Targeted Cancer Therapy

Abstract

The present disclosure provides a method of classifying cancer and cancer patients using a classifier, TME Panel 1, which stratifies patients and cancer according to tumor microenvironment. Thus, medical decisions are guided by the presence / absence of certain TME phenotype classes. Also provided is a method of treating a subject, e.g., a human subject, suffering from gastric cancer, breast cancer, prostate cancer, liver cancer, head and neck cancer, melanoma, colorectal cancer, or ovarian cancer, comprising administering a specific therapy according to the classification of the TME of the cancer by the TME Panel 1 classifier. Also provided is a personalized treatment that can be administered to a patient according to the TME Panel 1 classification of a specific type of cancer, e.g., left or right colorectal cancer or dMMR colorectal cancer.

Description

[Technical Field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This international application claims priority to U.S. Provisional Application No. 63 / 166,167, filed March 25, 2021, and U.S. Provisional Application No. 63 / 188,321, filed May 13, 2021, both of which are incorporated herein by reference in their entireties.

[0002] Reference to electronic application sequence listing The contents of the electronic submission sequence listing (file name: 4488_024PC01_Seqlisting_ST25.txt; size: 20,480 bytes; and creation date: March 11, 2022) are incorporated herein by reference in their entirety.

[0003] FIELD OF THE INVENTION The present disclosure relates to methods for stratifying cancer patients suffering from colorectal cancer, gastric cancer, breast cancer, prostate cancer, liver cancer, head and neck cancer, melanoma, ovarian cancer, glioma, glioblastoma, or lung cancer based on diagnostic panels using gene expression data, classifying patients based on the predominant tumor microenvironment biology, methods for identifying subpopulations of cancer patients for treatment with specific therapies, and individualized therapies for treating patients with specific biology of the tumor microenvironment. [Background technology]

[0004] Few diagnostic tools exist to match individuals with cancer to treatment regimens that will optimally alter survival, and for the majority of patients, clinicians must choose therapy without the benefit of accurate tools that are supposed to indicate the best course of treatment.

[0005] In 2015, a collaborative study across six research groups resulted in the Consensus Molecular Subtype (CMS) model for classifying CRC patients (Guinney et al. 2015 Nat. Med. 21:1350-1356). The CMS model represents a synthesis of six different classification schemes, returning four distinct subtypes based on gene expression data returns: CMS groups 1–4 (along with some unclassifiable patients). These four groups were further annotated based on analysis of additional molecular features: briefly, CMS1 is immunogenic and includes MSI-H; CMS2 is WNT and MYC activity; CMS3 includes KRAS mutations and metabolic dysregulation; and CMS4 is stromal or angiogenic in nature. CMS subgroups generally correspond to known pathological features of the disease (Baran et al. 2018 Gastroenterol. Res. 11:264-273) and are prognostic for overall survival (OS) and progression-free survival (PFS) (Lenz et al. 2019 J. Clin. Oncol. 37:1876-1885). Nevertheless, CMS has not proven to be a predictor of response to targeted therapies such as bevacizumab, and there have been some disappointing results between different trials (e.g., CALBG / SWOG 80405 vs. FIRE-3) (Lenz et al. 2019 J. Clin. Oncol. 37:1876-1885; Stintzing et al. 2019 Ann. Oncol. 30:1796-1803). Therefore, patients and clinicians still need predictive diagnostic tools to guide precise treatment of colorectal, gastric, breast, prostate, liver, head and neck, melanoma, or ovarian cancer. Summary of the Invention

[0006] The present disclosure provides a method of treating a human subject suffering from cancer, the method comprising administering a TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is determined prior to administration by applying an artificial neural network (ANN) classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample obtained from the subject, and the cancer tumor is assigned to a TME phenotype class selected from the group consisting of IS (immunosuppressed), A (angiogenic), IA (immune active), ID (immune desert), and combinations thereof.

[0007] Also provided is a method of treating a human subject suffering from cancer, the method comprising: (i) applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel derived from a cancer tumor sample obtained from the subject, wherein the cancer tumor is assigned to a TME phenotype class selected from the group consisting of IS, A, IA, ID, and combinations thereof; and administering a TME phenotype class-specific therapy to the subject.

[0008] Also provided is a method for identifying a human subject suffering from cancer suitable for treatment with a TME phenotype class-specific therapy, the method comprising applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample obtained from the subject, and assigning the cancer tumor to a TME phenotype class selected from the group consisting of IS, A, IA, ID, and combinations thereof, wherein the assigned TME phenotype class indicates that a TME phenotype class-specific therapy can be administered to treat the cancer. In some embodiments, the ANN classifier comprises an input layer, a hidden layer, and an output layer. In some embodiments, the input layer comprises between 2 and 100 nodes.

[0009] In some embodiments, each node in the input layer corresponds to a gene in a gene panel selected from the genes shown in Tables 1 and 2, and the gene panel includes: (i) 1 to 63 genes selected from Table 1 and 1 to 61 genes selected from Table 2; (ii) a gene panel including genes selected from Tables 3 and 4; (iii) a gene panel in Table 5; or (iv) any of the gene panels (gene sets) disclosed in Figures 9A-G. In some embodiments, the sample includes intratumoral tissue. In some embodiments, the RNA expression level is a transcript RNA expression level determined using next-generation sequencing (NGS), such as RNA-Seq, EdgeSeq, PCR, NanoString, WES, or a combination thereof. In some embodiments, the hidden layer includes two nodes, and the output layer includes four output nodes, and any one of the four output nodes in the output layer corresponds to a TME phenotype class, and the four TME phenotype classes are IA, IS, ID, and A. In some embodiments, the method further comprises applying a logistic regression classifier comprising a softmax function to the output of the ANN, wherein the softmax function assigns a probability to each TME phenotype class. In some embodiments, the TME phenotype class-specific therapy is an IA, IS, ID, or A TME phenotype class-specific therapy, or a combination thereof. In some embodiments, the TME phenotype class-specific therapy is an IA TME phenotype class-specific therapy, including a checkpoint modulator therapy.

[0010] In some embodiments, the checkpoint modulator therapy is selected from the group consisting of (i) an activator of a stimulatory immune checkpoint molecule, such as an antibody molecule against GITR, OX-40, ICOS, 4-1BB, or a combination thereof; (ii) a RORy agonist; or (iii) an antibody against PD-1 (e.g., nivolumab, pembrolizumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, TSR-042, or an antigen-binding portion thereof), an antibody against PD-L1 (e.g., avelumab, atezolizumab, durvalumab, CX-072, LY330 ... 054, or an antigen-binding portion thereof), an antibody against PD-L2, or an antibody against CTLA-4, alone or in combination, or in combination with an inhibitor of TIM-3, LAG-3, BTLA, TIGIT, VISTA, TGF-β, LAIR1, CD160, 2B4, GITR, OX40, 4-1BB, CD2, CD27, CDS, ICAM-1, LFA-1, ICOS, CD30, CD40, BAFFR, HVEM, CD7, LIGHT, NKG2C, SLAMF7, NKp80, or CD86.

[0011] In some embodiments, the checkpoint modulator therapy comprises administering (i) an anti-PD-1 antibody selected from the group consisting of nivolumab, pembrolizumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, or TSR-042; (ii) an anti-PD-L1 antibody selected from the group consisting of avelumab, atezolizumab, CX-072, LY3300054, and durvalumab; or (iii) a combination thereof. In some embodiments, the TME phenotype class-specific therapy is an IS class TME therapy comprising administering (1) a checkpoint modulator therapy and an anti-immunosuppressive therapy, and / or (2) an anti-angiogenic therapy. In some embodiments, the checkpoint modulator therapy comprises administering an inhibitor of an inhibitory immune checkpoint molecule. In some embodiments, the inhibitor of an inhibitory immune checkpoint molecule is (i) an antibody against PD-1 selected from the group consisting of pembrolizumab, nivolumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, TSR-042, antigen-binding portions thereof, and combinations thereof; (ii) an antibody against PD-L1 selected from the group consisting of avelumab, atezolizumab, CX-072, LY3300054, durvalumab, antigen-binding portions thereof, and combinations thereof; (iii) an antibody against PD-L2 or antigen-binding portion thereof; (iv) an antibody against CTLA-4 selected from ipilimumab and the bispecific antibody XmAb20717 (anti-PD-1 / anti-CTLA-4); or (v) a combination thereof.

