit 242 may be configured to filter the portions of the biomarker data records 220, 222, 224 and outcome data records 320 used to generate the input data structure 260 for processing by the machine learning model 270 from the portions of the outcome data records 320 used as labels for the generated input data structure 260. Such filtering includes the extraction unit 242 separating the biomarker data and a first portion of the outcome data including the disease or disorder, treatment, treatment details, or a combination thereof from the treatment results. The application server 240 may then generate the input data structure 260 using the biomarker data 220a-1, 222a-1, 224a-1, 320a-1, 320a-2 and the first portion of the outcome data including the disease or disorder 320a-1, treatment 320a-2, treatment details (not shown in FIG. 1B), or a combination thereof. In addition, the application server 240 may use a second portion of the outcome data describing the treatment result 320a-3 as a label for the generated data structure.

[0107] The application server 240 may process the extracted data stored in the memory unit 244 and correlate the biomarker data 220a-1, 222a-1, 224a-1 extracted from the biomarker data records 220, 222, 224 with a first portion of the outcome data 320a-1, 320a-2. The purpose of this correlation is to cluster the biomarker data with the outcome data such that the subject's outcome data is clustered with the subject's biomarker data. In some embodiments, the correlation of the biomarker data with the first portion of the outcome data may be based on keyed data associated with each of the biomarker data records 220, 222, 224 and the outcome data record 320. For example, the keyed data may include a subject identifier.

[0108] The application server 240 provides the extracted first portions of the extracted biomarker data 220a-1, 222a-1, 224a-1 and outcome data 320a-1, 320a-2 as input to the vector generation unit 250. The vector generation unit 250 is used to generate a data structure based on the extracted first portions of the extracted biomarker data 220a-1, 222a-1, 224a-1 and outcome data 320a-1, 320a-2. The generated data structure is a feature vector 260 that includes a plurality of values ​​that numerically represent the extracted first portions of the extracted biomarker data 220a-1, 222a-1, 224a-1 and outcome data 320a-1, 320a-2. The feature vector 260 may include a field for each type of biomarker and each type of outcome data. For example, feature vector 260 may include one or more fields corresponding to (i) one or more types of next generation sequencing data, such as single variants, insertions and deletions, substitutions, translocations, fusions, truncations, duplications, amplifications, losses, copy number, repeats, total gene mutation dose, microsatellite instability, (ii) one or more types of in situ hybridization data, such as DNA copies, gene copies, gene translocations, (iii) one or more types of RNA data, such as gene expression or gene fusions, (iv) one or more types of protein data, such as obtained using immunohistochemistry, (v) one or more types of ADAPT data, such as complex numbers, and (vi) one or more types of outcome data, such as a disease or disorder, treatment type, details of each treatment type, etc.

[0109] The vector generation unit 250 is configured to assign a weight to each field of the feature vector 260 that indicates the extent to which the extracted first portions of the extracted biomarker data 220a-1, 222a-1, 224a-1 and outcome data 320a-1, 320a-2 contain the data represented by each field. In one embodiment, for example, the vector generation unit 250 may assign a "1" to each field of the feature vector that corresponds to features found in the extracted first portions of the extracted biomarker data 220a-1, 222a-1, 224a-1 and outcome data 320a-1, 320a-2. In such an embodiment, the vector generation unit 250 may also assign a "0" to each field of the feature vector that corresponds to features not found in the extracted first portions of the extracted biomarker data 220a-1, 222a-1, 224a-1 and outcome data 320a-1, 320a-2, for example. The output of the vector generation unit 250 may include a data structure, such as a feature vector 260, that can be used to train a machine learning model 270.

[0110] The application server 240 can label the training feature vector 260. Specifically, the application server can use the extracted second portion of the patient outcome data 320a-3 to label the generated feature vector 260 with a treatment outcome 320a-3. The label of the training feature vector 260 generated based on the treatment outcome 320a-3 can provide an indication of the effectiveness of the treatment 320a-2 for the disease or disorder 320a-1 of interest as determined by a particular set of biomarkers 220a-1, 222a-1, 224a-1 (each of which is described in the training data structure 260).

[0111] The application server 240 can train the machine learning model 270 by providing the feature vector 260 as an input to the machine learning model 270. The machine learning model 270 can process the generated feature vector 260 and generate an output 272. The application server 240 can use a loss function 280 to determine the amount of error between the output 272 of the machine learning model 280 and the values ​​specified by the training labels (generated based on a second portion of the extracted patient outcome data describing the treatment results 320a-3). The output 282 of the loss function 280 can be used to adjust parameters of the machine learning model 282.

[0112] In some embodiments, adjusting the parameters of the machine learning model 270 may include manual tuning of the machine learning model parameters. Alternatively, in some embodiments, the parameters of the machine learning model 270 may be tuned automatically by one or more algorithms executed by the application server 242.

[0113] The application server 240 may perform multiple iterations of the process described above with reference to FIG. 1B for each outcome data record 320 stored in the outcome database corresponding to the subject's set of biomarker data. This may include hundreds of iterations, thousands of iterations, tens of thousands of iterations, hundreds of thousands of iterations, millions of iterations, or more iterations until each outcome data record 320 having a corresponding set of biomarker data for the subject stored in the outcome database 312 is exhausted, until the machine learning model 270 is trained to within a particular error range, or a combination thereof. The machine learning model 270 is trained to within a particular error range, for example, when the machine learning model 270 can predict the efficacy of a treatment for a subject having biomarker data based on the set of unlabeled biomarker data, the disease or disorder data, and the treatment data. The efficacy may include, for example, a probability, a general indication of whether the treatment is successful or unsuccessful, etc.

[0114] FIG. 1C is a block diagram of a system for using a trained machine learning model to predict the efficacy of a treatment for a disease or disorder in a subject having a particular set of biomarkers.

[0115] The machine learning model 370 includes a machine learning model trained using the process described with reference to the system of FIG. 1B above. The trained machine learning model 370 can predict a level of efficacy of a treatment in treating a disease or disorder in a subject having a biomarker based on an input feature vector representing a set of one or more biomarkers, a disease or disorder, and a treatment. In some embodiments, the "treatment" may include a drug, details of the treatment (e.g., dosage, regimen, missed doses, etc.), or any combination thereof.

[0116] The application server 240 hosting the machine learning model 370 is configured to receive unlabeled biomarker data records 320, 322, 324. The biomarker data records 320, 322, 324 include one or more data structures having fields that structure data representing one or more particular biomarkers, such as DNA biomarkers 320a, protein biomarkers 322a, RNA biomarkers 324a, or any combination thereof. As mentioned above, the received biomarker data records may include biomarkers of types not depicted by FIG. 1C, such as (i) one or more types of next generation sequencing data, such as single variants, insertions and deletions, substitutions, translocations, fusions, truncations, duplications, amplifications, losses, copy number, repeats, total gene mutation dose, microsatellite instability, (ii) one or more types of in situ hybridization data, such as DNA copies, gene copies, gene translocations, (iii) one or more types of RNA data, such as gene expression or gene fusions, (iv) one or more types of protein data, such as obtained using immunohistochemistry, or (v) one or more types of ADAPT data, such as complex numbers.

[0117] The application server 240 hosting the machine learning model 370 is also configured to receive data representing suggested treatment data 422a for a disease or disorder described by the disease or disorder data 420a for subjects having the biomarkers represented by the received biomarker data records 320, 322, 324. The suggested treatment data 422a for the disease or disorder 422a is also unlabeled and is merely a suggestion for treating a subject having the biomarkers represented by the biomarker data records 320, 322, 324.

[0118] In some embodiments, the disease or disorder data 420a and the proposed treatment 422a are provided 305 by the terminal 405 over the network 230, and the biomarker data is obtained from the second distributed computer 310. The biomarker data may be derived from laboratory equipment used to perform various assays. In other embodiments, the disease or disorder data 420a, the proposed treatment 422a, and the biomarker data 320, 322, 324 may each be received from the terminal 405. For example, the terminal 405 may be a user device of a physician, an employee working for the physician or a representative of the physician, or any other person who inputs data representing the disease or disorder, data representing the proposed treatment, and data representing one or more biomarkers of a subject having the disease or disorder. In some embodiments, the treatment data 422 may include a data structure structuring fields of data representing the proposed treatment described by drug name. In other embodiments, the treatment data 422 may include a data structure structuring fields of data representing more complex treatment data, such as dosage, dosing regimen, number of missed doses allowed, etc.

[0119] The application server 240 receives the biomarker data records 320, 322, 324, the disease or disorder data 420, and the treatment data 422. The application server 240 provides the biomarker data records 320, 322, 324, the disease or disorder data 420, and the treatment data 422 to an extraction unit 242, which is configured to extract (i) specific biomarker data, e.g., DNA biomarker data 320a-1, protein expression data 322a-1, 324a-1, (ii) disease or disorder data 420a-1, and (iii) suggested treatment data 420a-1 from fields of the biomarker data records 320, 322, 324 and the outcome data records 420, 422. In some embodiments, the extracted data is stored in the memory unit 244 as a buffer, cache, etc., and then provided as input to the vector generation unit 250 when the vector generation unit 250 has the bandwidth to receive the input for processing. In other embodiments, the extracted data is provided directly to vector generation unit 250 for processing. For example, in some embodiments, multiple vector generation units 250 may be used to allow parallel processing of inputs to reduce latency.

[0120] The vector generation unit 250 can generate a data structure, such as a feature vector 360, that includes multiple fields, including one or more fields for each type of biomarker data and one or more fields for each type of outcome data. For example, each field of the feature vector 360 can correspond to (i) each type of extracted biomarker data that can be extracted from the biomarker data records 320, 322, 324, such as each type of next generation sequencing data, each type of in situ hybridization data, each type of RNA data, each type of immunohistochemistry data, and each type of ADAPT data, and (ii) each type of outcome data that can be extracted from the outcome data records 420, 422, such as each type of disease or disorder, each type of treatment, and each type of treatment details.

[0121] The vector generation unit 250 is configured to assign a weight to each field of the feature vector 360 that indicates the extent to which the extracted biomarker data 320a-1, 322a-1, 324a-1, extracted disease or disorder 420a-1, and extracted treatment 422a-1 include the data represented by each field. In one embodiment, for example, the vector generation unit 250 may assign a "1" to each field of the feature vector 360 that corresponds to features found in the extracted biomarker data 320a-1, 322a-1, 324a-1, extracted disease or disorder 420a-1, and extracted treatment 422a-1. In such an embodiment, the vector generation unit 250 may also assign a "0" to each field of the feature vector that corresponds to features not found in the extracted biomarker data 320a-1, 322a-1, 324a-1, extracted disease or disorder 420a-1, and extracted treatment 422a-1, for example. The output of the vector generation unit 250 may include a data structure such as a feature vector 360, which can be provided as input to a trained machine learning model 370.

[0122] The trained machine learning model 370 processes the generated feature vector 360 based on the adjusted parameters determined during the training phase and described with reference to FIG. 1B. The output 272 of the trained machine learning model provides an indication of the effectiveness of the treatment 422a-1 of the disease or disorder 420a-1 of the subject having the biomarkers 320a-1, 322a-1, 324a-1. In some embodiments, the output 272 may include a probability indicating the effectiveness of the treatment 422a-1 of the disease or disorder 420a-1 of the subject having the biomarkers 320a-1, 322a-1, 324a-1. In such embodiments, the output 272 may be provided 311 to the terminal 405 using the network 230. The terminal 405 may then generate an output on the user interface 420 indicating a predicted level of effectiveness of the treatment of the disease or disorder of the person having the biomarkers represented by the feature vector 360.

[0123] In other embodiments, the output 272 may be provided to a prediction unit 380 configured to decipher the meaning of the output 272. For example, the prediction unit 380 can be configured to map the output 272 to one or more categories of validity. The output of the prediction unit 328 can then be used as part of a message 390 provided 311 to the terminal 305 using the network 230 for review by the subject, the subject's guardian, a nurse, a doctor, etc.

[0124] 1D is a flowchart of a process 400 for generating training data for training a machine learning model to predict the effectiveness of a treatment for a disease or disorder of a subject having a particular set of biomarkers. In one aspect, the process 400 may include obtaining a first data structure from a first distributed data source, the first data structure including fields structuring data representing a set of one or more biomarkers associated with the subject (410), storing the first data structure in one or more memory devices (420), obtaining a second data structure from a second distributed data source, the second data structure including fields structuring data representing outcome data of the subject having one or more biomarkers (430), storing the second data structure in one or more memory devices (440), generating a labeled training data structure based on the first data structure and the second data structure, the labeled training data structure including (i) data representing one or more biomarkers, (ii) a disease or disorder, (iii) a treatment, and (iv) the effectiveness of the treatment for the disease or disorder (450), and training a machine learning model using the generated labeled training data (460).

[0125] 1E is a flow chart of a process 500 using a trained machine learning model to predict the efficacy of a treatment for a disease or disorder of a subject having a particular set of biomarkers. In one aspect, the process 500 may include obtaining a data structure representing a set of one or more biomarkers associated with a subject (510), obtaining data representing a disease or disorder type of the subject (520), obtaining data representing a treatment type for the subject (530), generating a data structure for input to a machine learning model representing (i) the one or more biomarkers, (ii) the disease or disorder, and (iii) the treatment type (540), providing the generated data structure as an input to a machine learning model trained using labeled training data representing the one or more obtained biomarkers, the one or more treatment types, and the one or more diseases or disorders (550), obtaining an output generated by the machine learning model based on the machine learning model processing of the provided data structure (560), and determining a predicted outcome for the treatment of a disease or disorder of the subject having the one or more biomarkers based on the obtained output generated by the machine learning model (570).

[0126] Provided herein is a method for improving classification performance using multiple machine learning models. Conventionally, a single model is selected to perform the desired prediction / classification. For example, during the training phase, various model parameters or model types, such as random forests, support vector machines, logistic regression, k-nearest neighbors, artificial neural networks, naive Bayes, quadratic discriminant analysis, or Gaussian process models, may be compared to identify the model with the optimal desired performance. Applicants have realized that the selection of a single model may not provide optimal performance in every setting. Instead, multiple models can be trained to perform prediction / classification, and a joint prediction can be used to perform classification. In this scenario, each model is allowed to "vote," and the classification that receives the majority of votes is considered the winner.

[0127] The voting strategy disclosed herein can be applied to any machine learning classification, including both model building (e.g., using training data) and applications for classifying naive samples. Such settings include, but are not limited to, data in the fields of biology, finance, communications, media and entertainment. In some preferred embodiments, the data is high-dimensional "big data". In some embodiments, the data includes biological data, including biological data obtained by molecular profiling as described herein. See, e.g., Example 1. Molecular profiling data can include, but is not limited to, high-dimensional next generation sequencing data for a particular biomarker panel (see, e.g., Example 1) or whole-exome, whole-genome and / or whole-transcriptome data. The classification can be any classification useful, e.g., to characterize a phenotype. For example, the classification can provide a diagnosis (e.g., diseased or healthy), a prognosis (e.g., predicting a good or bad outcome) or a theranostic (e.g., predicting or monitoring treatment efficacy or lack thereof). Examples of the application of the voting strategy are provided herein in Examples 2-4.

[0128] 1F is a block diagram of a system 600 for interpreting outputs generated by multiple machine learning models by using a voting unit. System 600 is similar to system 300 of FIG. 1C. However, instead of a single machine learning model 370, system 600 includes multiple machine learning models 370-0, 370-1...370-x (where x is any non-zero integer greater than 1). In addition, system 600 also includes a voting unit 480.

[0129] As a non-limiting example, system 600 can be used to predict the efficacy of a treatment for a disease or disorder in a subject having a particular set of biomarkers. See Examples 2-4.

[0130] Each machine learning model 370-0, 370-1, 370-x may include a machine learning model trained to classify a particular type of input data 320-0, 320-1... 320-x (where x is any non-zero integer greater than 1 and equal to the number x of machine learning models). In some embodiments, each of the machine learning models 370-0, 370-1, 370-x may be of the same type. For example, each of the machine learning models 370-0, 370-1, 370-x may be a random forest classification algorithm, for example, trained using different parameters. In other embodiments, the machine learning models 370-0, 370-1, 370-x may be of different types. For example, there may be one or more random forest model classifiers, one or more neural networks, one or more k-nearest neighbor classifiers, other types of machine learning models, or any combination thereof.

[0131] Input data, such as input data 0 (320-0), input data 1 (320-1), input data x (320-x), etc., may be obtained by the application server 240. In some embodiments, the input data 320-0, 320-1, 320-x may be obtained from one or more distributed computers 310, 405 over the network 230. Illustratively, one or more of the input data items 320-0, 320-1, 320-x may be generated by correlating data from multiple different data sources 210, 405. In such an embodiment, (i) first data describing a subject's biomarkers may be obtained from a first distributed computer 310, and (ii) second data describing a disease or disorder and associated treatment may be obtained from a second computer 405. The application server 240 may correlate the first data and the second data to generate an input data structure, such as input data structure 320-0. This process is described in further detail in FIG. 1C. The input data items 320-0, 320-1, 320-x may be provided sequentially, one at a time, as respective inputs, to, for example, a vector generation unit. The vector generation unit may generate input vectors 360-0, 360-1, 360-x corresponding to the respective input data 320-0, 320-1, 320-x. Although some embodiments may generate the vectors 360-0, 360-1, 360-x sequentially, the disclosure need not be so limited.

[0132] Alternatively, in some embodiments, the vector generation unit 250 may be configured to operate multiple parallel vector generation units that may parallelize the vector generation process. In such embodiments, the vector generation unit 250 may receive input data 320-0, 320-1, 320-x in parallel, process the input data 320-0, 320-1, 320-x in parallel, and generate respective vectors 360-0, 360-1, 360-x in parallel, each corresponding to one of the input data 320-0, 320-1, 320-x.

[0133] In some embodiments, vectors 360-0, 360-1, 360-x may each be generated based on a respective corresponding input data, such as input data 320-0, 320-1, 320-x. That is, vector 360-0 is generated based on input data 320-0 and represents input data 320-0. Similarly, vector 360-1 is generated based on input data 320-1 and represents input data 320-1. Similarly, vector 360-x is generated based on input data 320-x and represents input data 320-x.

[0134] In some embodiments, each input data structure 320-0, 320-1, 320-x can include data representing a biomarker of the subject, data describing a disease or disorder associated with the subject, data describing a proposed treatment for the subject, or any combination thereof. The data representing the biomarkers of the subject can include data describing a particular subset or panel of genes from the subject. Alternatively, in some embodiments, the data representing the biomarkers of the subject can include data representing a complete set of known genes for the subject. The complete set of known genes for the subject can include all of the genes of the subject. In some embodiments, each of the machine learning models 370-0, 370-1, 370-x is the same type of machine learning model, e.g., a neural network trained to classify input data vectors as corresponding to subjects likely to respond or not likely to respond to a treatment identified as associated by the vector processed by the machine learning model. In such an embodiment, each of the machine learning models 370-0, 370-1, 370-x is the same type of machine learning model, but each of the machine learning models 370-0, 370-1, 370-x may be trained in a different manner. The machine learning models 370-1, 370-1, 370-x may generate output data 272-0, 272-1, 272-x, respectively, that represent whether a subject associated with the input vector 360-0, 360-1, 360-x is likely to respond or not respond to a treatment associated with the input vector 360-0, 360-1, 360-x. In this example, the input data sets and their corresponding input vectors are the same. For example, each set of input data has the same biomarkers, the same disease or disorder, the same treatment, or any combination.Nonetheless, given the various training methods used to train each machine learning model 370-0, 370-1, 370-x, different outputs 272-0, 272-1, 272-x may be generated based on each machine learning model 370-0, 370-1, 370-x processing the input vectors 360-0, 361-1, 361-x, respectively, as shown in FIG. 1F.

[0135] Alternatively, each of the machine learning models 370-0, 370-1, 370-x can be a different type of machine learning model trained or otherwise configured to classify input data as representing subjects likely to respond or likely not respond to a treatment for a disease or disorder. For example, the first machine learning model 370-1 can include a neural network, the machine learning model 370-1 can include a random forest classification algorithm, and the machine learning model 370-x can include a k-nearest neighbor algorithm. In this example, each of these different types of machine learning models 370-0, 370-1, 370-x can be trained or otherwise configured to receive and process an input vector and determine whether the input vector is associated with a subject likely to respond or likely not respond to a treatment also associated with the input vector. In this example, the input data sets and their corresponding input vectors can be the same. For example, each set of input data has the same biomarkers, the same disease or disorder, the same treatment, or any combination. Thus, machine learning model 370-0 can be a neural network trained to process input vector 360-0 and generate output data 272-0 indicative of whether a subject associated with input vector 360-0 is likely to respond or not respond to a treatment also associated with input vector 360-0. Additionally, machine learning model 370-1 can be a random forest classification algorithm trained to process input vector 360-1, which in this example is the same as input vector 360-0, and generate output data 272-1 indicative of whether a subject associated with input vector 360-1 is likely to respond or not respond to a treatment also associated with input vector 360-1. This input vector analysis method can continue with each of the x inputs, x input vectors, and x machine learning models.Continuing with the example with reference to FIG. 1F, machine learning model 370-x can be a k-nearest neighbor algorithm trained to process input vector 360-x, which in this example is the same as input vectors 360-0 and 360-1, and generate output data 272-x indicative of whether a subject associated with input vector 360-x is likely to respond or not respond to a treatment also associated with input vector 360-x.

[0136] Alternatively, each of the machine learning models 370-0, 370-1, 370-x can be the same type of machine learning model, or different types of machine learning models, each configured to receive different inputs. For example, the input to the first machine learning model 370-0 can include a vector 360-0 including data representing a first subset or panel of genes of the subject, and then predict whether the subject is likely to respond or not respond to the treatment based on the machine learning model 370-0 processing of the vector 360-0. Additionally, in this example, the input to the second machine learning model 370-1 can include a vector 360-1 including data representing a second subset or panel of genes of the subject, different from the first subset or panel of genes. The second machine learning model can then generate second output data 272-1 indicative of whether the subject associated with the input vector 360-1 is likely to respond or not respond to the treatment associated with the input vector 360-2. This input vector analysis method can continue with each of the x inputs, x input vectors, and x machine learning models. The input to the xth machine learning model 370-x can include a vector 360-x that includes data representing an xth subset or xth panel of genes of the subject that is distinct from (i) at least one, (i) two or more, or (iii) each of the other x-1 input data vectors 370-0 through 370-x-1. In some embodiments, at least one of the x input data vectors can include data representing a complete set of genes from the subject. The xth machine learning model 370-x can then generate second output data 272-x, which indicates whether the subject associated with the input vector 360-x is likely to respond or not respond to the treatment associated with the input vector 360-x.

[0137] The above embodiments of the system 400 are not intended to be limiting, but instead are merely examples of configurations of machine learning models 370-0, 370-1, 370-x and their respective inputs that can be used when using the present disclosure. When referring to these examples, the subject can be any human, non-human animal, plant or other subject. As described above, the input feature vectors can be generated based on the input data and represent the input data. Thus, each input vector can represent data including one or more biomarkers, a disease or disorder and a treatment, a level of effectiveness of the treatment in treating the disease or disorder of the subject having the biomarker. A "treatment" can include data describing any therapeutic substance, such as a small molecule drug or biologic, details of the treatment (e.g., dosage, regimen, missed doses, etc.), or any combination thereof.

[0138] In the embodiment of FIG. 1F, the output data 272-0, 272-1, 272-x can be analyzed using a voting unit 480. For example, the output data 272-0, 272-1, 272-x can be input to the voting unit 480. In some embodiments, the output data 272-0, 272-1, 272-x can be data indicative of whether a subject associated with an input vector processed by the machine learning model is likely to respond or not respond to a treatment associated with the vector processed by the machine learning model. The data indicative of the subject associated with the input vector and generated by each machine learning model can include a "0" or a "1". A "0" generated by the machine learning model 370-0 based on the machine learning model's 370-0 processing of the input vector 360-0 can indicate that a subject associated with the input vector 360-0 is likely to not respond to a treatment associated with the input vector 360-0. Similarly, a "1" generated by the machine learning model 360-0 based on the machine learning model's 370-0 processing of the input vector 360-0 may indicate that the subject associated with the input vector 360-0 is likely to respond to the treatment associated with the input vector 360-0. Although this example uses a "0" as "non-response" and a "1" as "response," the disclosure is not so limited. Instead, any value may be generated as the output data to represent the "response" and "non-response" classes. For example, in some embodiments, a "1" may be used to represent the "non-response" class, and a "0" may be used to represent the "response" class. In still other embodiments, the output data 272-0, 272-1, 272-x may include a probability indicating the likelihood that the subject associated with the input vector processed by the machine learning model will be associated with the "response" or "non-response" class. In such an embodiment, for example, the generated probability may be applied to a threshold, and if the threshold is met, the subject associated with the input vector processed by the machine learning model may be determined to be in the "response" class.

[0139] The voting unit 480 can evaluate the received output data 270-0, 272-1, 272-x and determine whether the subject associated with the processed input vector 360-0, 360-1, 360-x is likely to respond or not respond to the treatment associated with the processed input vector 360-0, 360-1, 360-x. The voting unit 480 can then determine whether the subject associated with the input vector 360-0, 360-1, 360-x is likely to respond to the treatment associated with the input vector 360-0, 360-2, 360-x based on the set of received output data 270-0, 272-1, 272-x. In some embodiments, the voting unit 480 can apply a "majority rule". Applying the majority rule, the voting unit 480 can tally up the outputs 272-0, 272-1, and 272-x that indicate that the subject will respond and the outputs 272-0, 272-1, 272-x that indicate that the subject will not respond. The class with the majority of predictions or votes (e.g., a responsive or non-responsive class) is then selected as the appropriate classification for the subject associated with the input vector 360-0, 360-1, 360-x. This selected class can be referred to as the actual entity class, and each of the predictions or votes output by the machine learning models 370-0, 370-1, 370-x can be referred to as the initial entity class.

[0140] Thus, in some embodiments, determining the majority of predictions or votes may be accomplished by the voting unit 480 tallying the number of occurrences of predictions or votes for each initial entity class. For example, the system 600 may determine the number of times each initial entity class is predicted or voted for by the machine learning models 370-0, 370-1, 370-x, and then select the entity class associated with the greatest number of occurrences of predictions or votes.

[0141] In some embodiments, the voting unit 480 can complete a more nuanced analysis. For example, in some embodiments, the voting unit 480 can store a confidence score for each machine learning model 370-0, 370-1, 370-x. The confidence score for each machine learning model 370-0, 370-1, 370-x can initially be set to a default value, such as 0, 1, etc. Thereafter, for each round of processing of input vectors, the voting unit 480 or other modules of the application server 240 can adjust the confidence score of the machine learning model 370-0, 370-1, 370-x based on whether the machine learning model correctly predicted the target classification selected by the voting unit 480 during the immediately preceding iteration. Thus, the stored confidence score for each machine learning model can provide an indication of the historical accuracy of each machine learning model.

[0142] In a more subtle approach, the voting unit 480 may adjust the output data 272-0, 272-0, 272-x generated by each machine learning model 370-0, 370-1, 370-x, respectively, based on the confidence score calculated for the machine learning model. Thus, a confidence score indicating that the machine learning model is historically accurate may be used to boost the value of the output data generated by the machine learning model. Similarly, a confidence score indicating that the machine learning model is historically inaccurate may be used to decrease the value of the output data generated by the machine learning model. Such boosting or decreasing the value of the output data generated by the machine learning model may be achieved, for example, by using the confidence score as a multiplier, less than 1 for a decrease and greater than 1 for a boost. Other operations may also be used to adjust the value of the output data, for example, by subtracting the confidence score from the value of the output data to decrease the value of the output data, or by adding the confidence score to the value of the output data to boost the value of the output data. The use of confidence scores to boost or reduce the value of output data generated by a machine learning model is particularly useful when the machine learning model is configured to output a probability that applies to one or more thresholds for determining whether a subject will respond or not respond to a treatment, because the confidence scores to adjust the output of the machine learning model can be used to move the generated output value above or below a class threshold, thereby altering predictions made by the machine learning model based on its historical accuracy.

[0143] The use of the voting unit 480 to evaluate the output of multiple machine learning models can result in greater accuracy in predicting the efficacy of a treatment for a particular set of biomarkers of interest, since a consensus between multiple machine learning models can be evaluated instead of the output of only a single machine learning model.

[0144] FIG. 1G is a block diagram of system components that can be used to implement the systems of FIGS.

[0145] Computing device 600 is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. Computing device 650 is intended to represent various forms of mobile devices, such as personal digital assistants, mobile phones, smart phones, and other similar computing devices. In addition, computing device 600 or 650 may include a Universal Serial Bus (USB) flash drive. A USB flash drive may store an operating system and other applications. A USB flash drive may include input / output components, such as a wireless transmitter or a USB connector, that may be inserted into a USB port of another computing device. The components shown, their connections and relationships, and their functions are exemplary only and are not intended to limit the embodiments of the invention described and / or claimed herein.

[0146] The computing device 600 includes a processor 602, a memory 604, a storage device 608, a high-speed interface 608 connecting to the memory 604 and a high-speed expansion port 610, and a low-speed interface 612 connecting to a low-speed bus 614 and the storage device 608. Each of the components 602, 604, 608, 608, 610, and 612 can be interconnected using various buses, implemented on a common motherboard, or attached in any other suitable manner. The processor 602 can process instructions for execution within the computing device 600, including instructions stored in the memory 604 or the storage device 608 for displaying graphical information for a GUI on an external input / output device, such as a display 616 coupled to the high-speed interface 608. In other embodiments, multiple processors and / or multiple buses can be used, along with multiple memories and types of memories, as appropriate. Multiple computing devices 600, each providing a portion of the required operations, can also be connected, for example, as a server bank, a group of blade servers, or a multiprocessor system.

[0147] The memory 604 stores information within the computing device 600. In one embodiment, the memory 604 is one or more volatile memory units. In another embodiment, the memory 604 is one or more non-volatile memory units. The memory 604 can also be another form of computer readable medium, such as a magnetic or optical disk.

[0148] The storage device 608 can provide mass storage for the computing device 600. In one embodiment, the storage device 608 can be or include a computer-readable medium, such as a floppy disk device, a hard disk device, an optical disk device or a tape device, a flash memory or other similar solid-state memory device or an array of devices including the device in a storage area network or other configuration. A computer program product can be tangibly embodied in an information carrier. The computer program product can also include instructions that, when executed, perform one or more methods as described above. The information carrier is a computer-readable or machine-readable medium, such as the memory 604, the storage device 608 or the on-processor memory 602.

[0149] The high-speed controller 608 manages bandwidth-intensive operations for the computing device 600, and the low-speed controller 612 manages low-bandwidth intensive operations. Such an allocation of functions is merely exemplary. In one embodiment, the high-speed controller 608 is coupled to the memory 604, the display 616 (e.g., via a graphics processor or accelerator), and a high-speed expansion port 610 that can accept various expansion cards (not shown). In an embodiment, the low-speed controller 612 is coupled to the storage device 608 and the low-speed expansion port 614. The low-speed expansion port, which can include various communication ports, e.g., USB, Bluetooth, Ethernet, wireless Ethernet, can be coupled, e.g., via a network adapter, to one or more input / output devices, e.g., a keyboard, a pointing device, a microphone / speaker pair, a scanner, or a networking device, e.g., a switch or a router. The computing device 600 can be implemented in several different forms, as shown. For example, it can be implemented as a standard server 620, or multiplexed as a group of such servers. It can also be implemented as part of a rack server system 624. In addition, it may be implemented as a personal computer, such as a laptop computer 622. Alternatively, components from computing device 600 may be combined with other components in a mobile device (not shown), such as device 650. Each such device may include one or more computing devices 600, 650, or the entire system may be made up of multiple computing devices 600, 650 in communication with each other.

[0150] The computing device 600 can be implemented in a number of different forms, as shown in the drawing. For example, it can be implemented as a standard server 620, or multiplexed as a group of such servers. It can also be implemented as part of a rack server system 624. In addition, it can be implemented as a personal computer, such as a laptop computer 622. Alternatively, components from the computing device 600 can be combined with other components in a mobile device (not shown), such as device 650. Each such device can include one or more computing devices 600, 650, or the entire system can be made up of multiple computing devices 600, 650 in communication with each other.

[0151] Computing device 650 includes, among other things, a processor 652, a memory 664, and input / output devices, such as a display 654, a communication interface 666, and a transceiver 668. Device 650 may also include a storage device, such as a microdrive or other device, to provide additional storage. Each of the components 650, 652, 664, 654, 666, and 668 are interconnected using various buses, and some of the components may be mounted on a common motherboard or in other suitable manners.

[0152] The processor 652 can execute instructions within the computing device 650, including instructions stored in the memory 664. The processor can be implemented as a chipset of chips including separate and multiple analog and digital processors. In addition, the processor can be implemented using any of a number of architectures. For example, the processor 610 can be a Complex Instruction Set Computers (CISC) processor, a Reduced Instruction Set Computer (RISC) processor, or a Minimal Instruction Set Computer (MISC) processor. The processor can provide, for example, coordination of other components of the device 650, such as control of a user interface, applications executed by the device 650, and wireless communication by the device 650.

[0153] The processor 652 can communicate with a user via a control interface 658 and a display interface 656 coupled to a display 654. The display 654 can be, for example, a Thin-Film-Transistor Liquid Crystal Display (TFT) display or an Organic Light Emitting Diode (OLED) display or other suitable display technology. The display interface 656 can include suitable circuitry for driving the display 654 to present graphical and other information to the user. The control interface 658 can receive commands from a user and convert them for submission to the processor 652. Additionally, an external interface 662 can be provided in communication with the processor 652 to enable near-field communication between the device 650 and other devices. The external interface 662 can provide, for example, wired communication in some embodiments and wireless communication in other embodiments, and multiple interfaces can be used.

[0154] The memory 664 stores information within the computing device 650. The memory 664 can be realized as one or more of a computer readable medium, a volatile memory unit, or a non-volatile memory unit. An expansion memory 674 can also be provided and connected to the device 650 via an expansion interface 672, which can include, for example, a Single In Line Memory Module (SIMM) card interface. Such expansion memory 674 can provide extra storage space for the device 650 or can store applications or other information for the device 650. In particular, the expansion memory 674 can include instructions for performing or supplementing the above processes or can include security information. Thus, for example, the expansion memory 674 can be provided as a security module for the device 650 and can be programmed with instructions that allow secure use of the device 650. In addition, secure applications can also be provided via a SIMM card, along with further information, for example by placing identification information on the SIMM card in an unhackable manner.

[0155] The memory may include, for example, flash memory and / or NVRAM memory, as described in more detail below. In one embodiment, a computer program product is tangibly embodied in an information carrier. The computer program product includes instructions that, when executed, perform one or more methods, such as those described above. The information carrier is a computer-readable or machine-readable medium, such as memory 664, expansion memory 674, or on-processor memory 652, which may be received, for example, via transceiver 668 or external interface 662.

[0156] The device 650 can communicate wirelessly via a communication interface 666, which can include digital signal processing circuitry if necessary. The communication interface 666 can provide communications under various modes or protocols, such as GSM voice calls, SMS, EMS or MMS messaging, CDMA, TDMA, PDC, WCDMA, CDMA2000 or GPRS, among others. Such communications can be implemented, for example, via a radio frequency transceiver 668. In addition, short-range communications can be implemented, for example, using Bluetooth, Wi-Fi or other such transceivers (not shown). In addition, a Global Positioning System (GPS) receiver module 670 can provide further navigation-related and location-related wireless data to the device 650, which can be used appropriately by applications running on the device 650.

[0157] Device 650 can also communicate audibly using audio codec 660, which can receive voice information from a user and convert it into usable digital information. Audio codec 660 can also generate audible sounds for the user, such as through a speaker in a handset of device 650. Such sounds can include sounds from a telephone call, can include recordings, such as voice messages, music files, etc., and can include sounds generated by applications running on device 650.

[0158] The computing device 650 may be implemented in a number of different forms, as shown, for example as a mobile phone 680, or as part of a smartphone 682, personal digital assistant, or other similar mobile device.

[0159] Various embodiments of the systems and methods described herein can be realized as digital electronic circuitry, integrated circuits, specially designed ASICs (application specific integrated circuits), computer hardware, firmware, software, and / or combinations of such embodiments. These various embodiments can include implementation as one or more computer programs executable and / or interpretable on a programmable system including at least one programmable processor, which can be special purpose or general purpose, coupled to receive data and instructions from a storage system, at least one input device, and at least one output device, and to send data and instructions to the storage system, the at least one input device, and the at least one output device.

[0160] These computer programs (also known as programs, software, software applications or codes) contain machine instructions for a programmable processor and can be implemented in high-level procedural and / or object-oriented programming languages ​​and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus and / or device used to provide machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal, such as a magnetic disk, optical disk, memory, programmable logic device (PLD). The term "machine-readable signal" refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0161] To provide for user interaction, the systems and techniques described herein can be implemented on a computer having a display device, such as a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, for displaying information to the user, as well as a keyboard and a pointing device, such as a mouse or trackball, through which the user can provide input to the computer. Other types of devices can also be used to provide for user interaction. For example, the feedback provided to the user can be any form of sensory feedback, such as visual feedback, auditory feedback or tactile feedback; input from the user can be received in any form, including acoustic, speech or tactile input.

[0162] The systems and techniques described herein can be implemented as a computing system that includes a back-end component, e.g., a data server, or includes a middleware component, e.g., an application server, or includes a front-end component, e.g., a client computer having a graphical user interface or web browser through which a user can interact with embodiments of the systems and techniques described herein, or includes any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication, e.g., a communications network. Examples of communications networks include a local area network ("LAN"), a wide area network ("WAN"), and the Internet.

[0163] A computing system may include clients and servers. Clients and servers are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

[0164] Computer Systems The implementation of the method may also use computer-related software and systems. A computer software product as described herein typically includes a computer-readable medium having computer-executable instructions for performing the logical steps of the method as described herein. Suitable computer-readable media include floppy disks, CD-ROM / DVD / DVD-ROM, hard disk drives, flash memory, ROM / RAM, magnetic tape, etc. The computer-executable instructions may be written in a suitable computer language or a combination of several languages. Basic computational biology methods are described, for example, in Setubal and Meidanis at al., Introduction to Computational Biology Methods (PWS Publishing Company, Boston, 1997); Salzberg, Searles, Kasif, (Ed.), Computational Methods in Molecular Biology (Elsevier, Amsterdam, 1998); Rashidi and Buehler, Bioinformatics Basics: Application in Biological Science and Medicine (CRC Press, London, 2000) and Ouelette and Bzevanis Bio informatics: A Practical Guide for Analysis of Genes and Proteins (Wiley & Sons, Inc., 2.sup.nd ed., 2001). See U.S. Patent No. 6,420,108.

[0165] The method may also utilize various computer program products and software for a variety of purposes, such as probe design, data management, analysis, and instrument operation. See U.S. Patent Nos. 5,593,839, 5,795,716, 5,733,729, 5,974,164, 6,066,454, 6,090,555, 6,185,561, 6,188,783, 6,223,127, 6,229,911, and 6,308,170.

[0166] In addition, the method relates to an embodiment that includes a method of providing genetic information via a network such as the Internet, as shown in US Patent Application Nos. 10 / 197,621, 10 / 063,559 (US Patent Application Publication No. 20020183936), 10 / 065,856, 10 / 065,868, 10 / 328,818, 10 / 328,872, 10 / 423,403 and 60 / 482,389. For example, one or more molecular profiling techniques can be performed in one location, such as a city, state, country or continent, and the results can be transmitted to a different city, state, country or continent. Treatment selection can then be made in whole or in part at the second location. The method as described herein includes a process of transferring information between different locations.

[0167] Conventional data networking, application development and other functional aspects of the system (as well as components of the individual operating components of the system) may not be described in detail herein, but are part of as described herein. Additionally, the connecting lines shown in the various figures contained herein are intended to represent example functional relationships and / or physical couplings between the various elements. It should be noted that many alternative or additional functional relationships or physical connections may be present in an actual system.

[0168] The various system components detailed herein may include one or more of the following: a host server or other computing system including a processor for processing digital data; a memory coupled to the processor for storing digital data; an input digitizer coupled to the processor for inputting digital data; an application program stored in the memory and accessible by the processor for directing the processing of the digital data by the processor; a display device coupled to the processor and the memory for displaying information derived from the digital data processed by the processor; and a number of databases. The various databases used herein may include patient data, such as family history, age group and environmental data, biosample data, previous treatment and protocol data, patient clinical data, molecular profiling data of biosamples, data regarding therapeutic and / or investigational drugs, gene libraries, disease libraries, drug libraries, patient follow-up data, file management data, financial management data, billing data and / or the like useful for the operation of the system. As one skilled in the art will appreciate, the user computer may include an operating system (e.g., Windows NT, 95 / 98 / 2000, OS2, UNIX, Linux, Solaris, MacOS, etc.), as well as various conventional support software and drivers that typically accompany a computer. The computer may include any suitable personal computer, network computer, workstation, minicomputer, mainframe, etc. The user computer may be in a home or medical / business environment with network access. In an exemplary embodiment, the access is via a network or via the Internet via a commercially available web browser software package.

[0169] The term "network" as used herein is intended to include any electronic communication means incorporating both hardware and software components. Communication between the parties may be accomplished through any suitable communication channel, such as a telephone network, an extranet, an intranet, the Internet, a point of interaction device, a personal digital assistant (e.g., Palm Pilot, Blackberry, mobile phone, kiosk, etc.), online communication, satellite communication, offline communication, wireless communication, transponder communication, a local area network (LAN), a wide area network (WAN), a networked or linked device, a keyboard, a mouse, and / or any suitable communication or data entry modality. Moreover, although the system is often described herein as being implemented using TCP / IP communication protocols, the system may be implemented using IPX, Appletalk, IP-6, NetBIOS, OSI, or any number of existing or future protocols. If the network has the nature of a public network such as the Internet, it may be advantageous to assume that the network is not secure and subject to eavesdropping. Specific information relating to protocols, standards, and application software used in connection with the Internet is generally known to those skilled in the art and thus need not be detailed herein. See, for example, Dilip Naik, Internet Standards and Protocols (1998); Java 2 Complete, various authors, (Sybex 1999); Deborah Ray and Eric Ray, Mastering HTML 4.0 (1997); and Loshin, TCP / IP Clearly Explained (1997) and David Gourley and Brian Totty, HTTP, The Definitive Guide (2002), the contents of which are incorporated herein by reference.