[0012] In some embodiments, the anti-angiogenic therapy comprises administering (i) an anti-VEGF antibody selected from the group consisting of valisacumab, bevacizumab, nabicikizumab (anti-DLL4 / anti-VEGF bispecific), ABL101 (NOV1501) (anti-DLL4 / anti-VEGF), ABT165 (anti-DLL4 / anti-VEGF), and combinations thereof; (ii) an anti-VEGFR2 antibody, including ramucirumab; or (iii) a combination thereof. In some embodiments, the anti-immunosuppressive therapy comprises administering an anti-PS antibody, an anti-PS targeting antibody, an antibody that binds to beta-2 glycoprotein 1, an inhibitor of PI3Kγ, an adenosine pathway inhibitor, an inhibitor of IDO, an inhibitor of TIM, an inhibitor of LAG3, an inhibitor of TGFβ, a CD47 inhibitor, or a combination thereof, wherein (i) the anti-PS targeting antibody is bavituximab or an antibody that binds to beta-2 glycoprotein 1; (ii) the PI3Kγ inhibitor is LY3023414 (samotricisib) or IPI-549; (iii) the adenosine pathway inhibitor is AB-928; (iv) the TGFβ inhibitor is LY2157299 (galunisertib) or the TGFβR1 inhibitor is LY3200882; the CD47 inhibitor is magrolimab (5F9); and (vi) the CD47 inhibitor targets SIRPα.

[0013] In some embodiments, the anti-immunosuppressive therapy comprises administering an inhibitor of TIM-3, LAG-3, BTLA, TIGIT, VISTA, TGF-β or its receptor, LAIR1, CD160, 2B4, GITR, OX40, 4-1BB, CD2, CD27, CDS, ICAM-1, LFA-1, ICOS, CD30, CD40, BAFFR, HVEM, CD7, LIGHT, NKG2C, SLAMF7, an inhibitor of NKp80, an agonist of CD86, or a combination thereof. In some embodiments, the TME phenotype class-specific therapy comprises administering a VEGF-targeted therapy, an inhibitor of Angiopoietin 1 (Ang1), an inhibitor of Angiopoietin 2 (Ang2), an inhibitor of DLL4, an anti-VEGF and anti-DLL4 bispecific, a TKI inhibitor, an anti-FGF antibody, an anti-FGFR1 antibody, an anti-FGFR2 antibody, a small molecule that inhibits FGFR1, a small molecule that inhibits FGFR2, an anti-PLGF antibody, a small molecule against the PLGF receptor, an antibody against the PLGF receptor, an anti-VEGFB antibody, an anti-VEGFC antibody, an anti-VEGFD antibody, an antibody against a VEGF / PLGF trapping molecule such as aflibercept or ziv-aflibercept, an anti-DLL4 antibody, an anti-Notch therapy such as an inhibitor of gamma secretase, or any combination thereof.

[0014] In some embodiments, the TKI inhibitor is selected from the group consisting of cabozantinib, vandetanib, tivozanib, axitinib, lenvatinib, sorafenib, regorafenib, sunitinib, fruquintinib, pazopanib, and any combination thereof.

[0015] In some embodiments, the VEGF-targeted therapy comprises administering (i) an anti-VEGF antibody, including valisacumab, bevacizumab, an antigen-binding portion thereof, or a combination thereof; (ii) an anti-VEGFR2 antibody, including ramucirumab or an antigen-binding portion thereof; or (iii) a combination thereof.

[0016] In some embodiments, the A TME phenotype class-specific therapy comprises administering an angiopoietin / TIE2-targeted therapy including endoglin and / or angiopoietin. In some embodiments, the A TME phenotype class-specific therapy comprises administering a DLL4-targeted therapy including nabicikizumab, ABL101 (NOV1501), ABT165, or a combination thereof. In some embodiments, the TME phenotype class-specific therapy is an ID TME phenotype class-specific therapy comprising administering a checkpoint modulator therapy simultaneously with or after administration of a therapy that initiates an immune response. In some embodiments, the therapy that initiates an immune response is a vaccine, CAR-T, or neoepitope vaccine. In some embodiments, the checkpoint modulator therapy comprises administering an inhibitor of an inhibitory immune checkpoint molecule. In some embodiments, the inhibitor of an inhibitory immune checkpoint molecule is an antibody against PD-1, PD-L1, PD-L2, CTLA-4, or a combination thereof. In some embodiments, the anti-PD-1 antibody comprises nivolumab, pembrolizumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, or TSR-042, or an antigen-binding portion thereof. In some embodiments, the anti-PD-L1 antibody comprises avelumab, atezolizumab, CX-072, LY3300054, durvalumab, or an antigen-binding portion thereof. In some embodiments, the anti-CTLA-4 antibody comprises ipilimumab or the bispecific antibody XmAb20717 (anti-PD-1 / anti-CTLA-4), or an antigen-binding portion thereof. In some embodiments, the checkpoint modulator therapy comprises administration of (i) an anti-PD-1 antibody selected from the group consisting of nivolumab, pembrolizumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, and TSR-042; (ii) an anti-PD-L1 antibody selected from the group consisting of avelumab, atezolizumab, CX-072, LY3300054, and durvalumab; (iv) an anti-CTLA-4 antibody, which is ipilimumab or the bispecific antibody XmAb20717 (anti-PD-1 / anti-CTLA-4), or (iii) a combination thereof.

[0017] In some embodiments, the method further comprises (a) administering chemotherapy; (b) performing surgery; (c) administering radiation therapy; or (d) any combination thereof. In some embodiments, the cancer is recurrent, refractory, metastatic, dMMR, or a combination thereof. In some embodiments, the cancer is refractory after at least one prior treatment including administration of at least one anticancer agent. In some embodiments, the cancer is selected from the group consisting of gastric cancer, breast cancer, prostate cancer, liver cancer, head and neck cancer, melanoma, colorectal cancer, ovarian cancer, glioma, glioblastoma, or lung cancer. In some embodiments, the gastric cancer is locally advanced, metastatic, or previously untreated gastric cancer. In some embodiments, the breast cancer is locally advanced or metastatic HER2-negative breast cancer. In some embodiments, the prostate cancer is castration-resistant metastatic prostate cancer. In some embodiments, the liver cancer is hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma. In some embodiments, the head and neck cancer is recurrent or metastatic head and neck squamous cell carcinoma. In some embodiments, the colorectal cancer is advanced colorectal cancer metastatic to the liver. In some embodiments, the ovarian cancer is platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer. In some embodiments, the glioma is metastatic glioma. In some embodiments, the lung cancer is NSCLC.

[0018] In some embodiments, administration of the TME phenotype class-specific therapy reduces the cancer burden by at least about 10%, 20%, 30%, 40%, or 50% compared to the cancer burden prior to administration. In some embodiments, the subject exhibits progression-free survival of at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 months, or at least about 1, 2, 3, 4, or 5 years after the initial administration of the TME phenotype class-specific therapy. In some embodiments, the subject exhibits stable disease for about 1 month, about 2 months, about 3 months, about 4 months, about 5 months, about 6 months, about 7 months, about 8 months, about 9 months, about 10 months, about 11 months, about 1 year, about 18 months, about 2 years, about 3 years, about 4 years, or about 5 years after the initial administration of the TME phenotype class-specific therapy. In some embodiments, the subject exhibits a partial response about 1 month, about 2 months, about 3 months, about 4 months, about 5 months, about 6 months, about 7 months, about 8 months, about 9 months, about 10 months, about 11 months, about 1 year, about 18 months, about 2 years, about 3 years, about 4 years, or about 5 years after the first administration of the TME phenotype class-specific therapy. In some embodiments, the subject exhibits a complete response about 1 month, about 2 months, about 3 months, about 4 months, about 5 months, about 6 months, about 7 months, about 8 months, about 9 months, about 10 months, about 11 months, about 1 year, about 18 months, about 2 years, about 3 years, about 4 years, or about 5 years after the first administration of the TME phenotype class-specific therapy.

[0019] In some embodiments, administration of the TME phenotype class-specific therapy improves progression-free survival probability by at least about 10%, at least about 20%, at least about 30%, at least about 40%, at least about 50%, at least about 60%, at least about 70%, at least about 80%, at least about 90%, at least about 100%, at least about 110%, at least about 120%, at least about 130%, at least about 140%, or at least about 150% compared to the progression-free survival probability of a subject who did not receive the TME phenotype class-specific therapy as assigned using an ANN classifier such as TME Panel-1. In some embodiments, administration of the TME phenotype class-specific therapy improves overall survival probability by at least about 25%, at least about 50%, at least about 75%, at least about 100%, at least about 125%, at least about 150%, at least about 175%, at least about 200%, at least about 225%, at least about 250%, at least about 275%, at least about 300%, at least about 325%, at least about 350%, or at least about 375% compared to the overall survival probability of a subject who did not receive the TME phenotype class-specific therapy as assigned using an ANN classifier such as TME Panel-1.