[0170] The various system components may be suitably coupled independently, individually, or collectively into a network via data links, including, by way of example, standard modem communications, cable modem, Dish network, ISDN, Digital Subscriber Line (DSL), or a connection to an Internet Service Provider (ISP) via a local loop such as those typically used in connection with various wireless communication methods. See, for example, Gilbert Held, Understanding Data Communications (1996), which is incorporated herein by reference. Note that the network may be implemented as other types of networks, for example, an interactive television (ITV) network. Moreover, the system contemplates the use, sale, or distribution of any goods, services, or information over any network having similar functionality as described herein.

[0171] As used herein, "transmitting" may include sending electronic data from one system component to another over a network connection. Additionally, as used herein, "data" may include including information in digital or any other form, such as commands, queries, files, data for storage, etc.

[0172] The system contemplates use in connection with web services, utility computing, pervasive and personalized computing, security and identity solutions, autonomic computing, commodity computing, mobility and wireless solutions, open source, biometric authentication, grid computing and / or mesh computing.

[0173] Any database detailed herein may include a relational, hierarchical, graphical or object-oriented structure and / or any other database organization. Common database products that may be used to implement the database include DB2 from IBM (White Plains, NY), various database products available from Oracle Corporation (Redwood Shores, CA), Microsoft Access or Microsoft SQL Server from Microsoft Corporation (Redmond, Washington), or any other suitable database product. Moreover, the database may be organized in any suitable manner, such as, for example, a data table or a lookup table. Each record may be a single file, a set of files, a set of linked data fields, or any other data structure. Association of specific data may be achieved by any desired data association technique as known or implemented in the art. For example, association may be achieved manually or automatically. Automatic association techniques may include, for example, database search, database merge, GREP, AGREP, SQL, using key fields in tables to speed up searches, sequentially searching all tables and files, sorting records in a file according to a known order to simplify searching, and the like. The association step may be accomplished, for example, by a database merge function using preselected "key fields" of databases or data sectors.

[0174] More specifically, a "key field" divides the database according to the superclass of objects defined by the key field. For example, a particular type of data may be designated as a key field in multiple related data tables, in which case the data tables may be linked based on the type of data in the key field. The data corresponding to the key field in each of the linked data tables is preferably the same or of the same type. However, data tables having similar but not identical data in the key field may also be linked, for example, using AGREP. According to one embodiment, any suitable data storage technology may be used to store the data without a standard format. Data sets may be stored using any suitable technique, including, for example, storing individual files using the ISO / IEC 7816-4 file structure; realizing a domain where a dedicated file exposing one or more base files containing one or more datasets is selected; using datasets stored in individual files using a hierarchical filing system; using datasets stored as records in a single file (compressed, SQL accessible, hashed key(s), numeric, alphabetic with first tuple, etc.); BLOB (Binary Large Object); stored as ungrouped data elements coded using ISO / IEC 7816-6 data elements; stored as ungrouped data elements coded using ISO / IEC Abstract Syntax Notation (ASN.1) as in ISO / IEC 8824 and 8825; and / or using other proprietary techniques which may include fractal compression schemes, image compression methods, etc.

[0175] In one exemplary embodiment, the ability to store a wide variety of information in different formats is facilitated by storing the information as a BLOB. Thus, any binary information can be stored in the storage space associated with a dataset. The BLOB method may store the dataset as ungrouped data elements formatted as blocks of binary over fixed memory offsets using either fixed storage allocation, circular queue techniques, or best practices for memory management (e.g., paged memory, least recently used, etc.). By using the BLOB method, the ability to store various datasets having different formats facilitates storage of data by multiple unrelated owners of the datasets. For example, a first dataset that may be stored may be provided by a first party, a second dataset that may be stored may be provided by an unrelated second party, and yet a third dataset that may be stored may be provided by a third party that is unrelated to the first and second parties. Each of these three exemplary datasets may include different information that is stored using different data storage formats and / or techniques. Additionally, each dataset may include a subset of data that may also differ from the other subsets.

[0176] As noted above, in various aspects, data may be stored without regard to a common format. However, in one exemplary aspect, data sets (e.g., BLOBs) may be annotated in a standard manner when provided for manipulating the data. The annotations may include short headers, trailers, or other suitable indicators associated with each data set that are configured to convey information useful in managing the various data sets. For example, the annotations may be referred to herein as "conditional headers," "headers," "trailers," or "status," and may include an indication of the status of the data set, or may include an identifier correlated to a particular publisher or owner of the data. Subsequent bytes of data may be used to indicate, for example, the identity of the publisher or owner of the data, a user, a transaction / membership account identifier, or the like. Each of these conditional annotations is described in further detail herein.

[0177] Dataset annotations may also be used for other types of status information and various other purposes. For example, the dataset annotations may include security information that establishes access levels. The access levels may be configured, for example, such that only certain individuals, levels of employees, companies or other entities are allowed to access the dataset, or may be configured to allow access to certain datasets based on transactions, data issuers or owners, users, etc. Furthermore, the security information may restrict / allow only certain actions, such as accessing the dataset, modifying it, and / or deleting it. In one example, the dataset annotation indicates that only the dataset owner or user is allowed to delete the dataset, various identified users may be allowed to access the dataset for reading, and all other users are excluded from accessing the dataset. However, other access restriction parameters may be used that allow various entities to access the dataset with various permission levels as appropriate. The data, including the header or trailer, may be received by a standalone interactive device configured to add, delete, modify, or augment the data according to the header or trailer.

[0178] Those skilled in the art will also understand that for security reasons, any database, system, device, server or other component of a system may consist of any combination thereof, in one or multiple locations, and each database or system may include any of a variety of appropriate security mechanisms, such as firewalls, access codes, encryption, decryption, compression, decompression, etc.

[0179] The web client's computing unit may further include an Internet browser connected to the Internet or an intranet using standard dial-up, cable, DSL, or any other Internet protocol known in the art. Transactions occurring at the web client may pass through a firewall to prevent unauthorized access from users of other networks. Additionally, additional firewalls may be deployed between various components of the CMS to further enhance security.

[0180] The firewall may include any hardware and / or software suitably configured to protect the CMS components and / or enterprise computing resources from users of other networks. Additionally, the firewall may be configured to limit or restrict access to various systems and components behind the firewall in the case of web clients connecting through a web server. Firewalls may exist in a variety of configurations including stateful inspection, proxy-based firewalls and packet filtering, among others. The firewall may be integrated into the web server or any other CMS component, or may even exist as a separate entity.

[0181] The computers detailed herein may provide suitable websites or other Internet-based graphical user interfaces that are accessible by users. In one embodiment, Microsoft Internet Information Server (IIS), Microsoft Transaction Server (MTS) and Microsoft SQL Server are used in conjunction with the Microsoft operating system, Microsoft NT web server software, Microsoft SQL Server database system and Microsoft Commerce Server. In addition, components such as Access or Microsoft SQL Server, Oracle, Sybase, Informix MySQL, Interbase, etc. may be used to provide an Active Data Object (ADO) compliant database management system.

[0182] Any of the communication, input, storage, database or display detailed herein may be facilitated through a website having a web page. The term "web page" as used herein is not meant to limit the types of documents and applications that may be used to interact with a user. For example, a typical website may include various forms, Java applets, JavaScript, Active Server Pages (ASP), Common Gateway Interface (CGI) scripts, Extensible Markup Language (XML), Dynamic HTML, Cascading Style Sheets (CSS), helper applications and plug-ins, etc., in addition to standard HTML documents. The server may include a web service that receives a request from a web server, including a URL (http: / / yahoo.com / stockquotes / ge) and an IP address (123.56.789.234). The web server retrieves the appropriate web page and sends the data or application for the web page to the IP address. A web service is an application that can interact with other applications via a communication means such as the Internet. Web services are typically based on standards or protocols such as XML, XSLT, SOAP, WSDL and UDDI. Web services methods are well known in the art and are covered in many standard textbooks, see, for example, Alex Nghiem, IT Web Services: A Roadmap for the Enterprise (2003), incorporated herein by reference.

[0183] The web-based clinical database for the subject systems and methods preferably has the ability to upload and store clinical data files in native format and is searchable by any clinical parameter. The database is also extensible and can populate with clinical annotations from any study using the EAV data model (metadata) for easy integration with other studies. In addition, the web-based clinical database is flexible and can be XML and XSLT enabled to allow users to dynamically add customized queries. Additionally, the database includes export capabilities to CDISC ODM.

[0184] Implementers will also appreciate that there are many ways to display data within a browser-based document: data may be displayed as standard text, in fixed lists, scrollable lists, drop-down lists, editable text fields, fixed text fields, pop-up windows, etc. Similarly, there are many ways available to change data within a web page, e.g., free text entry using the keyboard, selecting menu items, check boxes, option boxes, etc.

[0185] The system and method may be described herein in terms of functional block components, screenshots, optional selections, and various processing steps. It should be understood that such functional blocks may be implemented by any number of hardware and / or software components configured to perform the specified functions. For example, the system may employ various integrated circuit components, such as memory elements, processing elements, logic elements, look-up tables, etc., that may perform various functions under the control of one or more microprocessors or other control devices. Similarly, the software elements of the system may be implemented in any programming or scripting language, such as C, C++, Macromedia Cold Fusion, Microsoft Active Server Pages, Java, COBOL, Assembler, PERL, Visual Basic, SQL Stored Procedures, Extensible Markup Language (XML), and the various algorithms are implemented in any combination of data structures, objects, processes, routines, or other programming elements. Additionally, it should be noted that the system may employ any number of conventional techniques for data transmission, signaling, data processing, network control, and the like. Still further, the system may also be used to detect or prevent security issues in client-side scripting languages, such as JavaScript, VBScript, and the like.For an introduction to the basics of cryptography and network security, see any of the following references, all of which are incorporated herein by reference: (1) "Applied Cryptography: Protocols, Algorithms, And Source Code In C," by Bruce Schneier, published by John Wiley & Sons (second edition, 1995); (2) "Java Cryptography" by Jonathan Knudson, published by O'Reilly & Associates (1998); (3) "Cryptography & Network Security: Principles & Practice" by William Stallings, published by Prentice Hall.

[0186] As used herein, the terms "end user", "consumer", "customer", "client", "treating physician", "hospital" or "business" may be used interchangeably and shall mean any person, entity, machine, hardware, software or business, respectively. Each participant is equipped with a computing device to interact with the system and facilitate online data access and data entry. The customer has a computing unit in the form of a personal computer, although other types of computing units may be used including laptops, notebooks, handheld computers, set-top boxes, mobile phones, touch-tone phones, etc. The system and the owner / operator of the method have a computing unit realized in the form of a computer server, although other embodiments are contemplated by systems including computing centers depicted as mainframe computers, minicomputers, PC servers, networks of computers in the same or different geographic locations, etc. Moreover, the system contemplates the use, sale or distribution of any goods, services or information over any network having similar functionality as described herein.

[0187] In one exemplary embodiment, each client customer may be issued an "account" or "account number." Account or account number as used herein may include any device, code, number, letter, symbol, digital certificate, smart chip, digital signal, analog signal, biometric or other identifier / indicia (e.g., one or more of an authentication / access code, personal identification number (PIN), Internet code, other identification code, etc.) suitably configured to allow a consumer to access, interact with, or communicate with the system. The account number may optionally be located on or associated with a charge card, credit card, debit card, prepaid card, embossed card, smart card, magnetic stripe card, bar code card, transponder, radio frequency card, or associated account. The system may include or interface with any of the aforementioned cards or devices, or a fob having a transponder and an RFID reader in RF communication with the fob. While the system may include a fob embodiment, the method is not so limited. In fact, the system may include any device having a transponder configured to communicate with an RFID reader via RF communication. Exemplary devices may be, for example, key rings, tags, cards, mobile phones, watches, or any such form that can be presented for interrogation. Moreover, the systems, computing units, or devices detailed herein may include "pervasive computing devices," which may include traditional non-computerized devices that have embedded computing units. Account numbers may be distributed and stored in any form of plastic, electronic, magnetic, radio frequency, wireless, audio, and / or optical device that can transmit or download data from itself to a second device.

[0188] As will be appreciated by those skilled in the art, the system may be embodied as a customization of an existing system, an add-on product, upgraded software, a standalone system, a distributed system, a method, a data processing system, a device for data processing, and / or a computer program product. Thus, the system may take the form of an all-software embodiment, an all-hardware embodiment, or an embodiment combining both software and hardware aspects. Furthermore, the system may take the form of a computer program product on a computer-readable storage medium having computer-readable program code means embodied in the storage medium. Any suitable computer-readable storage medium may be used, including hard disks, CD-ROMs, optical storage devices, magnetic storage devices, and the like.

[0189] Systems and methods are described herein with reference to screenshots, block diagrams, and flowchart illustrations of methods, apparatus (e.g., systems), and computer program products according to various aspects. It will be understood that each functional block of the block diagrams and flowchart illustrations, and combinations of functional blocks in the block diagrams and flowchart illustrations, respectively, can be implemented by computer program instructions.

[0190] These computer program instructions may be loaded into a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, where the instructions executing on the computer or other programmable data processing apparatus create means for implementing the functions specified in one or more of the flowchart blocks. These computer program instructions may also be stored in a computer readable memory that can instruct the computer or other programmable data processing apparatus to function in a particular manner, where the instructions stored in the computer readable memory produce an article of manufacture including instruction means for implementing the functions specified in one or more of the flowchart blocks. The computer program instructions may also be loaded into a computer or other programmable data processing apparatus to produce a computer implemented process, where a series of operating steps executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more of the flowchart blocks.

[0191] Thus, the functional blocks of the block diagrams and flowchart diagrams support combinations of means for performing the specified functions, combinations of steps for performing the specified functions, and program instruction means for performing the specified functions. It will also be understood that each functional block of the block diagrams and flowchart diagrams, and combinations of functional blocks of the block diagrams and flowchart diagrams, can be implemented by a dedicated hardware-based computer system that performs the specified functions or steps, or by an appropriate combination of dedicated hardware and computer instructions. Furthermore, the illustrations and descriptions of the process flows may refer to user windows, web pages, websites, web forms, prompts, and the like. The practitioner will understand that the illustrated steps described herein may include any number of configurations, including the use of windows, web pages, web forms, pop-up windows, prompts, and the like. Furthermore, it should be understood that the multiple steps illustrated and described may be combined into a single web page and / or window, but have been expanded for clarity. In other cases, steps illustrated and described as a single process step may be separated into multiple web pages and / or windows, but have been combined for clarity.

[0192] Molecular profiling Molecular profiling approaches provide a method for selecting candidate treatments for individuals that can improve the clinical course for individuals with a disease or disorder, such as cancer. Molecular profiling approaches provide clinical benefits for individuals, such as identifying treatment regimens that provide longer progression-free survival (PFS), longer disease-free survival (DFS), longer overall survival (OS) or longer life span. The methods and systems as described herein relate to molecular profiling of cancer on an individual basis, which can identify optimal treatment regimens. Molecular profiling provides a personalized approach to selecting candidate treatments that are likely to benefit the cancer. The molecular profiling methods described herein can be used to guide treatment in any desired setting, including first-line / standard-of-care settings, or in the case of patients with poor prognosis, such as patients with metastatic disease, or patients whose cancer has progressed on standard first-line treatment, or patients whose cancer has progressed on previous chemotherapy or hormonal therapy.

[0193] Using the systems and methods of the invention, patients may be classified as more or less likely to benefit from or respond to various treatments. Unless otherwise specified, the term "response" or "non-response" as used herein refers to any suitable indication that a treatment provides a benefit or increased benefit to the patient ("responder" or "beneficiary"), or has a reduced or lack of benefit to the patient ("non-responder" or "non-beneficiary" or "reduced benefit"). A predicted responder or beneficiary may have an increased probability of benefit or be more likely to benefit from a treatment, and a predicted non-responder or non-beneficiary may have a reduced probability of benefit or be less likely to benefit. Indications of benefit or lack thereof may be determined using any suitable criteria, for example, accepted clinical response criteria such as standard RECIST (Response Evaluation Criteria in Solid Tumors) criteria or other useful patient response criteria, such as progression-free survival (PFS), time to progression (TTP), disease-free survival (DFS), time to next treatment (TNT, TTNT), tumor shrinkage or disappearance, etc. RECIST is a set of rules published by an international consortium that defines whether tumors improve ("respond"), remain stable ("stable"), or worsen ("progress") during treatment in cancer patients. As used herein, unless otherwise specified, a patient "benefit" from a treatment may refer to any suitable measure of improvement, including a RECIST response or longer PFS / TTP / DFS / TNT / TTNT, and a "lack of benefit" from a treatment may refer to any suitable measure of disease worsening during treatment. Generally, disease stabilization is considered a benefit, although in certain circumstances, if so specified herein, stabilization may be considered a lack of benefit. If there is no prediction of an acceptable level of benefit or lack of benefit, the predicted or indicated benefit may be described as "undetermined."In some cases, if the benefit cannot be calculated, for example due to lack of necessary data, the benefit is considered to be indeterminate.

[0194] Personalized medicine based on pharmacogenetic insights such as those provided by molecular profiling as described herein is increasingly taken for granted by some practitioners and the general press, but it forms the basis of hope for improved cancer treatment. However, molecular profiling as taught herein represents a radical departure from traditional approaches to tumor treatment, where patients are, for the most part, grouped and treated with approaches based on findings from light microscopy and disease stage. Traditionally, differential responses to specific therapeutic strategies have been determined only after treatment has been administered, i.e., a posteriori. The "standard" approach to disease treatment relies on what is generally true for a given cancer diagnosis, and treatment responses are vetted by randomized Phase III clinical trials to form the "standard of care" in medical practice. The results of these trials are compiled in consensus statements by guideline organizations such as the National Comprehensive Cancer Network and the American Society of Clinical Oncology. The NCCN Compendium™ contains authoritative, scientifically derived information designed to aid decision-making regarding the appropriate use of drugs and biologics in cancer patients. The NCCN Compendium™ is recognized by the Centers for Medicare & Medicaid Services (CMS) and United Healthcare as the authoritative reference source for oncology insurance coverage. On Compendium treatments are those recommended by such guides. Biostatistical methods used to validate the results of clinical trials rely on minimizing differences between patients and are based on declaring the probability of error that one method is superior to another for groups of patients determined solely by light microscopy and stage (not by individual differences in tumors). The molecular profiling methods described herein exploit such individual differences. The methods can provide candidate treatments, which can then be selected by a physician to treat the patient.

[0195] Molecular profiling can be used to provide a comprehensive view of the biological state of a sample. In some embodiments, molecular profiling is used for whole-tumor profiling. Thus, several molecular techniques are used to assess the state of a tumor. Whole-tumor profiling can be used to select candidate treatments for a tumor. Molecular profiling can be used to select candidate therapeutics for any sample at any stage. In some embodiments, the methods as described herein are used to profile a newly diagnosed cancer. Candidate treatments indicated by molecular profiling can be used to select a therapy for treating the newly diagnosed cancer. In other embodiments, the methods as described herein are used to profile a cancer that has already been treated, for example, by one or more standard therapies. In some embodiments, the cancer is refractory to a previous treatment. For example, the cancer can be refractory to a standard treatment for the cancer. The cancer can be a metastatic cancer or other recurrent cancer. The treatment can be an on-compendium treatment or an off-compendium treatment.

[0196] Molecular profiling can be carried out by any known means for detecting molecules in biological samples.Molecular profiling includes methods such as nucleic acid sequencing, for example DNA sequencing or RNA sequencing; immunohistochemistry (IHC); in situ hybridization (ISH); fluorescent in situ hybridization (FISH); chromogenic in situ hybridization (CISH); PCR amplification (for example qPCR or RT-PCR); various types of microarrays (mRNA expression arrays, low density arrays, protein arrays, etc.); various types of sequencing (Sanger, pyrosequencing, etc.); comparative genomic hybridization (CGH); high-throughput or next-generation sequencing (NGS); Northern blots; Southern blots; immunoassays; and any other suitable techniques for assaying the presence or amount of biomolecules of interest.In various embodiments, any one or more of these methods can be used in parallel or sequential with each other to evaluate the target genes disclosed herein.

[0197] Molecular profiling of individual samples is used to select one or more candidate treatments for the disorder in the subject, for example, by identifying targets for drugs that may be effective for a given cancer. For example, the candidate treatments can be treatments known to affect cells that differentially express the genes identified by molecular profiling techniques, experimental drugs, government or regulatory approved drugs, or any combination of such drugs (which may have been researched and approved for a specific indication that is the same or different from the indication of the subject for which the biological sample is taken and molecularly profiled).