[0020] Also provided is a method for assigning a TME phenotype class to cancer in a subject in need thereof, the method comprising: (i) generating an ANN classifier by training the ANN with a training set comprising RNA expression levels for each gene in a gene panel of multiple samples from multiple subjects, wherein each sample is assigned a TME phenotype class; and (ii) using the ANN classifier to assign a TME phenotype class to the cancer in the subject, wherein input to the ANN classifier comprises RNA expression levels for each gene in the gene panel of test samples from the subject.

[0021] Also provided is a method for assigning a TME phenotype class to a cancer in a subject in need thereof, the method comprising: generating an ANN classifier by training the ANN with a training set comprising RNA expression levels for each gene in a gene panel of a plurality of samples from a plurality of subjects, where each sample is assigned a TME phenotype class, and the ANN classifier uses the RNA expression levels for each gene in the gene panel of a test sample from the subject as input to assign a TME phenotype class to the cancer in the subject. Also provided is a method for assigning a TME phenotype class to a cancer in a subject in need thereof, the method comprising predicting the TME phenotype class of the cancer in the subject using an ANN classifier, the ANN classifier being generated by training the ANN with a training set comprising RNA expression levels for each gene in the gene panel of a plurality of samples from a plurality of subjects, where each sample is assigned a TME phenotype class or a combination thereof. In some embodiments, the method is implemented in a computer system comprising at least one processing device and at least one storage device, and the at least one storage device comprises instructions, executed by the at least one processing device, that cause the at least one processing device to implement a machine learning model. In some embodiments, the method further includes (i) inputting the ANN classifier code into a storage device of the computer system; (ii) inputting gene panel input data corresponding to the subject into a storage device of the computer system, the input data comprising RNA expression levels; (iii) executing the ANN classifier code; or (v) any combination thereof.

[0022] Also provided is a method of treating a subject with locally advanced or metastatic gastric cancer using an IA TME phenotype, the method comprising administering to the subject an IA TME phenotype class-specific therapy, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with locally advanced or metastatic gastric cancer using an A TME phenotype, the method comprising administering to the subject an A TME phenotype class-specific therapy, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with locally advanced or metastatic gastric cancer using an IS TME phenotype, the method comprising administering to the subject an IS TME phenotype class-specific therapy, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a previously untreated subject with gastric cancer using an IS TME phenotype, the method comprising administering an IS TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a previously untreated subject with gastric cancer using an A TME phenotype, the method comprising administering an A TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with locally advanced / metastatic HER2-negative breast cancer using an A TME phenotype, the method comprising administering an A TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject.Also provided is a method of treating a subject with locally advanced / metastatic HER2-negative breast cancer using an IS TME phenotype, the method comprising administering an IS TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with castration-resistant metastatic prostate cancer using an A TME phenotype, the method comprising administering an A TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with castration-resistant metastatic prostate cancer using an IS TME phenotype, the method comprising administering an IS TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with advanced, metastatic hepatocellular carcinoma using the IA TME phenotype, the method comprising administering to the subject an IA TME phenotype class-specific therapy, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with advanced, metastatic hepatocellular carcinoma using the IS TME phenotype, the method comprising administering to the subject an IS TME phenotype class-specific therapy, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with recurrent / metastatic head and neck squamous cell carcinoma using the IA TME phenotype, the method comprising administering to the subject an IA TME phenotype class-specific therapy, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject.Also provided is a method of treating a subject with recurrent / metastatic head and neck squamous cell carcinoma with an IS TME phenotype, the method comprising administering an IA TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with melanoma with an IA TME phenotype, the method comprising administering an IA TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with melanoma with an IS TME phenotype, the method comprising administering an IS TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with advanced colorectal cancer metastatic to the liver with an ID TME phenotype, the method comprising administering to the subject an ID TME phenotype class-specific therapy, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject.Also provided is a method of treating a subject with platinum-resistant or platinum-sensitive recurrent ovarian cancer with an IA, IS, or A TME phenotype, the method comprising administering to the subject an IA, IS, or A TME phenotype class-specific therapy, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject.

[0023] Also provided are methods of treating a subject with platinum-resistant or platinum-sensitive recurrent triple-negative breast cancer with an IA, IS, or A TME phenotype, the method comprising administering an IA, IS, or A TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided are methods of treating a subject with melanoma with an IS TME phenotype, the method comprising administering an IS TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided are methods of treating a subject with metastatic colorectal cancer with an A or IS TME phenotype, the method comprising administering an A or IS TME phenotype class-specific therapy to the subject, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject. Also provided is a method of treating a subject with glioma or glioblastoma with an IS or IA TME phenotype, the method comprising administering to the subject an IS or IA TME phenotype class-specific therapy, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject.Also provided is a method of treating a subject with non-small cell lung cancer with an IS or IA TME phenotype, the method comprising administering to the subject an IS or IA TME phenotype class-specific therapy, wherein the TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject.

[0024] The disclosure also provides kits comprising (i) a plurality of oligonucleotide probes capable of specifically detecting RNA encoding the genetic biomarkers of Table 1, and (ii) a plurality of oligonucleotide probes capable of specifically detecting RNA encoding the genetic biomarkers of Table 2. Also provided are articles of manufacture comprising (i) a plurality of oligonucleotide probes capable of specifically detecting RNA encoding the genetic biomarkers of Table 1 (or Figures 9A-9G), and (ii) a plurality of oligonucleotide probes capable of specifically detecting RNA encoding the genetic biomarkers of Table 2 (or Figures 9A-9G), wherein the articles of manufacture comprise microarrays.

[0025] The present disclosure provides an ANN classifier that includes: (a) an input layer comprising 2 to 100 nodes, each node in the input layer corresponding to a gene in a gene panel selected from the genes set forth in Table 1 and Table 2, the gene panel comprising: (i) 1 to 63 genes selected from Table 1 and 1 to 61 genes selected from Table 2; (ii) a gene panel comprising genes selected from Table 3 and Table 4; (iii) a gene panel of Table 5; or (iv) any of the gene panels (gene sets) disclosed in Figures 9A-G; (b) a hidden layer containing two nodes; and (c) an output layer including four output nodes, any one of the four output nodes in the output layer corresponding to a TME phenotype class, the four TME phenotype classes being IA, IS, ID, and A, and optionally further including applying a logistic regression analysis classifier including a softmax function to the output of the ANN, the softmax function assigning a probability to each TME phenotype class.

[0026] The present disclosure provides an IA TME phenotype class-specific therapy comprising a checkpoint modulator therapy comprising administering: (i) activators of stimulatory immune checkpoint molecules, such as antibody molecules against GITR, OX-40, ICOS, 4-1BB, or combinations thereof; (ii) RORγ agonists; (iii) inhibitory antibodies, such as antibodies against PD-1 (e.g., nivolumab, pembrolizumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, TSR-042, or antigen-binding portions thereof), antibodies against PD-L1 (e.g., avelumab, atezolizumab, durvalumab, CX-072, LY3300054, or antigen-binding portions thereof), antibodies against PD-L2, or antibodies against CTLA-4 Inhibitors of immune checkpoint molecules, alone or in combination, or in combination with inhibitors of TIM-3, LAG-3, BTLA, TIGIT, VISTA, TGF-β, LAIR1, CD160, 2B4, GITR, OX40, 4-1BB, CD2, CD27, CDS, ICAM-1, LFA-1, ICOS, CD30, CD40, BAFFR, HVEM, CD7, LIGHT, NKG2C, SLAMF7, NKp80, or CD86; or (iv) Combinations of these.

[0027] The present disclosure provides IS-class TME therapies, the therapies comprising: (1) checkpoint modulator therapy and anti-immunosuppressant therapy, and / or (2) antiangiogenic therapy, and the checkpoint modulator therapy comprises administering an inhibitor of an inhibitory immune checkpoint molecule, wherein the inhibitory immune checkpoint molecule comprises: (i) an antibody against PD-1 selected from the group consisting of pembrolizumab, nivolumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, TSR-042, antigen-binding portions thereof, and combinations thereof; (ii) an antibody against PD-L1 selected from the group consisting of avelumab, atezolizumab, CX-072, LY3300054, durvalumab, antigen-binding portions thereof, and combinations thereof; (iii) an antibody or antigen-binding portion thereof against PD-L2; (iv) an antibody against CTLA-4 selected from ipilimumab and the bispecific antibody XmAb20717 (anti-PD-1 / anti-CTLA-4); or (v) combinations thereof, including Antiangiogenic therapy is (a) an anti-VEGF antibody selected from the group consisting of valisacumab, bevacizumab, nabicikizumab (anti-DLL4 / anti-VEGF bispecific), ABL101 (NOV1501) (anti-DLL4 / anti-VEGF), ABT165 (anti-DLL4 / anti-VEGF), and combinations thereof; (b) an anti-VEGFR2 antibody, including ramucirumab; or (c) a combination thereof, and anti-immunosuppressant therapy. (a) an anti-PS antibody, an anti-PS targeting antibody, an antibody that binds to β2 glycoprotein 1, an inhibitor of PI3Kγ, an adenosine pathway inhibitor, an inhibitor of IDO, an inhibitor of TIM, an inhibitor of LAG3, an inhibitor of TGFβ, a CD47 inhibitor, or a combination thereof (the anti-PS targeting antibody is bavituximab, or an antibody that binds to β2 glycoprotein 1; the PI3Kγ inhibitor is LY3023414 (samotricisib) or IPI-549; the adenosine pathway inhibitor is AB-928; the TGFβ inhibitor is LY2157299 (galunisertib) or the TGFβR1 inhibitor is LY3200882; the CD47 inhibitor is magrolimab (5F9); and the CD47 inhibitor targets SIRPα); (b) an inhibitor of TIM-3, LAG-3, BTLA, TIGIT, VISTA, TGF-β or its receptor, LAIR1, CD160, 2B4, GITR, OX40, 4-1BB, CD2, CD27, CDS, ICAM-1, LFA-1, ICOS, CD30, CD40, BAFFR, HVEM, CD7, LIGHT, NKG2C, SLAMF7, an inhibitor of NKp80, an agonist of CD86, or a combination thereof; or (c) a combination thereof.

[0028] The present disclosure provides an A TME phenotype class-specific therapy comprising administering: (i) VEGF-targeted therapy, inhibitors of angiopoietin 1 (Ang1), inhibitors of angiopoietin 2 (Ang2), inhibitors of DLL4, anti-VEGF and anti-DLL4 bispecifics, TKI inhibitors, anti-FGF antibodies, anti-FGFR1 antibodies, anti-FGFR2 antibodies, small molecules that inhibit FGFR1, small molecules that inhibit FGFR2, anti-PLGF antibodies, small molecules against PLGF receptors, antibodies against PLGF receptors, anti-VEGFB antibodies, anti-VEGFC antibodies, anti-VEGFD antibodies, antibodies against VEGF / PLGF trapping molecules such as aflibercept or ziv-aflibercept, anti-DLL4 antibodies, gamma-secretase inhibitors, etc. which anti-Notch therapy, or any combination thereof, wherein the TKI inhibitor is selected from the group consisting of cabozantinib, vandetanib, tivozanib, axitinib, lenvatinib, sorafenib, regorafenib, sunitinib, fruquintinib, pazopanib, and any combination thereof, and the VEGF-targeted therapy comprises administering (a) an anti-VEGF antibody, including valisacumab, bevacizumab, an antigen-binding portion thereof, or a combination thereof; (b) an anti-VEGFR2 antibody, including ramucirumab, or an antigen-binding portion thereof; or (c) a combination thereof; (ii) angiopoietin / TIE2 targeted therapy, including endoglin and / or angiopoietin; or (iii) DLL4-targeted therapy including nabi- cikizumab, ABL101 (NOV1501), ABT165, or a combination thereof.

[0029] The present disclosure provides an ID TME phenotype class-specific therapy comprising administering a checkpoint modulator therapy concurrently with or following administration of a therapy that initiates an immune response, wherein the checkpoint modulator therapy comprises administration of an inhibitor of an inhibitory immune checkpoint molecule, such as an antibody against PD-1, PD-L1, PD-L2, CTLA-4, or a combination thereof, and the therapy that initiates an immune response is a vaccine, CAR-T, or a neo-epitope vaccine; (i) the anti-PD-1 antibody includes nivolumab, pembrolizumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, or TSR-042, or an antigen-binding portion thereof; (ii) the anti-PD-L1 antibody includes avelumab, atezolizumab, CX-072, LY3300054, durvalumab, or an antigen-binding portion thereof; and (iii) The anti-CTLA-4 antibody includes ipilimumab or the bispecific antibody XmAb20717 (anti-PD-1 / anti-CTLA-4), or an antigen-binding portion thereof.

[0030] In some embodiments, the TME phenotype class-specific therapy disclosed herein is combined with (a) administering chemotherapy; (b) performing surgery; (c) administering radiation therapy; or (d) any combination thereof.

[0031] In some embodiments, the cancer is selected from the group consisting of: (i) gastric cancer, such as locally advanced, metastatic, or previously untreated gastric cancer; (ii) breast cancer, such as locally advanced, triple-negative breast cancer, or metastatic HER2-negative breast cancer; (iii) prostate cancer, such as castration-resistant metastatic prostate cancer; (iv) liver cancer, such as advanced metastatic hepatocellular carcinoma; (v) head and neck cancer, such as recurrent or metastatic head and neck squamous cell carcinoma; (vi) melanoma, such as metastatic melanoma; (vii) colorectal cancer, such as advanced colorectal cancer metastatic to the liver; (viii) ovarian cancer, such as platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer; (ix) glioma, such as metastatic glioma; (x) lung cancer, such as non-small cell lung cancer (NSCLC); and (xi) glioblastoma.

[0032] In some embodiments, administration of a TME phenotype class-specific therapy results in (i) a reduction in cancer burden by at least about 10%, 20%, 30%, 40%, or 50% compared to the cancer burden prior to administration; (ii) progression-free survival of at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 months, or at least about 1, 2, 3, 4, or 5 years after the first administration of the TME phenotype class-specific therapy; (iii) a reduction in the cancer burden by at least about 10%, 20%, 30%, 40%, or 50% compared to the cancer burden prior to administration; (iv) a reduction in the cancer burden by at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 months, or at least about 1, 2, 3, 4, or 5 years, after the first administration of the TME phenotype class-specific therapy; (iv) stable for about 1 month, about 2 months, about 3 months, about 4 months, about 5 months, about 6 months, about 7 months, about 8 months, about 9 months, about 10 months, about 11 months, about 1 year, about 18 months, about 2 years, about 3 years, about 4 years, or about 5 years after the first dose of TME phenotype class-specific therapy; (iv) stable for about 1 month, about 2 months, about 3 months, about 4 months, about 5 months, about 6 months, about 7 months, about 8 months, about 9 months, about 10 months, about 11 months, about 1 year, about 18 months, or about 5 years after the first dose of TME phenotype class-specific therapy; (v) a complete response at about 1 month, about 2 months, about 3 months, about 4 months, about 5 months, about 6 months, about 7 months, about 8 months, about 9 months, about 10 months, about 11 months, about 1 year, about 18 months, about 2 years, about 3 years, about 4 years, or about 5 years after the first dose of TME phenotype class-specific therapy; (vi) a complete response at about 1 month, about 2 months, about 3 months, about 4 months, about 5 months, about 6 months, about 7 months, about 8 months, about 9 months, about 10 months, about 11 months, about 1 year, about 18 months, about 2 years, about 3 years, about 4 years, or about 5 years after the first dose of TME phenotype class-specific therapy assigned using an ANN classifier such as TME Panel-1. an improved probability of progression-free survival of at least about 10%, at least about 20%, at least about 30%, at least about 40%, at least about 50%, at least about 60%, at least about 70%, at least about 80%, at least about 90%, at least about 100%, at least about 110%, at least about 120%, at least about 130%, at least about 140%, or at least about 150% compared to the probability of progression-free survival of subjects who did not receive the treatment;(vii) an improved overall survival probability of at least about 25%, at least about 50%, at least about 75%, at least about 100%, at least about 125%, at least about 150%, at least about 175%, at least about 200%, at least about 225%, at least about 250%, at least about 275%, at least about 300%, at least about 325%, at least about 350%, or at least about 375% compared to the overall survival probability of a subject who did not receive the TME phenotype class-specific therapy assigned using an ANN classifier such as TME Panel-1; or (viii) a combination thereof.