[0198] When multiple biomarker targets are revealed by evaluating target genes by molecular profiling, one or more decision rules can be applied to prioritize the selection of a particular therapeutic agent for the treatment of an individual on an individualized basis. Rules as described herein can assist in prioritizing treatments, e.g., direct results of molecular profiling, expected efficacy of the therapeutic agent, previous history of the same or other treatments, expected side effects, availability of the therapeutic agent, cost of the therapeutic agent, drug-drug interactions and other factors considered by the treating physician. Based on the recommended and prioritized therapeutic agent targets, the physician can determine the course of treatment for a particular individual. Thus, molecular profiling methods and systems as described herein can select candidate treatments based on the individual characteristics of diseased cells, e.g., tumor cells, in a subject requiring treatment and other individualizing factors, as opposed to relying on the conventional one-size-fits-all approach that is routinely used to treat individuals suffering from diseases, particularly cancer. In some cases, the recommended treatment is one that is not normally used to treat the disease or disorder afflicting the subject. In some cases, the recommended treatment is used after standard treatments no longer provide sufficient efficacy.

[0199] The treating physician can use the results of molecular profiling to optimize the treatment regimen for the patient.The candidate treatment identified by the method as described herein can be used to treat the patient, but such treatment is not required for the method.In fact, the analysis of molecular profiling results and the identification of candidate treatment based on such results can be automated and do not require the involvement of a physician.

[0200] Biological Entities Nucleic acids include deoxyribonucleotides or ribonucleotides and their polymers in either single-stranded or double-stranded form, or their complements. Nucleic acids can contain known nucleotide analogs, synthetic, natural, and non-natural, or modified backbone residues or bonds, which have similar binding properties as the reference nucleic acid, and which are metabolized in a similar manner as the reference nucleotide. Examples of such analogs include, but are not limited to, phosphorothioates, phosphoramidates, methyl phosphonates, chiral-methyl phosphonates, 2-O-methyl ribonucleotides, peptide-nucleic acids (PNAs). Nucleic acid sequences can include the sequences specified, in addition to their conservatively modified variants (e.g., degenerate codon substitutions) and complementary sequences. Specifically, degenerate codon substitutions can be achieved by generating sequences in which the third position of one or more selected (or all) codons is substituted with mixed-base and / or deoxyinosine residues (Batzer et al., Nucleic Acid Res. 19:5081 (1991); Ohtsuka et al., J. Biol. Chem. 260:2605-2608 (1985); Rossolini et al., Mol. Cell Probes 8:91-98 (1994)). The term nucleic acid can be used interchangeably with gene, cDNA, mRNA, oligonucleotide, and polynucleotide.

[0201] A particular nucleic acid sequence may implicitly encompass nucleic acid sequences that code for a particular sequence as well as "splice variants" and truncated forms. Similarly, a particular protein encoded by a nucleic acid may encompass any protein encoded by a splice variant or truncated form of that nucleic acid. A "splice variant," as the name suggests, is a product of alternative splicing of a gene. After transcription, an initial nucleic acid transcript may be spliced ​​such that different (alternative) nucleic acid splice products code for different polypeptides. The mechanisms for producing splice variants vary, but include alternative splicing of exons. Alternative polypeptides obtained by read-through transcription from the same nucleic acid are also encompassed by this definition. Any product of a splicing reaction, including recombinant forms of splice products, is included in this definition. Nucleic acids can be truncated at the 5' or 3' end. Polypeptides can be truncated at the N- or C-terminus. Truncated versions of nucleic acid or polypeptide sequences can be natural or can be generated using recombinant techniques.

[0202] The terms "gene variant" and "nucleotide variant" are used interchangeably herein to refer to changes or alterations to a reference human gene or cDNA sequence at a particular locus, including, but not limited to, deletions, insertions, inversions, and substitutions of nucleotide bases in coding and non-coding regions. Deletions can be of a single nucleotide base, a portion or region of the nucleotide sequence of a gene, or the entire gene sequence. Insertions can be of one or more nucleotide bases. Gene variants or nucleotide variants may occur in transcriptional regulatory regions, untranslated regions of mRNA, exons, introns, exon / intron junctions, etc. Gene variants or nucleotide variants can potentially result in stop codons, frameshifts, deletions of amino acids, altered gene transcript splice forms, or altered amino acid sequences.

[0203] An allele or gene allele generally includes a naturally occurring gene having a reference sequence or a gene containing particular nucleotide variants.

[0204] A haplotype refers to a combination of genetic (nucleotide) variants in a region of genomic DNA on a chromosome or mRNA found in an individual. Thus, a haplotype contains several genetically linked polymorphic variants that are typically inherited together as a unit.

[0205] As used herein, the term "amino acid variant" is used to refer to amino acid changes relative to a reference human protein sequence that result from genetic or nucleotide variants relative to a reference human gene encoding the reference protein. The term "amino acid variant" is intended to encompass not only single amino acid substitutions in the amino acid sequence of the reference protein, but also amino acid deletions, insertions, and other significant changes.

[0206] The term "genotype" as used herein refers to the nature of the nucleotide at a particular nucleotide variant marker (or locus) in either one allele or both alleles of a gene (or a particular chromosomal region). For a particular nucleotide position of a gene of interest, the nucleotide at that locus or its equivalent in one or both alleles forms the genotype of the gene at that locus. The genotype can be homozygous or heterozygous. Thus, "genotyping" means determining the genotype, i.e., the nucleotide at a particular gene locus. Genotyping can also be done by determining the amino acid variant at a particular position of a protein, which can be used to infer the corresponding nucleotide variant.

[0207] The term "locus" refers to a specific position or site in a gene sequence or protein. Thus, there may be one or more consecutive nucleotides at a particular gene locus, or one or more amino acids at a particular locus in a polypeptide. Moreover, a locus may refer to a specific position in a gene where one or more nucleotides have been deleted, inserted, or inverted.

[0208] Unless otherwise specified or understood by those skilled in the art, the terms "polypeptide", "protein" and "peptide" are used interchangeably herein to refer to an amino acid chain in which amino acid residues are linked by covalent peptide bonds. The amino acid chain can be of any length of at least two amino acids, including full-length proteins. Unless otherwise specified, polypeptides, proteins and peptides also encompass various modified forms thereof, including but not limited to glycosylated forms, phosphorylated forms, and the like. A polypeptide, protein or peptide can also be referred to as a gene product.

[0209] A list of genes and gene products that can be assayed by molecular profiling techniques is provided herein. The list of genes may be provided in conjunction with molecular profiling techniques that detect gene products (e.g., mRNA or protein). Those skilled in the art will understand that this refers to the detection of gene products of the listed genes. Similarly, the list of gene products may be provided in conjunction with molecular profiling techniques that detect gene sequence or copy number. Those skilled in the art will understand that this refers to the detection of genes corresponding to the gene products, including, for example, DNA that codes for the gene products. As will be recognized by those skilled in the art, "biomarker" or "marker" includes genes and / or gene products depending on the context.

[0210] The terms "label" and "detectable label" can refer to any composition detectable by spectroscopic, photochemical, biochemical, immunochemical, electrical, optical, chemical, or similar methods. Such labels include biotin for staining with labeled streptavidin conjugates, magnetic beads (e.g., DYNABEADS™), fluorescent dyes (e.g., fluorescein, Texas Red, rhodamine, green fluorescent protein, etc.), radioactive labels (e.g., 3 H, 125 I, 35 S, 14 C, or 32P), enzymes (e.g., horseradish peroxidase, alkaline phosphatase and others commonly used in ELISA), and colorimetric labels such as colloidal gold or colored glass or plastic (e.g., polystyrene, polypropylene, latex, etc.) beads. Patents teaching the use of such labels include U.S. Pat. Nos. 3,817,837; 3,850,752; 3,939,350; 3,996,345; 4,277,437; 4,275,149; and 4,366,241. Means for detecting such labels are well known to those skilled in the art. Thus, for example, radioactive labels may be detected using photographic film or a scintillation counter, and fluorescent markers may be detected using a light detector to detect emitted light. Enzymatic labels are typically detected by providing a substrate to the enzyme and detecting the reaction product produced by the action of the enzyme on the substrate, while colorimetric labels are detected by simply visualizing the colored label. Labels can include, for example, ligands that bind to labeled antibodies, fluorophores, chemiluminescent agents, enzymes, and antibodies that can serve as members of a binding pair specific for the labeled ligand. General descriptions of labels, labeling procedures, and detection of labels can be found in Polak and Van Noorden Introduction to Immunocytochemistry, 2nd ed., Springer Verlag, NY (1997); and Haugland Handbook of Fluorescent Probes and Research Chemicals (1996), a combined handbook and catalog published by Molecular Probes, Inc.

[0211] Detectable labels include, but are not limited to, nucleotides (labeled or unlabeled), compomers, sugars, peptides, proteins, antibodies, chemical compounds, conducting polymers, binding moieties such as biotin, mass tags, colorimetric agents, luminescent agents, chemiluminescent agents, light scattering agents, fluorescent tags, radioactive tags, charge tags (electric or magnetic charge), volatile tags and hydrophobic tags, biomolecules (e.g., members of the binding pairs antibody / antigen, antibody / antibody, antibody / antibody fragment, antibody / antibody receptor, antibody / protein A or protein G, hapten / anti-hapten, biotin / avidin, biotin / streptavidin, folate / folate binding protein, vitamin B12 / intrinsic factor, chemically reactive groups / complementary chemically reactive groups (e.g., sulfhydryl / maleimide, sulfhydryl / haloacetyl derivatives, amine / isotriocyanate, amine / succinimidyl ester, and amine / sulfonyl halide), and the like.

[0212] The terms "primer", "probe", and "oligonucleotide" are used interchangeably herein to refer to relatively short nucleic acid fragments or sequences. They can include DNA, RNA, or hybrids thereof, or chemically modified analogs or derivatives thereof. Typically, they are single-stranded. However, they can also be double-stranded, with two complementary strands that can be separated by denaturation. Usually, primers, probes and oligonucleotides have a length of about 8 nucleotides to about 200 nucleotides, preferably about 12 nucleotides to about 100 nucleotides, more preferably about 18 to about 50 nucleotides. They can be labeled with detectable markers or modified using conventional methods for various molecular biology applications.

[0213] The term "isolated", when used in connection with a nucleic acid (e.g., genomic DNA, cDNA, mRNA, or fragments thereof), is intended to mean that the nucleic acid molecule is present in a form that is substantially separated from other naturally occurring nucleic acids that are normally associated with the molecule. Because naturally occurring chromosomes (or their viral equivalents) contain long nucleic acid sequences, an isolated nucleic acid can be a nucleic acid molecule that has only a portion of the nucleic acid sequence in the chromosome, but does not have one or more other portions that are present on the same chromosome. More specifically, an isolated nucleic acid can include a naturally occurring nucleic acid sequence that is adjacent to a nucleic acid in a naturally occurring chromosome (or its viral equivalent). An isolated nucleic acid can be substantially separated from other naturally occurring nucleic acids on different chromosomes of the same organism. An isolated nucleic acid can also be a composition in which a particular nucleic acid molecule is significantly enriched such that it constitutes at least 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, 95%, or at least 99% of the total nucleic acid in the composition.

[0214] An isolated nucleic acid can be a hybrid nucleic acid having a specific nucleic acid molecule covalently linked with one or more nucleic acid molecules, and the one or more nucleic acid molecules are not adjacent to the specific nucleic acid in nature.For example, an isolated nucleic acid can be in a vector.In addition, a specific nucleic acid can have a nucleotide sequence that is identical to a natural nucleic acid, or its modified form or mutein with one or more mutations, such as nucleotide substitutions, deletions / insertions, inversions, etc.

[0215] An isolated nucleic acid can be prepared from a recombinant host cell (wherein the nucleic acid has been recombinantly amplified and / or expressed), or can be a chemically synthesized nucleic acid having a native nucleotide sequence or an artificially modified form thereof.

[0216] The term "high stringency hybridization conditions" when used in reference to nucleic acid hybridization includes hybridization overnight at 42°C in a solution containing 50% formamide, 5x SSC (750 mM NaCl, 75 mM sodium citrate), 50 mM sodium phosphate, pH 7.6, 5x Denhardt's solution, 10% dextran sulfate, and 20 micrograms / ml denatured, fragmented salmon sperm DNA, with the hybridization filter washed in 0.1x SSC at about 65°C. The term "moderately stringent hybridization conditions" when used in reference to nucleic acid hybridization includes hybridization overnight at 37° C. in a solution containing 50% formamide, 5×SSC (750 mM NaCl, 75 mM sodium citrate), 50 mM sodium phosphate, pH 7.6, 5×Denhardt's solution, 10% dextran sulfate, and 20 micrograms / ml denatured, fragmented salmon sperm DNA, with the hybridization filter washed in 1×SSC at about 50° C. It should be noted that many other hybridization methods, solutions, and temperatures can be used to achieve similarly stringent hybridization conditions, as will be apparent to one of skill in the art.

[0217] For purposes of comparing two different nucleic acid or polypeptide sequences, one sequence (test sequence) may be stated to be identical to another sequence (comparison sequence) at a certain percentage. The percentage of identity can be determined by the algorithm of Karlin and Altschul, Proc. Natl. Acad. Sci. USA, 90:5873-5877 (1993), which is incorporated into various BLAST programs. The percentage of identity can be determined by the "BLAST 2 Sequences" tool available on the National Center for Biotechnology Information (NCBI) website. See Tatusova and Madden, FEMS Microbiol. Lett., 174(2):247-250 (1999). For DNA-DNA pairwise comparisons, the BLASTN program is used with default parameters (e.g., match: 1; mismatch: -2; open gap: 5 penalty; extension gap: 2 penalty; gap x_dropoff: 50; expectation: 10; and word size: 11, with filter). For protein-protein pairwise comparisons, the BLASTP program can be employed with default parameters (e.g., matrix: BLOSUM62; gap open: 11; gap extension: 1; x_dropoff: 15; expectation: 10.0; and word size: 3, with filter). The percent identity of two sequences is calculated by using BLAST to align a test sequence with a comparison sequence, determining the number of amino acids or nucleotides in the aligned test sequence that are identical to amino acids or nucleotides at the same positions in the comparison sequence, and dividing the number of identical amino acids or nucleotides by the number of amino acids or nucleotides in the comparison sequence. When BLAST is used to compare two sequences, BLAST aligns the sequences and results in a percent identity over a given aligned region. When two sequences are aligned over their entire length, the percent identity generated by BLAST is the percent identity of these two sequences.If BLAST does not align two sequences over their entire length, the number of identical amino acids or nucleotides in the unaligned regions of the test and comparison sequences is assumed to be zero, and the percent identity is calculated by adding the number of identical amino acids or nucleotides in the aligned regions and dividing that number by the length of the comparison sequences. Various versions of the BLAST program can be used to compare sequences, for example, BLAST 2.1.2 or BLAST+ 2.2.22.

[0218] The subject or individual can be any animal that may benefit from the methods described herein, including, for example, humans and non-human mammals, such as primates, rodents, horses, dogs and cats. The subject includes, but is not limited to, eukaryotic organisms, most preferably mammals, such as primates, such as chimpanzees or humans, cows; dogs; cats; rodents, such as guinea pigs, rats, mice; rabbits; or birds; reptiles; or fish. The subjects specifically intended for treatment using the methods described herein include humans. The subject may also be referred to herein as an individual or patient. In the present method, the subject has colorectal cancer, for example, has been diagnosed with colorectal cancer. Methods for identifying subjects with colorectal cancer, such as using biopsy, are known in the art. See, e.g., Fleming et al., J Gastrointest Oncol. 2012 Sep; 3(3): 153-173; Chang et al., Dis Colon Rectum. 2012; 55(8):831-43.

[0219] Treatment of a disease or individual by the methods described herein is an approach to obtain a beneficial or desired medical result, including a clinical result, but not necessarily a cure. For the purposes of the methods described herein, beneficial or desired clinical results include, but are not limited to, alleviation or amelioration of one or more symptoms, whether detectable or undetectable, reduction in the extent of disease, stabilization of disease (i.e., not worsening), prevention of disease spread, delay or slowing of disease progression, amelioration or palliation of disease, and remission (whether partial or complete). Treatment also includes prolonging survival compared to the expected survival if no treatment or a different treatment was administered. Treatment can include administration of platinum-based chemotherapy, such as cisplatin, carboplatin, oxaliplatin, and / or nedaplatin. A biomarker generally refers to a molecule, including but not limited to a gene or its product, a nucleic acid (e.g., DNA, RNA), a protein / peptide / polypeptide, a glycostructure, a lipid, a glycolipid, which, when detected in a tissue or cell, has characteristics that can provide predictive, diagnostic, prognostic and / or theranostic information regarding susceptibility or resistance to a candidate treatment.

[0220] Biological samples As used herein, a sample includes any relevant biological sample that can be used for molecular profiling, such as biopsies or tissues taken during surgical or other procedures, body fluids, autopsy samples, and tissue sections such as frozen sections taken for histological purposes. Such samples include blood and blood fractions or products (e.g., serum, buffy coat, plasma, platelets, red blood cells, etc.), sputum, malignant effusions, buccal tissue, cultured cells (e.g., primary cultures, explants, and transformed cells), stool, urine, other biological fluids or fluids (e.g., prostatic fluid, gastric fluid, intestinal fluid, renal fluid, lung fluid, cerebrospinal fluid, etc.), and others. Samples can include biological materials that are fresh frozen and formalin-fixed paraffin-embedded (FFPE) blocks, formalin-fixed paraffin-embedded, or in RNA preservative + formalin fixative. More than one sample of more than one type can be used for each patient. In a preferred embodiment, the sample includes a fixed tumor sample.

[0221] The sample used in the system and method of the present invention can be a formalin-fixed paraffin-embedded (FFPE) sample. The FFPE sample can be one or more of fixed tissue, unstained slide, bone marrow core or clot, core needle biopsy, malignant fluid, and fine needle aspirate (FNA). In one embodiment, the fixed tissue comprises a tumor-containing formalin-fixed paraffin-embedded (FFPE) block from surgery or biopsy. In another embodiment, the unstained slide comprises an unstained, charged, unbaked slide from a paraffin block. In another embodiment, the bone marrow core or clot comprises a decalcified core. The formalin-fixed core and / or clot can be paraffin-embedded. In yet another embodiment, the core needle biopsy comprises 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 or more, e.g., 3-4, paraffin-embedded biopsy samples. An 18-gauge needle biopsy can be used. The malignant fluid may include a sufficient volume of fresh pleural / peritoneal fluid to produce a 5x5x2mm cell pellet. The fluid may be formalin fixed in a paraffin block. In some embodiments, the core needle biopsy includes 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 or more, e.g., 4-6, paraffin-embedded aspirates.

[0222] The sample may be processed according to techniques understood by those skilled in the art. The sample may be, but is not limited to, fresh, frozen, or fixed cells or tissues. In some embodiments, the sample includes formalin-fixed paraffin-embedded (FFPE) tissue, fresh tissue, or fresh frozen (FF) tissue. The sample may include cultured cells, including primary or immortalized cell lines, derived from a subject sample. The sample may also refer to an extract from a sample from a subject. For example, the sample may include DNA, RNA, or protein extracted from tissue or body fluid. Many techniques and commercial kits are available for such purposes. Fresh samples from individuals may be treated with agents to preserve RNA before further processing, such as cell lysis and extraction. The sample may include frozen samples collected for other purposes. The sample may be associated with relevant information, such as age, sex, and clinical symptoms present in the subject; origin of the sample; and method of sample collection and storage. The sample is typically obtained from a subject.

[0223] Biopsy includes the process of removing a tissue sample for diagnostic or prognostic evaluation, and the tissue specimen itself. Any biopsy technique known in the art can be applied to the molecular profiling method of the present disclosure. The biopsy technique applied can depend on the type of tissue to be evaluated (e.g., colon, prostate, kidney, bladder, lymph node, liver, bone marrow, blood cells, lung, breast, etc.), the size and type of tumor (e.g., solid or floating, blood or ascites), among other factors. Representative biopsy techniques include, but are not limited to, excision biopsy, incision biopsy, needle biopsy, surgical biopsy, and bone marrow biopsy. An "excision biopsy" refers to the removal of an entire tumor mass with a small margin of normal tissue surrounding it. An "incision biopsy" refers to the removal of a wedge of tissue that includes the cross-sectional diameter of the tumor. Molecular profiling can use a "core needle biopsy" of the tumor mass, or a "fine needle aspiration biopsy" that typically obtains a suspension of cells from within the tumor mass. Biopsy techniques are discussed, for example, in Chapter 70 and throughout Part V of Harrison's Principles of Internal Medicine, Kasper, et al., eds., 16th ed., 2005.

[0224] Unless otherwise stated, the "sample" referred to herein for molecular profiling of a patient may include more than one physical specimen.As a non-limiting example, a "sample" may include multiple sections from a tumor, such as multiple sections of an FFPE block or multiple core needle biopsy sections.As another non-limiting example, a "sample" may include multiple biopsy specimens, such as one or more surgical biopsy specimens, one or more core needle biopsy specimens, one or more fine needle aspiration biopsy specimens, or any useful combination thereof.As yet another non-limiting example, a molecular profile may be generated for a subject using a "sample" that includes a solid tumor specimen and a body fluid specimen.In some embodiments, a sample is a unit sample, i.e., a single physical specimen.