[0033] The present disclosure provides a method for assigning a TME phenotype class to cancer in a subject in need of treatment, the method comprising: (i) generating an ANN classifier by training the ANN with a training set comprising RNA expression levels for each gene in a gene panel of a plurality of samples from a plurality of subjects, each sample being assigned a TME phenotype classification; and using the ANN classifier to assign a TME phenotype class to the subject's cancer, the input to the ANN classifier comprising the RNA expression levels for each gene in the gene panel of test samples from the subjects; or (ii) generating an ANN classifier by training the ANN with a training set comprising RNA expression levels for each gene in the gene panel of a plurality of samples from a plurality of subjects, each sample being assigned a TME phenotype classification; the ANN classifier using as input the RNA expression levels for each gene in the gene panel of test samples from the subjects to assign a TME phenotype class to the subject's cancer; (iii) predicting the TME phenotype class of the subject's cancer using an ANN classifier, wherein the ANN classifier is generated by training the ANN with a training set comprising RNA expression levels for each gene in a gene panel of multiple samples from multiple subjects, and each sample is assigned to a TME phenotype class or a combination thereof.

[0034] The present disclosure provides a method of treating a subject with cancer with a specific TME phenotype, the method comprising administering a TME phenotype class-specific therapy to the subject, (i) the cancer is locally advanced, metastatic gastric cancer, and the TME phenotype is IA, A, or IS; (ii) the cancer is untreated gastric cancer and the TME phenotype is IS or A; (iii) the cancer is advanced / metastatic HER2-negative breast cancer and the TME phenotype is A or IS; (iv) the cancer is castration-resistant metastatic prostate cancer and the TME phenotype is A or IS; (v) the cancer is advanced metastatic hepatocellular carcinoma and the TME phenotype is IA or IS; (vi) the cancer is recurrent / metastatic head and neck squamous cell carcinoma and the TME phenotype is IA or IS; (vii) the cancer is melanoma and the TME phenotype is IA or IS; (viii) the cancer is advanced colorectal cancer metastatic to the liver and the TME phenotype is ID; (ix) the cancer is platinum-resistant or platinum-sensitive recurrent ovarian cancer and the TME phenotype is IA, IS, or A; (x) the cancer is platinum-resistant or platinum-sensitive recurrent triple-negative breast cancer and the TME phenotype is IA, IS, or A; (xi) the cancer is metastatic colorectal cancer and the TME phenotype is A or IS; (xii) the cancer is a glioma or glioblastoma and the TME phenotype is IS or IA; or (xiii) the cancer is non-small cell lung cancer and the TME phenotype is IS or IA; TME phenotype classes are assigned by applying an ANN classifier to multiple RNA expression levels obtained from a gene panel from a cancer tumor sample derived from a subject, the ANN classifier comprising: (a) an input layer comprising 2 to 100 nodes, each node in the input layer corresponding to a gene in a gene panel selected from the genes set forth in Table 1 and Table 2, the gene panel comprising: (i) 1 to 63 genes selected from Table 1 and 1 to 61 genes selected from Table 2; (ii) a gene panel comprising genes selected from Table 3 and Table 4; (iii) a gene panel of Table 5; or (iv) any of the gene panels (gene sets) disclosed in Figures 9A-G; (b) a hidden layer containing two nodes; and (c) an output layer including four output nodes, any one of the four output nodes in the output layer corresponding to a TME phenotype class, the four TME phenotype classes being IA, IS, ID, and A; and optionally further comprising a logistic regression classifier that includes a softmax function on the output of the ANN, the softmax function assigning a probability to each TME phenotype class. [Brief explanation of the drawings]

[0035] [Figure 1] Figure 1 shows the four TME (tumor microenvironment) phenotype classes assigned by the TME Panel 1 classifier. The angiogenic TME phenotype class, A, is characterized by high angiogenesis and a low immune signature score. Pathological angiogenesis promotes tumor growth and metastasis. The immunosuppressive TME phenotype class, IS, is characterized by high angiogenesis and a high immune signature score. The immune complement is primarily composed of suppressor cells. The immune desert TME phenotype class, ID, is characterized by low angiogenesis and a low immune signature score. Immune cells are absent, and the vasculature is functional. The immunoactive TME phenotype class, IA, is characterized by low angiogenesis and a high immune signature score. T cells are infiltrating but may not function optimally. [Figure 2A]Incidence of TME phenotype classes in the CIT and Wood-Hudson CRC datasets is shown. The CIT dataset was divided into early (0-2) and late (3-4) disease and compared to Wood-Hudson, where 89 of 93 patients were stage 3-4. For each data subset, the proportion of patients classified as angiogenic (A), immune-activated (IA), immune-desert (ID), and immune-suppressed (IS) was tallied. TME phenotype classes are color-coded according to the legend. [Figure 2B] Left- and right-sided CRC components of each TME phenotype class by stage, as classified by the TME Panel 1 classifier, are shown. The data subset shown in Figure 2A was further divided based on tumor side, left (distal) or right (proximal), and TME phenotype class proportions were recounted. [Figure 3A] Figure 1 shows a Kaplan-Meier plot of disease-free survival (DFS) for patients with early stage (0-2) disease in the CIT dataset classified by the TME Panel 1 classifier. Each survival curve represents a TME Panel 1 phenotype class, as indicated in the legend. Uppercase N is the number of patients, and lowercase n is the number of deaths. In the absence of antiangiogenic drug treatment, patients with tumors in the A TME phenotype class had the worst prognosis, followed by patients with tumors in the IS TME phenotype class. Patients in the IA TME phenotype class had the best outcome. [Figure 3B] Kaplan-Meier plot of overall survival (OS) for patients with late-stage (3-4) Wood-Hudson CRC classified by the TME Panel 1 classifier. Each survival curve represents a TME Panel 1 phenotype class, as indicated in the legend. Uppercase N represents the number of patients, and lowercase n represents the number of deaths. In the absence of antiangiogenic drug treatment, patients with tumors in the A TME phenotype class had the worst prognosis and exhibited the lowest median survival, followed by patients with tumors in the IS TME phenotype class. [Figure 4A] Diagram of the angiogenic and immune axes underlying the latent space analysis shown in Figures 4B-4F. Each patient was plotted on a TME Panel 1 landscape defined by their immune signature (X-axis) and angiogenic signature (Y-axis). [Figure 4] Figures 4B-4E. Latent space plots of CMS classes 1-4 after classification by the TME Panel 1 classifier. Grayscale contours are probability bands representing the probability of a particular TME phenotype classification by the TME Panel 1 classifier. [Figure 4F] Figure 1 shows a latent space plot of unclassified patients from the CIT dataset after classification by the TME Panel 1 classifier. The grayscale contours are probability bands representing the probability of a particular TME phenotype classification by the TME Panel 1 classifier. [Figure 5A] Figure 1 shows the TME phenotype class distribution of the CIT dataset within each CMS group. For each CMS group, the proportion of patients in each TME class is indicated by shading according to the legend. [Figure 5B] Figure 5 shows the CMS distribution of the CIT dataset within each TME phenotype class. For each TME class, the proportion of patients in each CMS group is indicated by shading according to the legend. Figures 5A and 5B represent the same but reversed tabulation analysis. [Figure 6] Figures 6A and 6B show the incidence of DNA mismatch repair (dMMR)-deficient patients among CMS groups and TME, or interstitial, phenotypic classes. Approximately three-quarters of dMMR patients were captured by CMS1 (77%) (Figure 6A), whereas 96% of dMMR patients were classified as hyperimmune TME phenotypes (IA) and (IS) (Figure 6B). Groups and classes are shaded according to the legend. [Figure 7] 1 shows a simplified diagram of the TME Panel1 classifier of the present disclosure. The TME Panel1 classifier includes an input layer with inputs corresponding to each gene in a gene panel (e.g., a 124-gene panel, a 105-gene panel, a 98-gene panel, or an 87-gene panel), a hidden layer containing two neurons (alternatively, three, four, or five neurons), and an output layer that can correspond to a TME phenotype class assignment (i.e., a stromal phenotype assignment). [Figure 8] 1 is a table showing TME phenotype class assignments based on application of the TME Panel 1 classifier disclosed herein, as well as the therapeutic classes assigned to each TME phenotype class. [Figure 9A] The presence (empty cells) or absence (solid cells) of 124 genes in gene sets 1–44 is shown. [Figure 9B] The presence (empty cells) or absence (solid cells) of 124 genes in gene set 45–88 is shown. [Figure 9C] The presence (empty cells) or absence (solid cells) of 124 genes in gene set 89–132 is shown. [Figure 9D] The presence (empty cells) or absence (solid cells) of 124 genes in gene set 133–177 is shown. [Figure 9E] The presence (empty cells) or absence (solid cells) of 124 genes in gene set 178–222 is shown. [Figure 9F] The presence (empty cells) or absence (solid cells) of 124 genes in gene set 223–267 is shown. [Figure 9G] The presence (empty cells) or absence (solid cells) of 124 genes in gene set 268–282 is shown. [Figure 10] 1 is a latent space plot corresponding to vidutolimod / CMP-001. DETAILED DESCRIPTION OF THE INVENTION

[0036] The present disclosure provides methods for stratifying patients with gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) according to a diagnostic panel that uses gene expression data to classify patients based on the predominant tumor microenvironment biology. Strand-Tibbitts et al.2020 Development of an RNA-based Diagnostic Platform Based on the Tumor Microenvironment Dominant Biology.SITC(2020). See Figure 1.

[0037] The TME Panel 1 classifier ("TME Panel 1") used for stratifying cancer patients disclosed herein employs a machine learning model trained on two gene signatures: an angiogenic signature and an immune signature, representing the angiogenic and immune biology that govern tumor stroma, respectively. The combination of these biologies results in four distinct tumor microenvironment (TME) or stromal phenotypic classes: angiogenic (A), immunosuppressive (IS), immune activation (IA), and immune desert (ID). In some embodiments, the terms "immune desert" and "microenvironment desert" are used interchangeably. The TME Panel 1 classifier assigns patients to one of these four TME phenotypic classes based on gene expression, e.g., RNA expression, data from the patient's tumor sample. These TME phenotypic classes confer distinct prognostic risks, independent of disease stage or demographics. The TME Panel 1 classifier predicts outcome to antiangiogenic and checkpoint inhibitor therapies, including approved and investigational drugs. See, for example, U.S. Patent Application No. 17 / 089,234, which is incorporated herein by reference in its entirety.

[0038] TME Panel1 learns (latent) gene expression patterns that classify individual patients into specific TME phenotype classes. TME Panel1 compresses high-dimensional data (gene expression of all genes in the input gene set) into a low-dimensional (latent) space. TME Panel1 was initially trained and validated on gastric cancer. Analysis of over 2,000 biobanked patient samples showed that the classifier can also be applied to other cancers, such as colorectal cancer. TME panel 1 can stratify cancer patients, such as patients with gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), to predict treatment outcome and guide the selection of specific therapy.

[0039] In some aspects, the present disclosure provides methods of treating a subject, e.g., a human subject, suffering from a particular type of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), the method comprising administering a particular therapy depending on the classification of the cancer or patient into a particular TME phenotype class by a classifier disclosed herein, e.g., a TME panel 1 classifier.

[0040] Also provided are personalized therapies that can be administered to subjects with particular types of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), where the cancer or subject has been classified into a particular TME phenotype class according to a classifier disclosed herein, e.g., the TME Panel 1 classifier, or has been identified as not having a cancer that would be classified into a particular TME phenotype class.

[0041] The present disclosure also provides methods of treating a subject, e.g., a human subject, suffering from a particular type of cancer selected from the group consisting of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), the method comprising administering a particular therapy depending on the classification of the cancer or patient into a particular TME phenotype class by a classifier disclosed herein, e.g., a TME panel 1 classifier.

[0042] Also provided is a personalized therapy that can be administered to a subject with a particular type of cancer selected from the group consisting of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), wherein the cancer or subject has been classified into a particular TME phenotype class according to a classifier disclosed herein, e.g., the TME panel 1 classifier, or the cancer has been determined not to be classified into a particular TME phenotype class.

[0043] Application of the methods and components disclosed herein can improve clinical outcomes for cancer patients, e.g., patients with gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), by matching them to TME phenotype class-specific therapies having mechanisms of action that target one or more specific TME phenotype classes.

[0044] Similarly, application of the methods and components disclosed herein can improve clinical outcomes for cancer patients, e.g., patients with gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), by matching them to TME phenotype class-specific therapies having mechanisms of action that target one or more specific TME phenotype classes.

[0045] The dominant TME phenotype class can be directional, but can vary for any particular drug based on the complexity of the mechanism of action of the drug, drugs, or clinical regimen. Combinations of drugs or clinical regimens (i.e., one or more TME phenotype class-specific therapies disclosed below) can be applied to patients with tumors that are biomarker positive for more than one TME phenotype class, if applicable, or that are predominantly one TME phenotype class but with contributions from other TME phenotype classes as observed in the probability function of a model, e.g., the TME panel 1 classifiers disclosed herein. Thus, the term "predominantly," when applied to the TME phenotype classes disclosed herein, indicates that a patient or sample is positive for a particular TME phenotype class (e.g., IA), but that other TME phenotype classes (e.g., IS, ID, or A), or combinations thereof, also contribute to the biomarker signal as observed in the model, e.g., the TME panel 1 classifier probability function disclosed herein.

[0046] An advantage of the disclosed ANN classifiers, e.g., the TME Panel 1 classifier, over other classifiers known in the art is that samples from patients who are, for example, part of a clinical trial or clinical regimen, can be assigned to a particular TME phenotype class without reference to any other current patient data. Thus, while the availability of potential plots with probabilities for each TME is useful, precise assignment of a particular TME phenotype class is not required.

[0047] The disclosed ANN classifiers, such as the TME Panel 1 classifier, are particularly advantageous over classifiers known in the art, such as CMS (consensus molecular subtypes), in the case of colorectal cancer, which lack clear predictive value. Prognostic biomarkers are used to predict the course of disease, regardless of treatment. For example, patients with hepatocellular carcinoma (HCC) and high levels of alpha-fetoprotein (AFP) tend to have worse outcomes regardless of therapy, and obesity is a known prognostic biomarker for COVID-19 patient outcomes. CMS subtypes resulting from unsupervised clustering of input data (DNA, RNA, proteomics) are not as predictive as the TME phenotype classes A, IA, IS, and ID. Gene sets used to classify tumors within TME phenotype classes or classes based on angiogenesis signature scores and immune signature scores are empirically developed from the biology of the tumor microenvironment. Therefore, ANN methods trained with these gene sets, such as the TME Panel 1 classifier, predict beneficial treatment.

[0048] The CMS approach in colorectal cancer addresses many aspects of cancer biology. Stintzing et al. attempted to distinguish between cetuximab and bevacizumab in colorectal cancer because both drugs were approved for wild-type KRAS colorectal cancer in their respective clinical trials (Stintzing et al. (2019) Annals of Oncology 30:1796-1803). The study described by Stintzing, FIRE1, was inconclusive. Because CMS4 mesenchymal lineage is associated with stromal invasion, TGFB activation, and angiogenesis, it follows that antiangiogenic therapy should benefit some of these patients. However, in Stintzing et al.'s study, bevacizumab did not benefit CMS3 and CMS4 patients as much as cetuximab. Furthermore, because CMS2 is associated with WNT and MYC activation, it follows that EGFR inhibitors, including the anti-EGFR monoclonal antibody cetuximab, should benefit some of these patients. However, in the study by Stintzing et al., CMS2 patients benefited more from the antiangiogenic drug bevacizumab than from cetuximab.

[0049] The disclosed ANN classifier, e.g., the TME Panel 1 classifier, is not limited to colorectal cancer but addresses two biologies represented by two signatures. While the empirically determined genes in each signature represent genes related to angiogenesis or immune processes, the ANN method relies on input from genes related to angiogenesis (e.g., those in Table 1) or immune processes (e.g., those in Table 2) for both hidden nodes or neurons. Furthermore, the classifier output can be simplified by designating one hidden node or neuron as the angiogenesis axis, corresponding to the angiogenesis signature score, and the other as the immune axis, corresponding to the immune signature score. See Figure 1.