[0225] Standard molecular biology techniques that are known in the art and not specifically described are generally described in Sambrook et al., Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory Press, New York (1989), and Ausubel et al., Current Protocols in Molecular Biology, John Wiley and Sons, Baltimore, Md. (1989), and Perbal, A Practical Guide to Molecular Cloning, John Wiley & Sons, New York (1988), and Watson et al., Recombinant DNA, Scientific American Books, New York, and Birren et al (eds) Genome Analysis: A Laboratory Manual Series, Vols. 1-4 Cold Spring Harbor Laboratory Press, New York (1998), and in U.S. Patent Nos. 4,666,828; 4,683,202; 4,801,531; 5,192,659 and 5,272,057, which are incorporated herein by reference. Polymerase chain reaction (PCR) can be generally performed as in PCR Protocols: A Guide to Methods and Applications, Academic Press, San Diego, Calif. (1990).

[0226] Vesicle The sample can include a vesicle. The method as described herein can include examining one or more vesicles, including examining a vesicle population. A vesicle, as used herein, is a membrane vesicle that has been shed from a cell. Vesicles or membrane vesicles include, but are not limited to, circulating microvesicles (cMVs), microvesicles, exosomes, nanovesicles, dexosomes, blebs, blebbies, prostasomes, microparticles, intraluminal vesicles, membrane fragments, intraluminal endosomal vesicles, endosome-like vesicles, exocytotic vehicles, endosome vesicles, endosomal vesicles, apoptotic bodies, multivesicular bodies, secretory vesicles, phospholipid vesicles, liposomal vesicles, argosomes, texas, secresomes, tolerosomes, melanosomes, oncosomes, or exocytosed vehicles. Additionally, although vesicles may be produced by different cellular processes, methods as described herein are not limited to or dependent on any one mechanism, so long as such vesicles are present in a biological sample and can be characterized by the methods disclosed herein. Unless otherwise specified, methods utilizing one type of vesicle can be applied to other types of vesicles. Vesicles comprise spherical structures with a lipid bilayer, similar to a cell membrane, surrounding an internal compartment that can contain soluble components, sometimes referred to as the payload. In some embodiments, methods as described herein utilize exosomes, which are small secreted vesicles with diameters of about 40-100 nm. For a review of membrane vesicles, including types and characterization, see Thery et al., Nat Rev Immunol. 2009 Aug;9(8):581-93. Some properties of different types of vesicles include those listed in Table 1:

[0227] Table 1. Vesicle characteristics TIFF0007624992000001.tif148169Abbreviation: Phosphatidylserine (PPS); Electron microscopy (EM)

[0228] Vesicles include excreted membrane-bound particles or "microparticles" derived from either the plasma membrane or endomembranes. Vesicles can be released from cells into the extracellular environment. Cells releasing vesicles include, but are not limited to, cells derived from or derived from the ectoderm, endoderm, or mesoderm. Cells may be genetic, environmental, and / or any other variation or modification. For example, the cells can be tumor cells. Vesicles can reflect any changes in the source cell, thereby reflecting changes in the derived cell, for example, cells with various genetic mutations. In one mechanism, vesicles are generated intracellularly when segments of the cell membrane naturally invaginate and are eventually exocytosed (see, e.g., Keller et al., Immunol. Lett. 107 (2): 102-8 (2006)). Vesicles also include cell-derived structures bound by lipid bilayer membranes that result from both separation of escaped blebbing and sealing of plasma membrane portions, or from the export of any intracellular membrane-bound vesicular structures containing various membrane-associated proteins of tumor origin, which contain surface-bound molecules from the host circulation that selectively bind tumor-derived proteins together with molecules contained in the vesicle lumen, including but not limited to tumor-derived microRNAs or intracellular proteins. Blebs and blebbing are further described in Charras et al., Nature Reviews Molecular and Cell Biology, Vol. 9, No. 11, p. 730-736 (2008). Vesicles that are released from tumor cells into circulation or body fluids may be referred to as "circulating tumor-derived vesicles". When such vesicles are exosomes, the vesicles may be referred to as circulating tumor-derived exosomes (CTEs). In some cases, the vesicles can be derived from a specific cell origin. CTEs, like cell origin-specific vesicles, typically have one or more unique biomarkers that allow for isolation of the CTEs or cell origin-specific vesicles, e.g., from bodily fluids, sometimes in a specific manner.For example, cell or tissue specific markers are used to identify cell origin.Examples of such cell or tissue specific markers are disclosed herein and can be found in the Tissue-specific Gene Expression and Regulation (TiGER) database available at bioinfo.wilmer.jhu.edu / tiger / ; Liu et al. (2008) TiGER: a database for tissue-specific gene expression and regulation. BMC Bioinformatics. 9:271; TissueDistributionDBs available at genome.dkfz-heidelberg.de / menu / tissue_db / index.html.

[0229] The vesicles can have a diameter of greater than about 10 nm, 20 nm, or 30 nm. The vesicles can have a diameter of greater than 40 nm, 50 nm, 100 nm, 200 nm, 500 nm, 1000 nm, or greater than 10,000 nm. The vesicles can have a diameter of about 30-1000 nm, about 30-800 nm, about 30-200 nm, or about 30-100 nm. In some embodiments, the vesicles have a diameter of less than 10,000 nm, 1000 nm, 800 nm, 500 nm, 200 nm, 100 nm, 50 nm, 40 nm, 30 nm, 20 nm, or less than 10 nm. As used herein, the term "about" in connection with a numerical value means that a variance of 10% above and below the numerical value is within the range ascribed to the particular value. Typical sizes for various types of vesicles are shown in Table 1. The vesicles can be examined to measure the diameter of a single vesicle or any number of vesicles. For example, the diameter range of a vesicle population or the average diameter of a vesicle population can be determined. The diameter of a vesicle can be examined using methods known in the art, such as imaging techniques, such as electron microscopy. In some embodiments, the diameter of one or more vesicles is determined using optical particle detection. For example, see U.S. Patent No. 7,751,053, entitled "Optical Detection and Analysis of Particles," issued on July 6, 2010; and U.S. Patent No. 7,399,600, entitled "Optical Detection and Analysis of Particles," issued on July 15, 2010.

[0230] In some embodiments, vesicles are assayed directly from a biological sample without prior isolation, purification, or enrichment from the biological sample. For example, the amount of vesicles in a sample can provide a biosignature that provides a diagnostic, prognostic, or theranostic determination by itself. Alternatively, vesicles in a sample may be isolated, captured, purified, or enriched from the sample prior to analysis. As described above, isolation, capture, or purification, as used herein, includes partial isolation, partial capture, or partial purification away from other components in the sample. Vesicles can be isolated using a variety of techniques, such as those described herein or known in the art, including, but not limited to, size exclusion chromatography, density gradient centrifugation, differential centrifugation, nanomembrane ultrafiltration, immunoadsorption capture, affinity purification, affinity capture, immunoassay, immunoprecipitation, microfluidic separation, flow cytometry, or combinations thereof.

[0231] Vesicles can be examined and vesicle characteristics compared to standards to provide phenotypic characterization. In some embodiments, surface antigens on the vesicles are examined. Vesicles or vesicle populations that carry a particular marker can be referred to as positive (biomarker+) vesicles or vesicle populations. For example, a DLL4+ population refers to a vesicle population that is bound to DLL4. Conversely, a DLL4- population is not bound to DLL4. Surface antigens can provide an indication of the anatomical and / or cellular origin of the vesicles as well as other phenotypic information, such as the state of a tumor. For example, vesicles found in a patient sample can be examined for surface antigens indicative of colorectal origin and the presence of cancer, thereby identifying vesicles associated with colorectal cancer cells. Surface antigens may include any information-bearing biological entity that can be detected on the vesicle membrane surface, including, but not limited to, surface proteins, lipids, carbohydrates, and other membrane components. For example, positive detection of vesicles obtained from the colon expressing a tumor antigen can indicate that the patient has colorectal cancer. In this manner, methods as described herein can be used to characterize any disease or condition associated with an anatomical or cellular origin, for example, by examining disease-specific and cell-specific biomarkers in one or more vesicles obtained from a subject.

[0232] In embodiments, one or more vesicle payloads are examined to provide phenotypic characterization. A vesicle-bearing payload includes any informative biological entity that can be detected as encapsulated within a vesicle, including, but not limited to, proteins and nucleic acids, e.g., genomes or cDNAs, mRNAs, or functional fragments thereof, as well as microRNAs (miRs). In addition, methods as described herein are directed to detecting vesicle surface antigens (in addition to or exclusively with the vesicle payload) to provide phenotypic characterization. For example, vesicles can be characterized by using binding agents (e.g., antibodies or aptamers) specific for vesicle surface antigens, and bound vesicles can be further examined to identify one or more payload components disclosed herein. As described herein, the level of a surface antigen of interest or a vesicle bearing a payload of interest can be compared to a reference to characterize a phenotype. For example, overexpression in a sample of a cancer-associated surface antigen or vesicle payload, e.g., tumor-associated mRNA or microRNA, compared to a reference can indicate the presence of cancer in the sample. The biomarkers examined may be present or absent, increased or decreased, based on the selection of the desired target sample and comparison of the target sample with the desired reference sample. Non-limiting examples of target samples include disease; treated / untreated; different time points, e.g., in longitudinal studies; non-limiting examples of reference samples include non-disease; normal; different time points; and sensitive or resistant to the candidate treatment.

[0233] In certain aspects, molecular profiling as described herein involves the analysis of microvesicles, such as circulating microvesicles.

[0234] MicroRNA A variety of biomarker molecules can be examined in biological samples or vesicles obtained from such biological samples. MicroRNAs comprise one class of biomarkers examined via methods as described herein. MicroRNAs, also referred to herein as miRNAs or miRs, are short RNA strands approximately 21-23 nucleotides in length. miRNAs are encoded by genes that are transcribed from DNA but not translated into proteins, and thus comprise non-coding RNAs. miRs are processed from primary transcripts known as pri-miRNAs to short stem-loop structures called pre-miRNAs and finally to the resulting single-stranded miRNAs. The pre-miRNAs typically form structures that fold back on themselves in self-complementary regions. These structures are then processed by the nuclease Dicer in animals or DCL1 in plants. Mature miRNA molecules are partially complementary to one or more messenger RNA (mRNA) molecules and can function to regulate protein translation. Identified sequences of miRNAs can be accessed in publicly available databases such as www.microRNA.org, www.mirbase.org, or www.mirz.unibas.ch / cgi / miRNA.cgi.

[0235] miRNAs are generally assigned numbers according to the naming convention "mir-[number]". miRNA numbers are assigned according to their order of discovery relative to previously identified miRNA species. For example, if the last published miRNA was mir-121, the next discovered miRNA would be named mir-122, and so on. If a miRNA is discovered to be homologous to a known miRNA from a different organism, its name can be given an optional organism identifier in the format [organism identifier]-mir-[number]. Identifiers include hsa for Homo sapiens and mmu for Mus Musculus. For example, the human homolog of mir-121 may be referred to as hsa-mir-121, while the mouse homolog may be referred to as mmu-mir-121.

[0236] Mature microRNAs are usually named with the prefix "miR", while genes or precursor miRNAs are named with the prefix "mir". For example, mir-121 is the precursor for miR-121. When different miRNA genes or precursors are processed into the same mature miRNA, the genes / precursors can be described by numbered suffixes. For example, mir-121-1 and mir-121-2 can refer to separate genes or precursors that are processed into miR-121. Lettered suffixes are used to indicate closely related mature sequences. For example, mir-121a and mir-121b can be processed into closely related miRNAs, miR-121a and miR-121b, respectively. In the context of this disclosure, any microRNA (miRNA or miR) named herein with the prefix mir-* or miR-* is understood to encompass both the precursor and / or mature species, unless otherwise specified.

[0237] Sometimes, it is observed that two mature miRNA sequences originate from the same precursor. When one of the sequences is more abundant than the other, the "*" suffix can be used to name the less common variant. For example, miR-121 is the main product, while miR-121* is the less common variant found in the opposite arm of the precursor. When the main variant is not identified, miRs can be identified by the suffix "5p" for variants from the 5' arm of the precursor and "3p" for variants from the 3' arm. For example, miR-121-5p originates from the 5' arm of the precursor, while miR-121-3p originates from the 3' arm. Less commonly, the 5p and 3p variants are referred to as sense ("s") and antisense ("as") forms, respectively. For example, miR-121-5p may be referred to as miR-121-s, while miR-121-3p may be referred to as miR-121-as.

[0238] The above naming conventions have evolved over time and are general guidelines rather than absolute prescriptions. For example, the let and lin families of miRNAs continue to be referred to by their nicknames. The mir / miR convention for precursor / mature forms is also a guideline and context should be taken into account to determine which form is referred to. Further details of miR nomenclature can be found at www.mirbase.org or in Ambros et al., A uniform system for microRNA annotation, RNA 9:277-279 (2003).

[0239] Plant miRNAs follow a different naming convention as described in Meyers et al., Plant Cell. 2008 20(12):3186-3190.

[0240] Several miRNAs are involved in gene regulation, and miRNAs are part of an expanding class of non-coding RNAs that are now recognized as a major layer of gene control. In some cases, miRNAs can disrupt translation by binding to regulatory sites embedded in the 3'-UTR of target mRNAs, resulting in translational repression. Target recognition involves complementary base pairing of the target site with the seed region of the miRNA (positions 2-8 at the 5' end of the miRNA). However, the exact degree of seed complementarity is not precisely determined and can be modified by 3' pairing. In other cases, miRNAs function like small interfering RNAs (siRNAs), binding to perfectly complementary mRNA sequences to disrupt target transcripts.

[0241] Characterization of some miRNAs has shown that they affect a variety of processes, including early development, cell proliferation and cell death, apoptosis and fat metabolism.For example, some miRNAs, such as lin-4, let-7, mir-14, mir-23 and bantam, have been shown to play important roles in cell differentiation and tissue development.Others are also believed to have important roles due to their differential spatial and temporal expression patterns.

[0242] The miRNA database available at miRBase (www.mirbase.org) contains a searchable database of published miRNA sequences and annotations. Further information regarding miRBase can be found in the following documents, each of which is incorporated herein by reference in its entirety: Griffiths-Jones et al., miRBase: tools for microRNA genomics. NAR 2008 36(Database Issue):D154-D158; Griffiths-Jones et al., miRBase: microRNA sequences, targets and gene nomenclature. NAR 2006 34(Database Issue):D140-D144; and Griffiths-Jones, S. The microRNA Registry. NAR 2004 32(Database Issue):D109-D111. Representative miRNAs included in release 16 of miRBase were made available in September 2010.

[0243] As described herein, microRNA is known to be involved in cancer and other diseases, and can be examined to characterize phenotypes in samples.See, for example, Ferracin et al., Micromarkers: miRNAs in cancer diagnosis and prognosis, Exp Rev Mol Diag, Apr 2010, Vol. 10, No. 3, Pages 297-308; Fabbri, miRNAs as molecular biomarkers of cancer, Exp Rev Mol Diag, May 2010, Vol. 10, No. 4, Pages 435-444.

[0244] In certain aspects, molecular profiling as described herein includes the analysis of microRNAs.

[0245] Techniques for isolating and characterizing vesicles and miRs are known to those of skill in the art. In addition to the methodologies presented herein, additional methods are described in U.S. Pat. No. 7,888,035, issued Feb. 15, 2011, entitled "METHODS FOR ASSESSING RNA PATTERNS"; and U.S. Pat. No. 7,897,356, issued Mar. 1, 2011, entitled "METHODS AND SYSTEMS OF USING EXOSOMES FOR DETERMINING PHENOTYPES"; as well as International Patent Publications WO / 2011 / 066589, issued Nov. 30, 2010, entitled "METHODS AND SYSTEMS FOR ISOLATING, STORING, AND ANALYZING VESICLES"; WO / 2011 / 088226, issued Jan. 13, 2011, entitled "DETECTION OF GASTROINTESTINAL DISORDERS"; and WO / 2011 / 088226, issued Jan. 13, 2011, entitled "BIOMARKERS FOR and WO / 2011 / 127219, published on April 6, 2011, entitled "CIRCULATING BIOMARKERS FOR DISEASE," each of which is incorporated by reference in its entirety.

[0246] Circulating biomarkers Circulating biomarkers include biomarkers that are detectable in bodily fluids, such as blood, plasma, and serum. Examples of circulating cancer biomarkers include cardiac troponin T (cTnT), prostate specific antigen (PSA) for prostate cancer, and CA125 for ovarian cancer. Circulating biomarkers according to the present disclosure include any suitable biomarker that can be detected in bodily fluids, including but not limited to proteins, nucleic acids, such as DNA, mRNA and microRNA, lipids, carbohydrates, and metabolites. Circulating biomarkers can include biomarkers that are not associated with cells, such as biomarkers that are membrane-bound, embedded in membrane fragments, part of a biological complex, or free in solution. In one embodiment, circulating biomarkers are biomarkers that are associated with one or more vesicles present in a biological fluid of a subject.

[0247] Circulating biomarkers have been identified for use in characterizing various phenotypes, including the detection of cancer. For example, Ahmed N, et al., Proteomic-based identification of haptoglobin-1 precursor as a novel circulating biomarker of ovarian cancer. Br. J. Cancer 2004; Mathelin _et al., Circulating proteinic biomarkers and breast cancer, Gynecol Obstet Fertil. 2006 Jul-Aug;34(7-8):638-46. Epub 2006 Jul 28; Ye et al., Recent technical strategies to identify diagnostic biomarkers for ovarian cancer. Expert Rev Proteomics. 2007 Feb;4(1):121-31; Carney, Circulating oncoproteins HER2 / neu, EGFR and CAIX (MN) as novel cancer biomarkers. Expert Rev Mol Diagn. 2007 May;7(3):309-19; Gagnon, Discovery and application of protein biomarkers for ovarian cancer, Curr Opin Obstet Gynecol. 2008 Feb;20(1):9-13; Pasterkamp et al., Immune regulatory cells: circulating biomarker factories in cardiovascular disease. Clin Sci (Lond). 2008 Aug;115(4):129-31; Fabbri, miRNAs as biomarker molecules of cancer, Exp Rev Mol Diag, May 2010, Vol. 10, No.4, Pages 435-444; PCT Patent Publication WO / 2007 / 088537; U.S. Patent Nos. 7,745,150 and 7,655,479; U.S. Patent Application Publication Nos. 20110008808, 20100330683, 20100248290, 20100222230, 20100203566, 20100173788, 20090291932, 20100203566, 20100173788, 20090291932, 20100203566, 20100203566, 20100173788, 20090291932, 20100203566, 20100203566, 20100173788, 2010020356 ... See US Pat. Nos. 20090239246, 20090226937, 20090111121, 20090004687, 20080261258, 20080213907, 20060003465, 20050124071, and 20040096915, each of which is incorporated by reference in its entirety. In certain aspects, molecular profiling as described herein includes analysis of circulating biomarkers.

[0248] Gene expression profiling The methods and systems as described herein include expression profiling, which involves examining the differential expression of one or more target genes disclosed herein. Differential expression can include overexpression and / or underexpression of a biological product, e.g., a gene, mRNA, or protein, compared to a control (or standard). The control can include cells similar to the sample but without the disease (e.g., expression profile obtained from a sample from a healthy individual). The control can be a previously determined level that indicates the effectiveness of a drug target associated with a particular disease and a particular drug target. The control can be from the same patient, e.g., a normal adjacent part of the same organ as the diseased cells, or the control can be obtained from healthy tissue from another patient, or it can be a previously determined threshold value that indicates that the disease is responsive or unresponsive to a particular drug target. The control can also be a control found in the same sample, e.g., a housekeeping gene or its product (e.g., mRNA or protein). For example, the control nucleic acid can be one that is known to have no difference depending on the cancerous or non-cancerous state of the cell. The expression level of the control nucleic acid can be used to normalize the signal level in the test population and the reference population. Illustrative control genes include, but are not limited to, for example, β-actin, glyceraldehyde 3-phosphate dehydrogenase, and ribosomal protein P1. Multiple controls or types of controls can be used. The cause of differential expression can vary. For example, gene copy number can increase in cells, resulting in increased expression of the gene. Alternatively, the transcription of the gene can be modified, for example, by chromatin remodeling, differential methylation, differential expression or activity of transcription factors, etc. Translation can also be modified, for example, by differential expression of factors that degrade mRNA, translate mRNA, or silence translation, such as microRNA or siRNA. In some embodiments, differential expression includes differential activity. For example, a protein can carry a mutation that increases the activity of the protein, such as constitutive activation, that contributes to a disease state.Molecular profiling, which reveals changes in activity, can be used to guide treatment choices.

[0249] Methods of gene expression profiling include methods based on polynucleotide hybridization analysis and methods based on polynucleotide sequencing.The commonly used methods known in the art for quantifying mRNA expression in a sample include Northern blotting and in situ hybridization (Parker & Barnes (1999) Methods in Molecular Biology 106:247-283); RNase protection assay (Hod (1992) Biotechniques 13:852-854); and reverse transcription polymerase chain reaction (RT-PCR) (Weis et al. (1992) Trends in Genetics 8:263-264).Alternatively, antibodies may be employed that can recognize specific duplexes, including DNA duplexes, RNA duplexes, and DNA-RNA hybrid duplexes or DNA-protein duplexes. Exemplary methods for sequencing-based gene expression analysis include Serial Analysis of Gene Expression (SAGE), gene expression analysis by massively parallel signature sequencing (MPSS), and / or next-generation sequencing.