[0050] Application of the disclosed ANN classifier, e.g., the TME Panel 1 classifier, to colorectal cancer patient data demonstrates that the TME Panel 1 classifier is superior to CMS subtypes in predicting, for example, the response of tumors belonging to specific TME phenotype classes to bevacizumab (AVASTIN®) in colorectal cancer patients. Furthermore, the disclosed ANN classifier, e.g., the TME Panel 1 classifier, identifies TME differences between left-sided and right-sided colorectal cancer, allowing for the selection of TME phenotype-specific therapies that match the different phenotypes observed in left-sided and right-sided colorectal cancer. The differences in TME phenotype between left-sided and right-sided colorectal cancer provide an explanation for the response to bevacizumab (AVASTIN®), which cannot be explained based on the CMS classification. Very commonly, left-sided colorectal cancer has a more angiogenic stromal phenotype, and right-sided colorectal cancer has a more immune-stromal phenotype.

[0051] A classifier of the present disclosure, e.g., the TME Panel 1 Classifier, can capture colorectal cancer patient populations with the A TME phenotype class more efficiently than CMS4, thus enabling more accurate selection of appropriate therapy and being a more effective predictor of treatment response. Similarly, a classifier of the present disclosure, e.g., the TME Panel 1 Classifier, can more efficiently and completely capture patient populations with the IA TME phenotype class, which are suitable for checkpoint inhibitor therapy, even if they are not dMMR or MSI-H. Thus, a classifier of the present disclosure, e.g., the TME Panel 1 Classifier, can stratify colorectal cancer patients into specific subpopulations based on, for example, whether the cancer is metastatic, the location of the cancer (e.g., left-sided or right-sided), or the presence or absence of specific molecular biomarkers or features (e.g., MMR status or MSI-H status), and can assign individualized TME phenotype class-specific therapy to patients more accurately than other classifiers known in the art. This stratification of cancer patients, e.g., colorectal cancer patients, into specific subpopulations with specific TME phenotypes allows for more accurate prediction of treatment response to each available therapy, allowing clinicians to design a course of treatment that maximizes the probability of a favorable outcome.

[0052] A key advantage of the classifiers disclosed herein is that they are cross-organ compatible, i.e., the same oncology prediction platform can be applied to multiple types of cancer. While other biomarker-based approaches to classify different types of cancer rely on cancer-specific biomarkers and probe sets (e.g., each type of tumor cell requires a specific set of RNA probes that are cancer-specific), the approach of the present invention is based on the use of a common set of genes (called a "training set" or "definition set") that can be applied to different types of cancer. In other words, conventional biomarker-based tumor classification systems require a large number of probes that are likely specific to each cancer type (e.g., breast cancer, liver cancer, ovarian cancer, prostate cancer, etc.). Thus, for example, a probe set (e.g., in a kit, array, etc.) that selects a therapeutic agent for breast cancer cannot be used to select a therapeutic agent for prostate cancer or liver cancer. In contrast, the present method relies on a proprietary set of genes that are expressed not in cancer cells but in the tumor stroma, i.e., the cells surrounding the cancer cells, the vasculature, etc. Thus, a single set of RNA probes (e.g., in a kit, array, etc.) can be used to obtain RNA expression data that, after processing by the machine learning-based artificial intelligence platform disclosed herein, can yield predictions of preferred therapeutic agents or treatment outcomes for multiple types of cancer.

[0053] term In order that this disclosure may be more readily understood, certain terms are first defined. Unless otherwise expressly provided herein, as used in this disclosure, each of the following terms shall have the meaning set forth below. Additional definitions are set forth throughout this disclosure.

[0054] "Administration" refers to the physical introduction of a composition containing a therapeutic agent (e.g., a monoclonal antibody) into a subject, using any of a variety of methods and delivery systems known to those of skill in the art. Preferred routes of administration include intravenous, intramuscular, subcutaneous, intraperitoneal, spinal or other parenteral routes of administration, for example, by injection or infusion.

[0055] The phrase "parenteral administration," as used herein, refers to methods of administration other than enteral and topical administration, usually by injection, including, but not limited to, intravenous, intramuscular, intraarterial, intrathecal, intranodal, intralesional, intraarticular, intraorbital, intracardiac, intradermal, intraperitoneal, transtracheal, subcutaneous, subcuticular, intraarticular, subcapsular, subarachnoid, intraspinal, intraocular, intravitreal, periorbital, epidural, and intrasternal injection and infusion, as well as in vivo electroporation. Other non-injection routes include oral, topical, epidermal, or mucosal administration routes, such as intranasal, vaginal, rectal, sublingual, or topical. Also, for example, administration can be performed once, multiple times, and / or over one or more extended periods.

[0056] An "antibody" (Ab) shall include, but is not limited to, a glycoprotein immunoglobulin that specifically binds to an antigen and comprises at least two heavy (H) chains and two light (L) chains, or antigen-binding portions thereof, interconnected by sulfide bonds. Each H chain comprises a heavy chain variable region (referred to herein as V H The heavy chain constant region contains three constant domains: C H1 , C H2 , and C H3 Each light chain comprises a light chain variable region (referred to herein as V L The light chain constant region comprises one domain, C L Includes V H and V L The regions can be further subdivided into regions of hypervariability called complementarity determining regions (CDRs) interspersed with more conserved regions called framework regions (FRs). H and V L contains three CDRs and four FRs, arranged from the amino terminus to the carboxy terminus in the following order: FR1, CDR1, FR2, CDR2, FR3, CDR3, and FR4. The variable regions of the heavy and light chains contain binding domains that interact with antigens. The constant region of the antibody can mediate the binding of the immunoglobulin to host tissues or factors, including various cells of the immune system (e.g., effector cells) and the first component (C1q) of the classical complement system.

[0057] Immunoglobulins can be derived from any of the well-known isotypes, including, but not limited to, IgA, secretory IgA, IgG, and IgM. IgG subclasses are also well known to those skilled in the art and include, but are not limited to, IgG1, IgG2, IgG3, and IgG4. "Isotype" refers to the antibody class or subclass (e.g., IgM or IgG1) that is encoded by the heavy chain constant region genes.

[0058] The term "antibody" includes, for example, monoclonal antibodies; chimeric and humanized antibodies; human or non-human antibodies; fully synthetic antibodies; and single-chain antibodies. Non-human antibodies can be humanized by recombinant methods to reduce their immunogenicity in humans. Unless expressly stated and unless the context indicates otherwise, the term "antibody" also includes antigen-binding fragments or portions of any of the above-mentioned immunoglobulins, including monovalent and bivalent fragments or portions, and single-chain antibodies. As used herein, the term "antibody" does not include naturally occurring antibodies or polyclonal antibodies. As used herein, the terms "naturally occurring antibody" and "polyclonal antibody" do not include antibodies resulting from an immune response induced by therapeutic intervention, e.g., a vaccine.

[0059] An "isolated antibody" refers to an antibody that is substantially free of other antibodies having different antigenic specificities (e.g., an isolated antibody that specifically binds VEGF-A is substantially free of antibodies that specifically bind to antigens other than VEGF-A). However, an isolated antibody that specifically binds VEGF-A (e.g., bevacizumab, or an antigen-binding portion thereof) may have cross-reactivity to other antigens, such as VEGF-A molecules from different species. Furthermore, an isolated antibody may be substantially free of other cellular material and / or chemicals.

[0060] The term "monoclonal antibody" (mAb) refers to a non-naturally occurring preparation of antibody molecules of single molecular composition, i.e., antibody molecules which are substantially identical in primary sequence and which display a single binding specificity and affinity for a particular epitope. A monoclonal antibody is an example of an isolated antibody. Monoclonal antibodies can be produced by hybridoma, recombinant, transgenic, or other techniques known to those skilled in the art.

[0061] A "human antibody" (HuMAb) refers to an antibody having variable regions in which both the framework and CDR regions are derived from human germline immunoglobulin sequences. Furthermore, if the antibody contains a constant region, the constant region is also derived from human germline immunoglobulin sequences. The human antibodies of the present disclosure may include amino acid residues not encoded by human germline immunoglobulin sequences (e.g., mutations introduced by random or site-specific mutagenesis in vitro or by somatic mutation in vivo). However, as used herein, the term "human antibody" is not intended to include antibodies in which CDR sequences derived from the germline of another mammalian species, such as a mouse, have been grafted onto human framework sequences. The terms "human antibody" and "fully human antibody" are used interchangeably.