[0250] RT-PCR Reverse transcription polymerase chain reaction (RT-PCR) is a variation of polymerase chain reaction (PCR). With this technique, RNA strand is reverse transcribed into its DNA complement (i.e., complementary DNA, or cDNA) using an enzyme called reverse transcriptase, and the resulting cDNA is amplified using PCR. Real-time polymerase chain reaction is another PCR variation, also called quantitative PCR, Q-PCR, qRT-PCR, or sometimes RT-PCR. Either reverse transcription PCR or real-time PCR can be used for molecular profiling according to the present disclosure, and RT-PCR can be represented unless otherwise specified or as understood by those skilled in the art.

[0251] RT-PCR can be used to determine the RNA level, for example, mRNA or miRNA level, of biomarkers as described herein.RT-PCR can be used to compare such RNA levels of biomarkers as described herein in different sample populations, in normal tissues and tumor tissues, with or without drug treatment, to characterize patterns of gene expression, to distinguish closely related RNAs, and to analyze RNA structure.

[0252] The first step is the isolation of RNA, for example, mRNA, from a sample.The starting material can be total RNA isolated from human tumor or tumor cell line, and corresponding normal tissue or cell line, respectively.Thus, RNA can be isolated from a sample, for example, tumor cell or tumor cell line, and compared with pooled DNA from healthy donors.When the origin of mRNA is primary tumor, mRNA can be extracted, for example, from frozen tissue sample or archival tissue sample that is paraffin-embedded and fixed (for example, formalin-fixed).

[0253] General methods for mRNA extraction are well known in the art and are disclosed in standard textbooks of molecular biology, including Ausubel et al. (1997) Current Protocols of Molecular Biology, John Wiley and Sons. Methods for extracting RNA from paraffin-embedded tissues are disclosed, for example, in Rupp & Locker (1987) Lab Invest. 56:A67, and De Andres et al., BioTechniques 18:42044 (1995). In particular, RNA isolation can be performed using purification kits, buffer sets and proteases from commercial manufacturers such as Qiagen, according to the manufacturer's instructions (QIAGEN Inc., Valencia, CA). For example, total RNA from cultured cells can be isolated using Qiagen RNeasy mini columns. Numerous RNA isolation kits are commercially available and can be used in the methods described herein.

[0254] Alternatively, the first step is the isolation of miRNA from target samples.The starting material is typically total RNA isolated from human tumors or tumor cell lines, and corresponding normal tissues or cell lines, respectively.Thus, RNA can be isolated from various primary tumors or tumor cell lines, together with pooled DNA from healthy donors.When the source of miRNA is primary tumor, miRNA can be extracted, for example, from frozen tissue samples or archival tissue samples that are paraffin-embedded and fixed (e.g., formalin-fixed).

[0255] General methods for miRNA extraction are well known in the art and are disclosed in standard textbooks of molecular biology, including Ausubel et al. (1997) Current Protocols of Molecular Biology, John Wiley and Sons. Methods for extracting RNA from paraffin-embedded tissues are disclosed, for example, in Rupp & Locker (1987) Lab Invest. 56:A67, and De Andres et al., BioTechniques 18:42044 (1995). In particular, RNA isolation can be performed using purification kits, buffer sets and proteases from commercial manufacturers such as Qiagen, according to the manufacturer's instructions. For example, total RNA from cultured cells can be isolated using Qiagen RNeasy mini columns. Numerous miRNA isolation kits are commercially available and can be used in the methods described herein.

[0256] Whether the RNA comprises mRNA, miRNA or other types of RNA, gene expression profiling by RT-PCR can involve reverse transcription of the RNA template into cDNA followed by amplification in a PCR reaction. Commonly used reverse transcriptases include, but are not limited to, avian myeloblastosis virus reverse transcriptase (AMV-RT) and Moloney murine leukemia virus reverse transcriptase (MMLV-RT). The reverse transcription step is typically primed using specific primers, random hexamers, or oligo-dT primers, depending on the context and goal of expression profiling. For example, extracted RNA can be reverse transcribed using a GeneAmp RNA PCR kit (Perkin Elmer, Calif., USA) according to the manufacturer's instructions. The resulting cDNA can then be used as a template in a subsequent PCR reaction.

[0257] The PCR step typically employs Taq DNA polymerase, which has 5'-3' nuclease activity but lacks 3'-5' proofreading endonuclease activity, although a variety of thermostable DNA-dependent DNA polymerases can be used. TaqMan PCR typically uses the 5'-nuclease activity of Taq or Tth polymerase to hydrolyze hybridization probes bound to the target amplicon, although any enzyme with comparable 5' nuclease activity can be used. Two oligonucleotide primers are used to generate an amplicon typical of a PCR reaction. A third oligonucleotide, or probe, is designed to detect the nucleotide sequence located between the two PCR primers. The probe cannot be extended by the Taq DNA polymerase enzyme and is labeled with a reporter fluorescent dye and a quencher fluorescent dye. Any laser-induced emission from the reporter dye is quenched by the quencher dye when the two dyes are located in close proximity to each other on the probe. During the course of the amplification reaction, the Taq DNA polymerase enzyme cleaves the probe in a template-dependent manner. The resulting probe fragments dissociate in solution, and the signal from the released reporter dye is freed from the quenching effect of the second fluorophore. Because one molecule of reporter dye is liberated for each new molecule synthesized, detection of the unquenched reporter dye provides the basis for quantitative interpretation of the data.

[0258] TaqMan™ RT-PCR can be performed using commercially available equipment, such as the ABI PRISM 7700™ Sequence Detection System™ (Perkin-Elmer-Applied Biosystems, Foster City, Calif., USA), or the LightCycler (Roche Molecular Biochemicals, Mannheim, Germany). In a particular embodiment, the 5' nuclease procedure is performed in a real-time quantitative PCR device, such as the ABI PRISM 7700 Sequence Detection System. The system consists of a thermocycler, a laser, a charge-coupled device (CCD), a camera, and a computer. The system amplifies samples in a 96-well format by the thermocycler. During amplification, the laser-induced fluorescent signal is collected in real time for all 96 wells through a fiber optic cable and detected by the CCD. The system includes software for running the instrument and for analyzing the data.

[0259] TaqMan data are initially expressed as Ct, or threshold cycle. As mentioned above, fluorescence values ​​are recorded during each cycle and represent the amount of product amplified to that point in the amplification reaction. The point at which the fluorescent signal is first recorded as statistically significant is the threshold cycle (Ct).

[0260] To minimize the effects of error and sample-to-sample variation, RT-PCR is usually performed using an internal standard. An ideal internal standard is expressed at a constant level among different tissues and is not affected by the experimental treatment. The RNAs most frequently used to normalize patterns of gene expression are the mRNAs for the housekeeping genes, glyceraldehyde-3-phosphate-dehydrogenase (GAPDH) and β-actin.

[0261] Real-time quantitative PCR (also known as quantitative real-time polymerase chain reaction, QRT-PCR or Q-PCR) is a more recent variation of the RT-PCR technique. Q-PCR can measure the accumulation of PCR products through a dual-labeled fluorogenic probe (i.e., TaqMan probe). Real-time PCR is compatible with both quantitative competitive PCR, in which an internal competitor for each target sequence is used for normalization, and quantitative comparative PCR, which uses normalization genes contained within the sample or housekeeping genes for RT-PCR. See, for example, Held et al. (1996) Genome Research 6:986-994.

[0262] Protein-based detection techniques are also useful for molecular profiling, especially when the nucleotide variants cause amino acid substitutions or deletions or insertions or frameshifts that affect the primary, secondary or tertiary structure of the protein. Protein sequencing techniques may be used to detect amino acid variations. For example, a protein or fragment thereof corresponding to a gene can be synthesized by recombinant expression using DNA fragments isolated from a test individual. Preferably, a cDNA fragment of 100-150 base pairs or less that encompasses the polymorphic locus to be determined is used. The amino acid sequence of the peptide can then be determined by conventional protein sequencing methods. Alternatively, HPLC-microscopy tandem mass spectrometry techniques can be used to determine amino acid sequence variations. In this technique, a proteolytic digestion is performed on the protein and the resulting peptide mixture is separated by reversed-phase chromatographic separation. Tandem mass spectrometry is then performed and the collected data is analyzed. See Gatlin et al., Anal. Chem., 72:757-763 (2000).

[0263] Microarrays Biomarkers as described herein can also be identified, confirmed, and / or measured using microarray techniques. Thus, expression profile biomarkers can be measured in cancer samples using microarray techniques. In this method, polynucleotide sequences of interest are plated or arrayed on a microchip substrate. The arrayed sequences are then hybridized with specific DNA probes from cells or tissues of interest. The mRNA source can be total RNA isolated from samples, such as human tumors or tumor cell lines and corresponding normal tissues or cell lines. Thus, RNA can be isolated from a variety of primary tumors or tumor cell lines. When the mRNA source is a primary tumor, mRNA can be extracted, for example, from frozen tissue samples or paraffin-embedded and fixed (e.g., formalin-fixed) preserved tissue samples, which are routinely prepared and preserved in daily clinical practice.

[0264] The expression profile of biomarkers can be measured in either fresh or paraffin-embedded tumor tissue or body fluid using microarray technique.In this method, the polynucleotide sequence of interest is plated or arrayed on a microchip substrate.The arrayed sequence is then hybridized with specific DNA probe from the cell or tissue of interest.Similar to RT-PCR method, the source of miRNA is typically the total RNA isolated from human tumor or tumor cell line and corresponding normal tissue or cell line, including body fluid such as serum, urine, tear and exosome.Thus, RNA can be isolated from various sources.When the source of miRNA is primary tumor, miRNA can be extracted from, for example, frozen tissue sample, which is routinely prepared and stored in daily clinical practice.

[0265] The cDNA microarray technique, also known as biochip, DNA chip or gene array, allows the identification of gene expression levels in biological samples.The cDNA or oligonucleotides representing each given gene are immobilized and tagged on a substrate, such as a small chip, bead or nylon membrane, and serve as probes to indicate whether they are expressed in the biological sample of interest.The simultaneous expression of several thousand genes can be monitored simultaneously.

[0266] In a particular embodiment of the microarray technique, the PCR amplified inserts of cDNA clones are applied to the substrate in a high-density array.In one aspect, at least 100, 200, 300, 400, 500, 600, 700, 800, 900, 1,000, 1,500, 2,000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, 10,000, 15,000, 20,000, 25,000, 30,000, 35,000, 40,000, 45,000 or at least 50,000 nucleotide sequences are applied to the substrate.Each sequence can correspond to a different gene, or multiple sequences can be arrayed per gene.The microarrayed genes immobilized on the microchip are suitable for hybridization under stringent conditions. Fluorescently labeled cDNA probes may be generated through the incorporation of fluorescent nucleotides by reverse transcription of RNA extracted from tissues of interest. The labeled cDNA probes applied to the chip specifically hybridize to each spot of DNA on the array. After stringent washing to remove non-specifically bound probes, the chip is scanned by confocal laser microscopy or by another detection method such as a CCD camera. Quantitation of hybridization of each arrayed element allows the corresponding mRNA abundance to be examined. cDNA probes generated from two RNA sources and separately labeled with dual-color fluorescence are hybridized to the array in pairs. Thus, the relative abundance of the transcripts from the two sources corresponding to each specific gene is determined simultaneously. Miniaturized scale hybridization allows convenient and rapid evaluation of expression patterns for a large number of genes. Such methods have been shown to have the sensitivity required to detect rare transcripts expressed at a few copies per cell and to reproducibly detect at least approximately two-fold differences in expression levels (Schena et al. (1996) Proc. Natl. Acad. Sci. USA 93(2):106-149).Microarray analysis can be performed using commercially available equipment following manufacturer's protocols, including but not limited to Affymetrix GeneChip technology (Affymetrix, Santa Clara, CA), Agilent (Agilent Technologies, Inc., Santa Clara, CA), or Illumina (Illumina, Inc., San Diego, CA) microarray technology.

[0267] The development of microarray methods for large-scale analysis of gene expression allows the systematic search for molecular markers of cancer classification and outcome prediction in diverse tumor types.

[0268] In some embodiments, the Agilent Whole Human Genome Microarray Kit (Agilent Technologies, Inc., Santa Clara, Calif.) The system is capable of analyzing more than 41,000 unique human genes and transcripts, all of which are represented by public domain annotations. The system is used according to the manufacturer's instructions.

[0269] In some embodiments, the Illumina Whole Genome DASL assay (Illumina Inc., San Diego, Calif.) is used. The system provides a method for simultaneously profiling over 24,000 transcripts in a high-throughput manner from minimal RNA input from both fresh frozen (FF) and formalin-fixed paraffin-embedded (FFPE) tissue sources.

[0270] Microarray expression analysis includes identifying whether a gene or gene product is up-regulated or down-regulated compared to a reference. Identification can be performed using a statistical test to determine the statistical significance of any observed differential expression. In some embodiments, statistical significance is determined using a parametric statistical test. Parametric statistical tests can include, for example, fractional factorial design, analysis of variance (ANOVA), t-test, least squares, Pearson correlation, linear simple regression, nonlinear regression, multiple linear regression, or multiple nonlinear regression. Alternatively, parametric statistical tests can include one-way analysis of variance, two-way analysis of variance, or repeated measures analysis of variance. In other embodiments, statistical significance is determined using a nonparametric statistical test. Examples include, but are not limited to, Wilcoxon signed rank test, Mann-Whitney test, Kruskal-Wallis test, Friedman test, Spearman's rank correlation coefficient, Kendall's tau analysis, and nonparametric regression tests. In some embodiments, statistical significance is determined by a p-value of less than about 0.05, 0.01, 0.005, 0.001, 0.0005, or 0.0001.Although the microarray system used in the method as described herein may assay thousands of transcripts, data analysis only needs to be performed on the transcript of interest, thereby reducing the problem of multiple comparisons inherent in performing multiple statistical tests.The p-value can also be corrected for multiple comparisons, for example, using Bonferroni correction, its variants, or other techniques known to those skilled in the art, such as Hochberg correction, Holm-Bonferroni correction, Sidak correction, or Dunnett correction.The degree of differential expression can also be taken into consideration. For example, a gene can be considered differentially expressed if the fold change in expression compared to the control level differs by at least 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0, 2.2, 2.5, 2.7, 3.0, 4, 5, 6, 7, 8, 9, or 10 fold in the sample relative to the control. Differential expression considers both overexpression and underexpression.If differential expression meets a statistical threshold, a fold change threshold, or both, a gene or gene product can be considered to be up-regulated or down-regulated.For example, the criteria for identifying differential expression can include both a p-value of 0.001 and a fold change of at least 1.5-fold (up or down).Those skilled in the art will understand that such statistical and threshold measures can be adapted to determine differential expression by any molecular profiling technique disclosed herein.

[0271] Various methods as described herein utilize many types of microarrays to detect the presence and potentially the amount of biological entities in a sample. Arrays typically contain addressable moieties that can detect the presence of an entity in a sample, for example, by a binding event. Microarrays include, but are not limited to, DNA microarrays, such as cDNA microarrays, oligonucleotide microarrays and SNP microarrays, microRNA arrays, protein microarrays, antibody microarrays, tissue microarrays, cell microarrays (also called transfection microarrays), chemical compound microarrays, and carbohydrate arrays (glycoarrays). DNA arrays typically contain addressable nucleotide sequences that can bind to sequences present in a sample. MicroRNA arrays, such as the MMChips array from University of Louisville or commercially available systems from Agilent, can be used to detect microRNA. Protein microarrays can be used to identify protein-protein interactions, including but not limited to identifying substrates of protein kinases, transcription factor protein activation, or to identify targets of biologically active small molecules. Protein arrays may include arrays of different protein molecules, typically antibodies, or nucleotide sequences that bind to a protein of interest. Antibody microarrays contain antibodies spotted on a protein chip that are used as capture molecules to detect proteins or other biological materials from samples, such as cell or tissue lysates. For example, antibody arrays can be used to detect biomarkers from bodily fluids, such as serum or urine, for diagnostic applications. Tissue microarrays contain separate tissue cores assembled in an array format to allow multiplexed tissue analysis. Cell microarrays, also called transfection microarrays, contain a variety of capture agents, such as antibodies, proteins, or lipids, that interact with cells to facilitate capture at addressable locations.A chemical compound microarray comprises an array of chemical compounds and can be used to detect proteins or other biological substances that bind to the compounds. A carbohydrate array (glycoarray) comprises an array of carbohydrates and can detect, for example, proteins that bind to sugar moieties. Those skilled in the art will recognize that similar techniques or improvements can be used in accordance with the methods as described herein.

[0272] Certain embodiments of the method include a multi-well reaction vessel, including but not limited to a multi-well plate or a multi-chamber microfluidic device, in which multiplex amplification reactions and in some embodiments detection are typically performed in parallel. In certain embodiments, one or more multiplex reactions to generate amplicons are performed in the same reaction vessel, including but not limited to a multi-well plate, such as a 96-well, 384-well, 1536-well plate; or a microfluidic device, such as but not limited to a TaqMan™ low-density array (Applied Biosystems, Foster City, CA). In some embodiments, the massively parallel amplification step includes a plate containing multiple reaction wells, such as a multi-well reaction vessel, including but not limited to a 24-well plate, a 96-well plate, a 384-well plate, or a 1536-well plate; or a multi-chamber microfluidic device, such as but not limited to a low-density array, in which each chamber or well contains the appropriate primer, primer set, and / or reporter probe, as appropriate. Typically such amplification steps occur in a series of parallel singleplex, 2-plex, 3-plex, 4-plex, 5-plex, or 6-plex reactions, although higher levels of parallel multiplexing are also within the contemplated scope of the present teachings. These methods can include PCR methodologies, such as RT-PCR, in each of the wells or chambers to amplify and / or detect the nucleic acid molecules of interest.

[0273] Low density arrays can include arrays that detect tens or hundreds of molecules as opposed to thousands of molecules. These arrays can be more sensitive than high density arrays. In some embodiments, low density arrays, such as TaqMan™ low density arrays, are used to detect one or more genes or gene products in any of Tables 5-12 of WO2018175501. For example, low density arrays can be used to detect at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 40, 50, 60, 70, 80, 90, or 100 genes or gene products selected from any of Tables 5-12 of WO2018175501.

[0274] In some embodiments, the disclosed methods include a microfluidic device, a "lab on a chip," or a micro total analysis system (pTAS). In some embodiments, the sample preparation is performed using a microfluidic device. In some embodiments, the amplification reaction is performed using a microfluidic device. In some embodiments, the sequencing or PCR reaction is performed using a microfluidic device. In some embodiments, the nucleotide sequence of at least a portion of the amplification product is obtained using a microfluidic device. In some embodiments, the detecting step includes a microfluidic device, including but not limited to a low density array, such as a TaqMan™ low density array. Descriptions of exemplary microfluidic devices can be found, among others, in published PCT application numbers WO / 0185341 and WO04 / 011666; Kartalov and Quake, Nucl. Acids Res. 32:2873-79, 2004; and Fiorini and Chiu, Bio Techniques 38:429-46, 2005.

[0275] Any suitable microfluidic device can be used in the methods described herein. Examples of microfluidic devices that may be used or adapted for use with molecular profiling include U.S. Patent Nos. 7,591,936, 7,581,429, 7,579,136, 7,575,722, 7,568,399, 7,552,741, 7,544,506, 7,541,578, 7,518,726, 7,488,596, 7,485,214, 7,467,928, and the like. No. 7,452,713, No. 7,452,509, No. 7,449,096, No. 7,431,887, No. 7,422,725, No. 7,422,669, No. 7,419,822, No. 7,419,639, 7,413,709, 7,411,184, 7,402,229, 7,390,463, 7,381,471, 7,357,864, 7,351,592, 7,351,380, No. 7,338,637, No. 7,329,391, No. 7,323,140, ​​No. 7,261,824, No. 7,258,837, No. 7,253,003, No. 7,238,324, No. 7,238,255, No. 7 ,233,865, 7,229,538, 7,201,881, 7,195,986, 7,189,581, 7,189,580, 7,189,368, 7,141,978, 7,1 Nos. 38,062, 7,135,147, 7,125,711, 7,118,910, 7,118,661, 7,640,947, 7,666,361, 7,704,735; U.S. Patent Application Publication No. 20060035243; and International Patent Publication No. WO2010 / 072410, each of which is incorporated by reference herein in its entirety.Another example for use with the methods disclosed herein is described in Chen et al., "Microfluidic isolation and transcriptome analysis of serum vesicles," Lab on a Chip, Dec. 8, 2009 DOI: 10.1039 / b916199f.

[0276] Gene expression analysis by massively parallel signature sequencing (MPSS) This method, described by Brenner et al. (2000) Nature Biotechnology 18:630-634, is a sequencing approach that combines non-gel-based signature sequencing with in vitro cloning of millions of templates on separate microbeads. First, a microbead library of DNA templates is constructed by in vitro cloning. This is followed by high-density assembly of planar arrays of template-containing microbeads in a flow cell. The free ends of the cloned templates on each microbead are simultaneously analyzed using a fluorescence-based signature sequencing method that does not require separation of DNA fragments. This method has been shown to simultaneously and accurately provide hundreds of thousands of gene signature sequences from a cDNA library in a single run.

[0277] MPSS data has many applications. The expression level of almost every transcript can be quantitatively determined; the abundance of the signature represents the expression level of the gene in the analyzed tissue. Quantitative methods for analyzing the frequency of tags and for detecting differences between libraries have been published and are incorporated into the public database for SAGE™ data, and are applicable to MPSS data. The availability of complete genome sequences allows direct comparison of signatures to genome sequences, further broadening the utility of MPSS data. Because targets for MPSS analysis are not preselected (as in microarrays), MPSS data can characterize the full complexity of the transcriptome. This is similar to sequencing millions of ESTs at once, and genome sequence data can be used so that the source of the MPSS signature can be easily identified by computational means.

[0278] Serial Analysis of Gene Expression (SAGE) Serial analysis of gene expression (SAGE) is a method that allows for the simultaneous quantitative analysis of many gene transcripts without the need to provide separate hybridization probes for each transcript. First, short sequence tags (e.g., about 10-14 bp) are generated that contain enough information to uniquely identify a transcript, provided that the tag is derived from a unique location within each transcript. Then, many transcripts are linked together to form long continuous molecules that can be sequenced, revealing the identity of multiple tags simultaneously. The expression pattern of any population of transcripts can be quantitatively assessed by determining the abundance of individual tags and identifying the gene that corresponds to each tag. See, e.g., Velculescu et al. (1995) Science 270:484-487; and Velculescu et al. (1997) Cell 88:243-51.

[0279] DNA copy number profiling Any method that can determine the DNA copy number profile of a particular sample can be used for molecular profiling according to the method described herein, as long as the resolution is sufficient to identify copy number polymorphisms in biomarkers as described herein.Those skilled in the art will recognize and can use several different platforms to examine whole genome copy number changes with sufficient resolution to identify the copy number of one or more biomarkers of the method described herein.Some of the platforms and techniques are described in the following embodiments.In some embodiments as described herein, next generation sequencing or ISH techniques as described herein or known in the art are used to determine copy number / gene amplification.

[0280] In some embodiments, copy number profile analysis involves amplification of whole genomic DNA using whole genome amplification techniques. Whole genome amplification techniques can use strand-displacing polymerases and random primers.

[0281] In some aspects of these embodiments, the copy number profile analysis involves hybridization of the whole genome amplified DNA with a high density array. In more particular aspects, the high density array has 5,000 or more different probes. In another particular aspect, the high density array has 5,000, 10,000, 20,000, 50,000, 100,000, 200,000, 300,000, 400,000, 500,000, 600,000, 700,000, 800,000, 900,000, or 1,000,000 or more different probes. In another particular aspect, each of the different probes on the array is an oligonucleotide having a length of about 15 to 200 bases. In another specific aspect, each of the different probes on the array is an oligonucleotide having a length of about 15-200, 15-150, 15-100, 15-75, 15-60, or 20-55 bases.

[0282] In some embodiments, microarrays are employed to help determine copy number profiles for samples, e.g., cells from tumors. Microarrays typically include a plurality of oligomers (e.g., DNA or RNA polynucleotides or oligonucleotides, or other polymers) synthesized or deposited in an array pattern on a substrate (e.g., a glass support). The oligomers bound to the support are "probes" that function to hybridize or bind to sample material (e.g., nucleic acids prepared or obtained from tumor samples) in hybridization experiments. The reverse situation can also be applied, where the sample can be bound to the microarray substrate and the oligomer probes are in solution for hybridization. In use, the array surface is contacted with one or more targets under conditions that promote specific high affinity binding of the target to one or more of the probes. In some configurations, the sample nucleic acid is labeled with a detectable label, such as a fluorescent tag, so that the hybridized sample and probes are detectable using a scanning instrument. DNA array technology offers the potential to analyze DNA copy number profiles using a large number (e.g., hundreds of thousands) of different oligonucleotides. In some embodiments, the substrate used for the array is a surface-derivatized glass or silica, or a polymer membrane surface (see, e.g., Z. Guo, et al., Nucleic Acids Res, 22, 5456-65 (1994); U. Maskos, EM Southern, Nucleic Acids Res, 20, 1679-84 (1992), and EM Southern, et al., Nucleic Acids Res, 22, 1368-73 (1994), each of which is incorporated herein by reference). Modification of the array substrate surface can be accomplished by a number of techniques. For example, siliceous or metal oxide surfaces can be modified with bifunctional silanes, i.e., a first functional group that allows for covalent bonding to the surface (e.g., --SiCl, respectively). 3 or --Si(OCH 3 ) 3The surface can be derivatized with silanes having a functional group (Si-halogen or Si-alkoxy group as seen in ...

[0283] Polymer array synthesis has also been widely described in the literature, including WO 00 / 58516, U.S. Pat. Nos. 5,143,854, 5,242,974, 5,252,743, 5,324,633, 5,384,261, 5,405,783, 5,424,186, 5,451,683, 5,482,867, 5,491, and the like. ,074, 5,527,681, 5,550,215, 5,571,639, 5,578,832, 5,593,839, 5,599,695, 5, 624,711, 5,631,734, 5,795,716, 5,831,070, 5,837,832, 5,856,101, 5,858,659, Nos. 5,936,324, 5,968,740, 5,974,164, 5,981,185, 5,981,956, 6,025,601, 6,033,860, 6,040,193, 6,090,555, 6,136,269, 6,269,846 and 6,428,752, 5,412,087, 6,14 Nos. 7,205, 6,262,216, 6,310,189, 5,889,165, and 5,959,098, PCT Application Nos. PCT / US99 / 00730 (International Publication No. WO99 / 36760) and PCT / US01 / 04285 (International Publication No. WO01 / 58593), all of which are incorporated by reference in their entirety for all purposes.

[0284] Nucleic acid arrays useful in the present disclosure include, but are not limited to, those commercially available under the trade name GeneChip™ from Affymetrix (Santa Clara, Calif.). Exemplary arrays are shown on the affymetrix.com website. Another microarray supplier is Illumina, Inc. of San Diego, Calif., and exemplary arrays are shown on the illumina.com website.

[0285] In some embodiments, the method of the present invention provides sample preparation.Depending on the microarray and the experiment to be carried out, sample nucleic acid can be prepared in several ways by methods known to those skilled in the art.In some aspects as described herein, before or at the same time as genotyping (analysis of copy number profile), sample can be amplified by several mechanisms.The most common amplification procedure used involves PCR. See, e.g., PCR Technology: Principles and Applications for DNA Amplification (Ed. H. A. Erlich, Freeman Press, NY, NY, 1992); PCR Protocols: A Guide to Methods and Applications (Eds. Innis, et al., Academic Press, San Diego, Calif., 1990); Mattila et al., Nucleic Acids Res. 19, 4967 (1991); Eckert et al., PCR Methods and Applications 1, 17 (1991); PCR (Eds. McPherson et al., IRL Press, Oxford); and U.S. Pat. Nos. 4,683,202, 4,683,195, 4,800,159, 4,965,188, and 5,333,675, each of which is incorporated by reference in its entirety for all purposes. In some embodiments, the sample may be amplified on an array (eg, US Pat. No. 6,300,070, incorporated herein by reference).

[0286] Other suitable amplification methods include ligase chain reaction (LCR) (e.g., Wu and Wallace, Genomics 4, 560 (1989), Landegren et al., Science 241, 1077 (1988) and Barringer et al. Gene 89:117 (1990)), transcription amplification (Kwoh et al., Proc. Natl. Acad. Sci. USA 86, 1173 (1989) and WO88 / 10315), self-sustained sequence replication (Guatelli et al., Proc. Nat. Acad. Sci. USA, 87, 1874), and genomic amplification (Garcia et al., 1999). (1990) and WO 90 / 06995), selective amplification of target polynucleotide sequences (U.S. Pat. No. 6,410,276), consensus sequence primed polymerase chain reaction (CP-PCR) (U.S. Pat. No. 4,437,975), arbitrarily primed polymerase chain reaction (AP-PCR) (U.S. Pat. Nos. 5,413,909, 5,861,245), and nucleic acid based sequence amplification (NABSA) (see U.S. Pat. Nos. 5,409,818, 5,554,517, and 6,063,603, each of which is incorporated herein by reference). Other amplification methods that may be used are described in US Pat. Nos. 5,242,794, 5,494,810, 4,988,617, and US patent application Ser. No. 09 / 854,317, each of which is incorporated herein by reference.

[0287] Additional methods of sample preparation and techniques for reducing the complexity of nucleic acid samples are described in Dong et al., Genome Research 11, 1418 (2001), U.S. Patent Nos. 6,361,947, 6,391,592, and U.S. Patent Application Nos. 09 / 916,135, 09 / 920,491 (U.S. Patent Application Publication No. 20030096235), 09 / 910,292 (U.S. Patent Application Publication No. 20030082543), and 10 / 013,598.

[0288] The method for carrying out polynucleotide hybridization assay has been fully developed in the art.The procedure and conditions of hybridization assay used in the method as described herein vary according to application and are selected according to known general binding methods, including the method mentioned in Maniatis et al. Molecular Cloning: A Laboratory Manual (2nd Ed. Cold Spring Harbor, NY, 1989); Berger and Kimmel Methods in Enzymology, Vol. 152, Guide to Molecular Cloning Techniques (Academic Press, Inc., San Diego, Calif., 1987); Young and Davism, PNAS, 80: 1194 (1983). Methods and apparatus for performing repeated and controlled hybridization reactions are described in U.S. Pat. Nos. 5,871,928, 5,874,219, 6,045,996, and 6,386,749 and 6,391,623, each of which is incorporated herein by reference.

[0289] Methods as described herein may also involve signal detection of hybridization between the ligands after (and / or during) hybridization. See U.S. Patent Nos. 5,143,854, 5,578,832; 5,631,734; 5,834,758; 5,936,324; 5,981,956; 6,025,601; 6,141,096; 6,185,030; 6,201,639; 6,218,803; and 6,225,625, U.S. patent application Ser. No. 10 / 389,194, and PCT application PCT / US99 / 06097 (published as WO99 / 47964), each of which is also incorporated by reference herein in its entirety for all purposes.

[0290] Methods and apparatus for signal detection and intensity data processing are described, for example, in U.S. Patent Nos. 5,143,854; 5,547,839; 5,578,832; 5,631,734; 5,800,992; 5,834,758; 5,856,092; 5,902,723; 5,936,324; 5,981,956; 6,025,601; 6,090,555; Nos. 6,141,096, 6,185,030, 6,201,639; 6,218,803; and 6,225,625, U.S. patent application Ser. Nos. 10 / 389,194, 60 / 493,495, and PCT application PCT / US99 / 06097 (published as WO99 / 47964), each of which is also incorporated by reference herein in its entirety for all purposes.

[0291] Immuno-based assays Protein-based molecular profiling techniques include immunoaffinity assays based on antibodies selectively immunoreactive with the proteins encoded by the mutant genes according to the present method. These techniques include, but are not limited to, immunoprecipitation, Western blot analysis, molecular binding assays, enzyme-linked immunosorbent assays (ELISA), enzyme-linked immunofiltration assays (ELIFA), fluorescence-activated cell sorting (FACS), and the like. For example, any method of detecting the expression of a biomarker in a sample includes contacting the sample with an antibody against the biomarker, or an immunoreactive fragment of the antibody, or a recombinant protein containing the antigen-binding region of the antibody against the biomarker; and then detecting the binding of the biomarker in the sample. Methods for producing such antibodies are known in the art. Antibodies can be used to immunoprecipitate specific proteins from solution samples, or to immunoblot proteins separated, for example, by polyacrylamide gels. Immunocytochemistry methods can also be used to detect specific protein polymorphisms in tissues or cells. Other well-known antibody-based techniques can also be used, including, for example, ELISA, radioimmunoassays (RIA), immunoradiometric assays (IRMA) and immunoenzymatic assays (IEMA), including sandwich assays using monoclonal or polyclonal antibodies. See, e.g., U.S. Patent Nos. 4,376,110 and 4,486,530, both of which are incorporated herein by reference.

[0292] In an alternative method, the sample may be contacted with an antibody specific for the biomarker under conditions sufficient for the formation of an antibody-biomarker complex, which may then be detected. The presence of the biomarker may be detected in several ways, for example, by Western blotting and ELISA procedures for assaying a wide variety of tissues and samples, including plasma or serum. A wide range of immunoassay techniques using such assay formats are available. See, for example, U.S. Patent Nos. 4,016,043, 4,424,279, and 4,018,653. These include both single-site and two-site or "sandwich" assays of the non-competitive type, as well as traditional competitive binding assays. These assays also include direct binding of labeled antibodies to target biomarkers.

[0293] There are several variations of the sandwich assay technique, all of which are intended to be encompassed by the present method. Briefly, in a typical forward assay, unlabeled antibody is immobilized on a solid substrate, and the test sample is contacted with the bound molecule. After a suitable incubation period sufficient to allow antibody-antigen complexes to form, a second antibody specific to the antigen, labeled with a reporter molecule capable of producing a detectable signal, is then added and incubated, allowing sufficient time for the formation of another complex of antibody-antigen-labeled antibody. Any unreacted material is washed away, and the presence of the antigen is determined by observing the signal produced by the reporter molecule. The results can be either qualitative, by simple observation of the visible signal, or quantified by comparison with a control sample containing known amounts of biomarker.

[0294] Variations of the forward assay include simultaneous assays, in which both the sample and the labeled antibody are added simultaneously to the bound antibody. These techniques are well known to those skilled in the art, including any minor variations that will be readily apparent. In a typical forward sandwich assay, a first antibody with specificity for a biomarker is either covalently or passively bound to a solid surface. The solid surface is typically glass or a polymer, the most commonly used polymers being cellulose, polyacrylamide, nylon, polystyrene, polyvinyl chloride or polypropylene. The solid support can be in the form of a tube, a bead, a disk of a microplate, or any other surface suitable for performing an immunoassay. The binding process is well known in the art and generally consists of a cross-linking, covalent binding or physical adsorption step, and the polymer-antibody complex is washed in preparation for the test sample. An aliquot of the test sample is then added to the solid phase complex and incubated under appropriate conditions (e.g., room temperature to 40°C, e.g., between 25°C and 32°C, inclusive) for a period of time sufficient to bind any subunits present in the antibody (e.g., 2-40 minutes or, more conveniently, overnight). Following the incubation period, the antibody subunit solid phase is washed, dried, and incubated with a second antibody specific for a portion of the biomarker. The second antibody is linked to a reporter molecule that is used to indicate binding of the second antibody to the molecular marker.

[0295] An alternative method involves immobilizing the target biomarker in the sample and then exposing the immobilized target to a specific antibody that may or may not be labeled with a reporter molecule. Depending on the amount of target and the signal strength of the reporter molecule, the bound target may be detectable by direct labeling with the antibody. Alternatively, a second labeled antibody specific to the first antibody is exposed to the target-first antibody complex to form a ternary complex of target-first antibody-second antibody. This complex is detected by the signal emitted by the reporter molecule. By "reporter molecule" as used herein is meant a molecule whose chemical nature provides an analytically identifiable signal that allows the antibody bound to the antigen to be detected. The reporter molecules most commonly used in this type of assay are either enzymes, fluorophores or radionuclide-containing molecules (i.e., radioisotopes) and chemiluminescent molecules.

[0296] In the case of enzyme immunoassays, the enzyme is conjugated to the second antibody, generally by glutaraldehyde or periodate. However, as will be readily recognized, there are a wide variety of different conjugation techniques readily available to those skilled in the art. Commonly used enzymes include horseradish peroxidase, glucose oxidase, β-galactosidase and alkaline phosphatase, among others. The substrate to be used with the specific enzyme is generally chosen for the production of a detectable color change upon hydrolysis by the corresponding enzyme. Examples of suitable enzymes include alkaline phosphatase and peroxidase. It is also possible to employ fluorogenic substrates that yield fluorescent products rather than the chromogenic substrates mentioned above. In any case, the enzyme-labeled antibody is added to the first antibody-molecular marker complex, allowed to bind, and then excess reagent is washed away. A solution containing the appropriate substrate is then added to the antibody-antigen-antibody complex. The substrate reacts with the enzyme linked to the second antibody to give a qualitative visible signal, which may then be quantified, typically spectrophotometrically, to provide an indication of the amount of biomarker present in the sample. Alternatively, fluorescent compounds such as fluorescein and rhodamine may be chemically coupled to the antibody without altering the antibody's binding ability. When activated by illumination with light of a specific wavelength, the fluorochrome-labeled antibody absorbs the light energy, inducing an excited state in the molecule, which subsequently emits a characteristic color of light that is visually detectable by light microscopy. As in EIA, the fluorescent-labeled antibody is allowed to bind to the first antibody-molecular marker complex. After washing away unbound reagents, the remaining ternary complex is then exposed to light of an appropriate wavelength, and the observed fluorescence indicates the presence of the molecular marker of interest. Both immunofluorescence and EIA techniques are very well established in the art. However, other reporter molecules such as radioisotopes, chemiluminescent or bioluminescent molecules may also be employed.

[0297] Immunohistochemistry (IHC) IHC is a process of locating an antigen (e.g., a protein) in cells of a tissue using an antibody that specifically binds to the antigen in the tissue. The antigen-binding antibody can be conjugated or fused to a tag that allows its detection, for example, by visualization. In some embodiments, the tag is an enzyme, such as alkaline phosphatase or horseradish peroxidase, that can catalyze a color reaction. The enzyme can be fused to the antibody or non-covalently bound, for example, using the biotin-avidin system. Alternatively, the antibody can be tagged with a fluorophore, such as fluorescein, rhodamine, DyLight Fluor, or Alexa Fluor. The antigen-binding antibody can be directly tagged, or a detection antibody carrying a tag can recognize the antigen-binding antibody itself. IHC may be used to detect one or more proteins. The expression of a gene product can be related to its staining intensity compared to a control level. In some embodiments, a gene product is considered to be differentially expressed if its staining varies by at least 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0, 2.2, 2.5, 2.7, 3.0, 4, 5, 6, 7, 8, 9, or 10 fold in a sample compared to a control.

[0298] IHC involves the application of antigen-antibody interactions to histochemical techniques. In an illustrative example, tissue sections are mounted on slides and incubated with antibodies (polyclonal or monoclonal) specific for the antigen (primary reaction). The antigen-antibody signal is then amplified using a second antibody conjugated to a peroxidase antiperoxidase (PAP), avidin-biotin-peroxidase (ABC) or avidin-biotin alkaline phosphatase complex. In the presence of a substrate and a chromogen, the enzyme forms a colored deposit at the antibody-antigen binding site. Immunofluorescence is an alternative technique for visualizing antigens. In this technique, the primary antigen-antibody signal is amplified using a second antibody conjugated to a fluorescent dye. When UV light is absorbed, the fluorescent dye itself emits light of a longer wavelength (fluorescence), thus allowing the location of the antibody-antigen complex to be identified.

[0299] Epigenetic state The molecular profiling method according to the present disclosure also includes measuring epigenetic changes, i.e., gene modifications caused by epigenetic mechanisms, such as changes in methylation status or histone acetylation. Frequently, epigenetic changes result in alterations in the expression levels of genes that can be detected (at the RNA or protein level, as appropriate) as indicators of epigenetic changes. Often, epigenetic changes result in silencing or downregulation of genes, referred to as "epigenetic silencing". The epigenetic changes most frequently investigated in the methods described herein involve determining the DNA methylation status of genes, where increased methylation levels are typically associated with associated cancers (as it can cause downregulation of gene expression). Aberrant methylation, sometimes referred to as hypermethylation, of one or more genes can be detected. Typically, the methylation status is determined in appropriate CpG islands, which are often found in the promoter regions of genes. The terms "methylation," "methylation status," or "methylation state" may refer to the presence or absence of 5-methylcytosine at one or more CpG dinucleotides within a DNA sequence. CpG dinucleotides are typically enriched in promoter regions and exons of human genes.

[0300] Reduced gene expression can be examined by DNA methylation status or expression level when determined by the methylation status of gene.One method for detecting epigenetic silencing is to determine that the gene expres...

Claims

1. 1. A system for selecting a treatment for cancer in a first subject, comprising: one or more computers; one or more memory devices storing instructions that, when executed by the one or more computers, cause the one or more computers to perform operations; and wherein the operation comprises: and obtaining, by the one or more computers, a plurality of copy numbers based on output data generated by a next generation sequencer, wherein the next generation sequencer generates the output data by sequencing a biological sample comprising cancer cells from a first subject, and the plurality of copy numbers are obtained by determining whether or not a plurality of copy numbers correspond to a plurality of genes in the following groups or genomic regions adjacent thereto: (a) Group 1, including multiple of MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; (b) group 2, including multiple of BCL9, PBX1, PRRX1, INHBA, and YWHAE; (c) group 3, including multiple of BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; (d) group 4, including multiples of BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CREB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR; and (e) Group 5, including multiple of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11. comprising the copy number for each of the providing, by the one or more computers, input data as input to a predictive model, the input data comprising a plurality of obtained copy numbers, the predictive model comprising a plurality of machine learning models, each machine learning model configured to process the obtained copy numbers for one of group 1, group 2, group 3, group 4, and group 5; processing, by the one or more computers, input data including the resulting copy numbers for each group through one of the plurality of machine learning models, wherein the processing includes: processing, by the one or more computers, the copy numbers obtained for group 1 through a first machine learning model of the plurality of machine learning models to generate first data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the first machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; processing, by the one or more computers, the copy numbers obtained for group 2 through a second machine learning model of the plurality of machine learning models to generate second data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the second machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including BCL9, PBX1, PRRX1, INHBA, and YWHAE; processing, by the one or more computers, the copy numbers obtained for group 3 through a third machine learning model of the plurality of machine learning models to generate third data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the third machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; and processing, by the one or more computers, the copy numbers obtained for group 4 through a fourth machine learning model of the plurality of machine learning models to generate fourth data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the fourth machine learning model is selected from the group consisting of BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, trained to generate output data indicative of the likelihood of benefiting from treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CREB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR; and processing, by the one or more computers, the copy numbers obtained for group 5 through a fifth machine learning model of the plurality of machine learning models to generate fifth data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the fifth machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11. including; determining, by the one or more computers, whether a majority of the plurality of machine learning models indicates that the first subject is likely to benefit from a treatment that includes platinum therapy based on the first data, the second data, the third data, the fourth data, and the fifth data; identifying, by the one or more computers, data identifying a treatment that includes a platinum therapy based on a determination that a majority of the plurality of machine learning models indicates that the first subject is likely to benefit from a treatment that includes a platinum therapy; and providing, by the one or more computers, an output identifying a treatment comprising a platinum therapy. Including, the system.

2. 2. The system of claim 1, wherein the platinum therapy comprises one or more of cisplatin, carboplatin, oxaliplatin, or nedaplatin.

3. 2. The system of claim 1, wherein the platinum therapy comprises a combination therapy comprising one or more of cisplatin, carboplatin, oxaliplatin, or nedaplatin.

4. Cancers include acute myeloid leukemia (AML), breast cancer, cholangiocarcinoma, colorectal adenocarcinoma, extrahepatic bile duct adenocarcinoma, female genital malignancies, gastric adenocarcinoma, gastroesophageal adenocarcinoma, gastrointestinal stromal tumor (GIST), glioblastoma, head and neck squamous cell carcinoma, leukemia, hepatocellular carcinoma, low-grade glioma, bronchioloalveolar carcinoma (BAC), non-small cell lung cancer (NSCLC), small cell lung cancer (SCLC), lymphoma, male genital malignancies, and isolated malignant pleural The system of claim 1, further comprising a tumor-specific fibrous tumor (MSFT), melanoma, multiple myeloma, neuroendocrine tumor, nodular diffuse large B-cell lymphoma, non-epithelial ovarian cancer (non-EOC), ovarian surface epithelial carcinoma, pancreatic adenocarcinoma, pituitary carcinoma, oligodendroglioma, prostate adenocarcinoma, retroperitoneal or peritoneal carcinoma, retroperitoneal or peritoneal sarcoma, small intestine malignant tumor, soft tissue tumor, thymic carcinoma, thyroid carcinoma, or uveal melanoma.

5. 2. The system of claim 1, wherein the cancer comprises colorectal cancer, ovarian cancer, esophageal cancer, gastroesophageal junction cancer, gastric cancer, head and neck cancer, bladder cancer, breast cancer, endometrial cancer, uterine cancer, cervical cancer, pancreatic cancer, or lung cancer.

6. The system of claim 1, wherein the biological sample comprises formalin-fixed paraffin-embedded (FFPE) tissue, fixed tissue, core needle biopsy, fine needle aspirate, unstained slide, fresh frozen (FF) tissue, formalin sample, tissue contained in a solution that preserves nucleic acid or protein molecules, fresh sample, malignant fluid, body fluid, tumor sample, tissue sample, or any combination thereof.

7. 10. The system of claim 1, wherein the biological sample comprises cells from a solid tumor.

8. 10. The system of claim 1, wherein the biological sample comprises a bodily fluid.

9. 9. The system of claim 8, wherein the bodily fluid comprises malignant fluid, pleural fluid, or peritoneal fluid.

10. 9. The system of claim 8, wherein the bodily fluid comprises peripheral blood, serum, plasma, ascites, urine, cerebrospinal fluid (CSF), sputum, saliva, bone marrow, synovial fluid, aqueous humor, amniotic fluid, earwax, breast milk, bronchoalveolar lavage fluid, semen, prostatic fluid, Cowper's fluid, pre-ejaculate fluid, female ejaculate, sweat, feces, tears, cyst fluid, pleural fluid, peritoneal fluid, pericardial fluid, lymph, chyme, chyle, bile, interstitial fluid, menstrual secretions, pus, sebum, vomit, vaginal fluid, mucosal secretions, stool water, pancreatic juice, nasal lavage fluid, bronchopulmonary aspirate, blastocytic cavity fluid, or umbilical cord blood.

11. 13. The system of claim 1, wherein the next generation sequencer is configured to sequence a panel of genes.

12. 2. The system of claim 1, wherein the next generation sequencer is configured to perform whole exome sequencing.

13. 2. The system of claim 1, wherein the first data, the second data, the third data, the fourth data, and the fifth data each represent an equivalent vote for each machine learning model.

14. 2. The system of claim 1, wherein the first data, the second data, the third data, the fourth data, and the fifth data each represent a vote of each machine learning model adjusted based on a confidence score associated with the respective machine learning model.

15. 2. The system of claim 1, wherein one or more of the plurality of machine learning models is a random forest model.

16. The operation, obtaining, by the one or more computers, a second biological sample from a second subject, the second biological sample comprising cancer cells; and obtaining, by the one or more computers, a plurality of second copy numbers based on second output data generated by a next generation sequencer, wherein the next generation sequencer generates the second output data by sequencing a second biological sample comprising cancer cells from a second subject, and the plurality of second copy numbers are obtained by determining whether the second copy numbers are related to genes or genomic regions adjacent thereto in the following groups: (f) group 1, including multiple of MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; (g) group 2, including one, two, three, four, or all five of BCL9, PBX1, PRRX1, INHBA, and YWHAE; (h) group 3, including multiple of BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; (i) group 4, which includes multiples of BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CREB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR; and (j) Group 5, including multiple of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11. comprising the copy number for each of the providing, by the one or more computers, second input data as input to the predictive model, the second input data comprising the obtained plurality of second copy numbers; processing, by the one or more computers, second input data including second copy numbers obtained for each group through one of a plurality of machine learning models, the processing including: processing, by the one or more computers, the second copy number obtained for group 1 through a first machine learning model of the plurality of machine learning models to generate sixth data indicative of whether a second subject is likely to benefit from a therapy that includes a platinum therapy; processing, by the one or more computers, the second copy number obtained for group 2 through a second machine learning model of the plurality of machine learning models to generate seventh data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy; processing, by the one or more computers, the second copy number obtained for group 3 through a third machine learning model of the plurality of machine learning models to generate eighth data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy; processing, by the one or more computers, the second copy number obtained for group 4 through a fourth machine learning model of the plurality of machine learning models to generate ninth data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy; and processing, by the one or more computers, the second copy number obtained for group 5 through a fifth machine learning model of the plurality of machine learning models to generate tenth data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy. including; determining, by the one or more computers, whether a majority of the plurality of machine learning models indicates that the second subject is likely to benefit from a treatment that includes platinum therapy based on the sixth data, the seventh data, the eighth data, the ninth data, and the tenth data; identifying, by the one or more computers, data that identifies a treatment alternative to the platinum therapy-containing treatment based on a determination that a majority of the plurality of machine learning models indicates that the second subject is unlikely to benefit from the platinum therapy-containing treatment; and providing, by the one or more computers, output data identifying the alternative treatment.

13. The system of claim 1, further comprising:

17. 17. The system of claim 16, wherein the alternative treatment is a treatment that does not include platinum therapy.

18. 1. A method for selecting a treatment for cancer in a first subject, comprising: Obtaining, by one or more computers, a plurality of copy numbers based on output data generated by a next generation sequencer, wherein the next generation sequencer generates the output data by sequencing a biological sample comprising cancer cells from a first subject, and the plurality of copy numbers are obtained by sequencing a plurality of copies of genes or genomic regions adjacent thereto of the following groups: (a) Group 1, including multiple of MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; (b) group 2, including multiple of BCL9, PBX1, PRRX1, INHBA, and YWHAE; (c) group 3, including multiple of BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; (d) group 4, including multiples of BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CREB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR; and (e) Group 5, including multiple of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11. including copy numbers for each of the following: providing, by the one or more computers, input data as input to a predictive model, the input data comprising a plurality of obtained copy numbers, the predictive model comprising a plurality of machine learning models, each machine learning model configured to process the obtained copy numbers for one of group 1, group 2, group 3, group 4, and group 5; processing, by the one or more computers, input data comprising the resulting copy numbers for each group through one of the plurality of machine learning models, the processing comprising: processing, by the one or more computers, the copy numbers obtained for group 1 through a first machine learning model of the plurality of machine learning models to generate first data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the first machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; processing, by the one or more computers, the copy numbers obtained for group 2 through a second machine learning model of the plurality of machine learning models to generate second data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the second machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including BCL9, PBX1, PRRX1, INHBA, and YWHAE; processing, by the one or more computers, the copy numbers obtained for group 3 through a third machine learning model of the plurality of machine learning models to generate third data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the third machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; and processing, by the one or more computers, the copy numbers obtained for group 4 through a fourth machine learning model of the plurality of machine learning models to generate fourth data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the fourth machine learning model is selected from the group consisting of BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, trained to generate output data indicative of the likelihood of benefiting from treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CREB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR; and and processing, by the one or more computers, the copy numbers obtained for group 5 through a fifth machine learning model of the plurality of machine learning models to generate fifth data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, the fifth machine learning model being trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11. a process comprising: determining, by the one or more computers, whether a majority of the plurality of machine learning models indicates that the first subject is likely to benefit from a treatment that includes platinum therapy based on the first data, the second data, the third data, the fourth data, and the fifth data; identifying, by the one or more computers, data identifying a treatment that includes a platinum therapy based on a determination that a majority of the plurality of machine learning models indicates that the first subject is likely to benefit from a treatment that includes a platinum therapy; and providing, by the one or more computers, an output identifying a treatment comprising a platinum therapy. A method comprising:

19. 20. The method of claim 18, wherein the platinum therapy comprises one or more of cisplatin, carboplatin, oxaliplatin, or nedaplatin.

20. 20. The method of claim 18, wherein the platinum therapy comprises a combination therapy comprising one or more of cisplatin, carboplatin, oxaliplatin, or nedaplatin.

21. Cancers include acute myeloid leukemia (AML), breast cancer, cholangiocarcinoma, colorectal adenocarcinoma, extrahepatic bile duct adenocarcinoma, female genital malignancies, gastric adenocarcinoma, gastroesophageal adenocarcinoma, gastrointestinal stromal tumor (GIST), glioblastoma, head and neck squamous cell carcinoma, leukemia, hepatocellular carcinoma, low-grade glioma, bronchioloalveolar carcinoma (BAC), non-small cell lung cancer (NSCLC), small cell lung cancer (SCLC), lymphoma, male genital malignancies, and malignant pleural orbital carcinoma.

20. The method of claim 18, comprising the cancer of any of the following: myeloma, myeloma, ovarian cancer, ovarian surface epithelial carcinoma, pancreatic adenocarcinoma, pituitary carcinoma, oligodendroglioma, prostate adenocarcinoma, retroperitoneal or peritoneal carcinoma, retroperitoneal or peritoneal sarcoma, small intestine malignant tumor, soft tissue tumor, thymic carcinoma, thyroid carcinoma, or uveal melanoma.

22. 20. The method of claim 18, wherein the cancer comprises colorectal cancer, ovarian cancer, esophageal cancer, gastroesophageal junction cancer, gastric cancer, head and neck cancer, bladder cancer, breast cancer, endometrial cancer, uterine cancer, cervical cancer, pancreatic cancer, or lung cancer.

23. 20. The method of claim 18, wherein one or more of the plurality of machine learning models is a random forest model.

24. obtaining, by the one or more computers, a second biological sample from a second subject, the second biological sample comprising cancer cells; Obtaining, by the one or more computers, a plurality of second copy numbers based on second output data generated by a next generation sequencer, wherein the next generation sequencer generates second output data by sequencing a second biological sample comprising cancer cells from a second subject, and the plurality of second copy numbers are obtained by sequencing the following groups of genes or genomic regions adjacent thereto: (f) group 1, including multiple of MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; (g) group 2, including one, two, three, four, or all five of BCL9, PBX1, PRRX1, INHBA, and YWHAE; (h) group 3, including multiple of BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; (i) group 4, which includes multiples of BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CREB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR; and (j) Group 5, including multiple of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11. including copy numbers for each of the following: providing, by the one or more computers, second input data as input to the predictive model, the second input data comprising the obtained plurality of second copy numbers; processing, by the one or more computers, second input data comprising the second copy numbers obtained for each group through one of a plurality of machine learning models, the processing comprising: processing, by the one or more computers, the second copy number obtained for group 1 through a first machine learning model of the plurality of machine learning models to generate sixth data indicative of whether a second subject is likely to benefit from a treatment that includes platinum therapy; processing, by the one or more computers, the second copy number obtained for group 2 through a second machine learning model of the plurality of machine learning models to generate seventh data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy; processing, by the one or more computers, the second copy number obtained for group 3 through a third machine learning model of the plurality of machine learning models to generate eighth data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy; processing, by the one or more computers, the second copy number obtained for group 4 through a fourth machine learning model of the plurality of machine learning models to generate ninth data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy; and processing, by the one or more computers, the second copy number obtained for group 5 through a fifth machine learning model of the plurality of machine learning models to generate tenth data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy. a process comprising: determining, by the one or more computers, whether a majority of the plurality of machine learning models indicates that the second subject is likely to benefit from a treatment that includes platinum therapy based on the sixth data, the seventh data, the eighth data, the ninth data, and the tenth data; based on a determination that a majority of the plurality of machine learning models indicates that the second subject is unlikely to benefit from a treatment that includes a platinum therapy, identifying, by the one or more computers, data that identifies an alternative treatment to the treatment that includes a platinum therapy; and providing, by the one or more computers, output data identifying the alternative treatment.

20. The method of claim 18, further comprising:

25. 25. The method of claim 24, wherein the alternative treatment is a treatment that does not include platinum therapy.

26. One or more non-transitory computer-readable storage media storing instructions that, when executed by one or more computers, cause the one or more computers to perform operations for selecting a treatment for cancer in a first subject, The operation is and obtaining, by the one or more computers, a plurality of copy numbers based on output data generated by a next generation sequencer, wherein the next generation sequencer generates the output data by sequencing a biological sample comprising cancer cells from a first subject, and the plurality of copy numbers are obtained by determining whether or not a plurality of copy numbers correspond to a plurality of genes in the following groups or genomic regions adjacent thereto: (a) Group 1, including multiple of MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; (b) group 2, including multiple of BCL9, PBX1, PRRX1, INHBA, and YWHAE; (c) group 3, including multiple of BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; (d) group 4, including multiples of BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CREB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR; and (e) Group 5, including multiple of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11. comprising the copy number for each of the providing, by the one or more computers, input data as input to a predictive model, the input data comprising a plurality of obtained copy numbers, the predictive model comprising a plurality of machine learning models, each machine learning model configured to process the obtained copy numbers for one of group 1, group 2, group 3, group 4, and group 5; processing, by the one or more computers, input data including the resulting copy numbers for each group through one of the plurality of machine learning models, wherein the processing includes: processing, by the one or more computers, the copy numbers obtained for group 1 through a first machine learning model of the plurality of machine learning models to generate first data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the first machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; processing, by the one or more computers, the copy numbers obtained for group 2 through a second machine learning model of the plurality of machine learning models to generate second data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the second machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including BCL9, PBX1, PRRX1, INHBA, and YWHAE; processing, by the one or more computers, the copy numbers obtained for group 3 through a third machine learning model of the plurality of machine learning models to generate third data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the third machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; processing, by the one or more computers, the copy numbers obtained for group 4 through a fourth machine learning model of the plurality of machine learning models to generate fourth data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the fourth machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, and TCL1A; and processing, by the one or more computers, the copy numbers obtained for group 5 through a fifth machine learning model of the plurality of machine learning models to generate fifth data indicative of whether the first subject is likely to benefit from a treatment that includes platinum therapy, wherein the fifth machine learning model has been trained to generate output data indicative of the likelihood of benefiting from a treatment that includes platinum therapy based on processing input data including copy numbers for a group of genes including BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11. including; determining, by the one or more computers, whether a majority of the plurality of machine learning models indicates that the first subject is likely to benefit from a treatment that includes platinum therapy based on the first data, the second data, the third data, the fourth data, and the fifth data; identifying, by the one or more computers, data identifying a treatment that includes a platinum therapy based on a determination that a majority of the plurality of machine learning models indicates that the first subject is likely to benefit from a treatment that includes a platinum therapy; and providing, by the one or more computers, an output identifying a treatment comprising a platinum therapy. Including, One or more non-transitory computer-readable storage media.

27. 27. The computer-readable storage medium of claim 26, wherein one or more of the plurality of machine learning models is a random forest model.

28. 27. The computer-readable storage medium of claim 26, wherein each machine learning model of the plurality of machine learning models is a random forest model.

29. The operation, obtaining, by the one or more computers, a second biological sample from a second subject, the second biological sample comprising cancer cells; and obtaining, by the one or more computers, a plurality of second copy numbers based on second output data generated by a next generation sequencer, wherein the next generation sequencer generates the second output data by sequencing a second biological sample comprising cancer cells from a second subject, and the plurality of second copy numbers are obtained by determining whether the second copy numbers are related to genes or genomic regions adjacent thereto in the following groups: (f) group 1, including multiple of MYC, EP300, U2AF1, ASXL1, MAML2, CNTRL, WRN, and CDX2; (g) group 2, including one, two, three, four, or all five of BCL9, PBX1, PRRX1, INHBA, and YWHAE; (h) group 3, including multiple of BCL9, PBX1, GNAS, LHFPL6, CASP8, ASXL1, FH, CRKL, MLF1, TRRAP, AKT3, ACKR3, MSI2, PCM1, and MNX1; (i) group 4, which includes multiples of BX1, GNAS, AURKA, CASP8, ASXL1, CRKL, MLF1, GAS7, MN1, SOX10, TCL1A, LMO1, BRD3, SMARCA4, PER1, PAX7, SBDS, SEPT5, PDGFB, AKT2, TERT, KEAP1, ETV6, TOP1, TLX3, COX6C, NFIB, ARFRP1, ARID1A, MAP2K4, NFKBIA, WWTR1, ZNF217, IL2, NSD3, CREB1, BRIP1, SDC4, EWSR1, FLT3, FLT1, FAS, CCNE1, RUNX1T1, and EZR; and (j) Group 5, including multiple of BCL9, PBX1, PRRX1, INHBA, YWHAE, GNAS, LHFPL6, FCRL4, BIRC3, AURKA, and HOXA11. comprising the copy number for each of the providing, by the one or more computers, second input data as input to the predictive model, the second input data comprising the obtained plurality of second copy numbers; processing, by the one or more computers, second input data including second copy numbers obtained for each group through one of a plurality of machine learning models, the processing including: processing, by the one or more computers, the second copy number obtained for group 1 through a first machine learning model of the plurality of machine learning models to generate sixth data indicative of whether a second subject is likely to benefit from a treatment that includes platinum therapy; processing, by the one or more computers, the second copy number obtained for group 2 through a second machine learning model of the plurality of machine learning models to generate seventh data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy; processing, by the one or more computers, the second copy number obtained for group 3 through a third machine learning model of the plurality of machine learning models to generate eighth data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy; processing, by the one or more computers, the second copy number obtained for group 4 through a fourth machine learning model of the plurality of machine learning models to generate ninth data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy; and processing, by the one or more computers, the second copy number obtained for group 5 through a fifth machine learning model of the plurality of machine learning models to generate tenth data indicative of whether the second subject is likely to benefit from a treatment that includes platinum therapy. including; determining, by the one or more computers, whether a majority of the plurality of machine learning models indicates that the second subject is likely to benefit from a treatment that includes platinum therapy based on the sixth data, the seventh data, the eighth data, the ninth data, and the tenth data; identifying, by the one or more computers, data that identifies a treatment alternative to the platinum therapy-containing treatment based on a determination that a majority of the plurality of machine learning models indicates that the second subject is unlikely to benefit from the platinum therapy-containing treatment; and providing, by the one or more computers, output data identifying the alternative treatment.

27. The computer-readable storage medium of claim 26, further comprising:

30. 30. The computer-readable storage medium of claim 29, wherein the alternative treatment is a treatment that does not include platinum therapy.

Images