[0062] "Humanized antibody" refers to an antibody in which some, most, or all of the amino acids outside the CDRs of a non-human antibody have been replaced with corresponding amino acids from a human immunoglobulin. In one embodiment of a humanized antibody, some, most, or all of the amino acids outside the CDRs have been replaced with amino acids from a human immunoglobulin, while some, most, or all of the amino acids within one or more CDRs remain unchanged. Minor additions, deletions, insertions, substitutions, or modifications of amino acids are permissible as long as they do not abrogate the antibody's ability to bind to a specific antigen. A "humanized antibody" retains antigen specificity similar to that of the original antibody.

[0063] "Chimeric antibody" refers to an antibody whose variable region is derived from one species and whose constant region is derived from another species, e.g., whose variable region is derived from a mouse antibody and whose constant region is derived from a human antibody.

[0064] As used herein, a "bispecific antibody" means an antibody that comprises two antigen-binding sites, a first binding site that has affinity for a first antigen or epitope and a second binding site that has binding affinity for a second antigen or epitope that is different from the first.

[0065] "Anti-antigen antibody" means an antibody that specifically binds to an antigen. For example, an anti-VEGF-A antibody (e.g., bevacizumab, or an antigen-binding portion thereof) specifically binds to VEGF-A.

[0066] An "antigen-binding portion" (also called an "antigen-binding fragment") of an antibody refers to one or more fragments of an antibody that retain the ability to specifically bind to the antigen bound by the full-length antibody. It has been shown that the antigen-binding function of an antibody can be performed by fragments of a full-length antibody. Examples of binding fragments encompassed by the term "antigen-binding portion" of an antibody, e.g., an anti-VEGF-A antibody described herein (e.g., bevacizumab, or an antigen-binding portion thereof), include: (i) V L , V H (ii) a Fab fragment (a fragment derived from papain cleavage) or similar monovalent fragment consisting of the LC and CH1 domains; (iii) a F(ab')2 fragment (a fragment derived from pepsin cleavage) or similar bivalent fragment comprising two Fab fragments linked by a disulfide bridge at the hinge region; H and an Fd fragment consisting of the CH1 domain; (iv) a V of a single arm of an antibody; L and V H Fv fragment consisting of domains; (v) V H (vi) isolated complementarity-determining regions (CDRs); and (vii) combinations of two or more isolated CDRs, optionally linked by synthetic linkers. Additionally, Fv fragments, V L and V HThe two domains are encoded by separate genes, but they can be synthesized by recombinant methods. L and V H The regions can be linked by a synthetic linker, allowing them to be produced as a single protein chain, pairing to form a monovalent molecule (known as single-chain Fv (scFv); see, e.g., Bird et al. (1988) Science 242:423-426; and Huston et al. (1988) Proc. Natl. Acad. Sci. USA 85:5879-5883). Such single-chain antibodies are also intended to be encompassed by the term "antigen-binding portion" of an antibody. These antibody fragments are obtained using techniques available in the art, and the fragments are screened for utility in the same manner as intact antibodies. Antigen-binding portions can be produced by recombinant DNA techniques or by enzymatic or chemical cleavage of intact immunoglobulins.

[0067] As used herein, the term "antibody," when applied to a specific antigen, also encompasses antibody molecules that contain other binding moieties with different binding specificities. Thus, in one embodiment, the term antibody also encompasses antibody-drug conjugates (ADCs). In another embodiment, the term antibody encompasses multispecific antibodies, e.g., bispecific antibodies. Thus, for example, the term anti-VEGF-A antibody can also encompass ADCs that include an anti-VEGF-A antibody or an antigen-binding portion thereof. Similarly, the term anti-VEGF-A antibody can encompass bispecific antibodies that include an antigen-binding portion capable of specifically binding to VEGF-A.

[0068] "Cancer" refers to a wide variety of diseases characterized by the uncontrolled growth of abnormal cells in the body. Uncontrolled cell division and proliferation form malignant tumors that can invade neighboring tissues and metastasize to distant parts of the body via the lymphatic system or bloodstream. In some embodiments, the cancer disclosed herein is selected from the group consisting of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC). In some embodiments, the cancer is gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer). In some embodiments, the cancer is breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer). In some embodiments, the cancer is prostate cancer (e.g., castration-resistant metastatic prostate cancer). In some embodiments, the cancer is liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma). In some embodiments, the cancer is head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma). In some embodiments, the cancer is melanoma (e.g., metastatic melanoma). In some embodiments, the cancer is colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver). In some embodiments, the cancer is ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer). In some embodiments, the cancer is glioma (e.g., metastatic glioma). In some aspects, the cancer is glioblastoma. In some embodiments, the cancer is lung cancer (e.g., NSCLC).

[0069] The term "tumor" refers to a solid tumor. The term "carcinoma" refers to a cancer of epithelial origin.

[0070] As used herein, the term "stroma" refers to the entire mixture of cells, including endothelial cells, smooth muscle cells, pericytes, immune cells (including lymphoid and myeloid cell types), and supportive or connective tissues characterized by connective and / or supportive tissues located in or around tissues or organs, particularly tumor tissues or whole tumors found in vivo. Stromal components may not be characterized by a single type or species of cells or proteins. Instead, for example, they may be characterized by a variety of molecular biomarker species that characterize the entire stromal tissue component as observed in vivo in association with whole organs or tumors. Tumor growth and spread are determined not only by cancer cells but also by non-malignant components of a malignant lesion, which are encompassed by the term stroma. Thus, in some embodiments, the term stroma refers to the non-malignant components of a tumor. In some embodiments, the term stroma further includes the malignant components of a tumor, i.e., cancer cells. In some embodiments of the present disclosure, the terms stroma and tumor microenvironment (TME) are synonymous.

[0071] As used herein, the term "CMS" refers to a classification of colorectal cancer based on gene self-clustering in genomic DNA and RNA analyses of colorectal cancer. Guinney et al. (2015) Nature Medicine 21:1350-6. The classification results in four gene clusters called CMS or consensus molecular subtypes. "CMS1" is called MSI-immune and is characterized by MSI (microsatellite instability), CIMP (CpG island methylator phenotype)-high, hypermutation, BRAF mutation, immune infiltration and activation, and poorer survival after recurrence. "CMS2" is called standard and is characterized by SCNA (somatic copy number alteration)-high, WNT and MYC activity. "CMS3" is called metabolic and is characterized by mixed MSI status, SCNA-low, CIMP-low, KRAS mutation, and metabolic deregulation. "CMS4" is termed mesenchymal and is characterized by high SCNA, stromal infiltration, TGFβ activation, angiogenesis and worse recurrence-free and overall survival.

[0072] As used herein, the term "MSI-H" refers to microsatellite instability-high (MSH-high). Generally, this refers to cancer cells that have more than the normal number of genetic markers called microsatellites. Microsatellites are short, repeated sequences of DNA. Cancer cells with a large number of microsatellites may have a defective ability to correct errors that occur when DNA is replicated in the cell. Microsatellite instability is most frequently found in colorectal cancer, other types of gastrointestinal cancer, and endometrial cancer. It can also be found in breast cancer, prostate cancer, bladder cancer, and thyroid cancer, for example.

[0073] The term "dMMR" refers to a deficiency in mismatch repair. MSI-H / dMMR can occur when cells are unable to repair errors made during the division process.

[0074] The term "immunotherapy" refers to treatment by methods that involve inducing, enhancing, suppressing, or otherwise modifying the immune response of a subject suffering from a disease or at risk of suffering from or experiencing a recurrence of a disease. "Treatment" or "therapy" of a subject refers to any type of intervention or procedure performed on a subject, or administration of an active agent to a subject, for the purpose of ameliorating, alleviating, ameliorating, suppressing, slowing, or preventing the onset, progression, development, severity or recurrence of symptoms, complications or conditions, or biochemical indicators associated with a disease.

[0075] In the context of this disclosure, the term "immunosuppressed" or "immunosuppression" refers to the state of the immune response to cancer. A patient's immune response to cancer can be suppressed by immunosuppressive cells in the tumor microenvironment, thereby blocking, preventing, or diminishing the immune system's attack on the cancer. In immunosuppressive therapy, the goal is to alleviate immunosuppression by providing the patient with specific drugs so that the immune system can attack the cancer (rather than causing immunosuppression, as in, for example, organ transplantation).

[0076] The term "small molecule" refers to an organic compound having a molecular weight of less than about 900 daltons, or less than about 500 daltons. This term includes compounds that can be administered orally or by injection, including drugs with desirable pharmacological properties. This term includes organic compounds that modulate the activity of TGFβ and / or other molecules involved in enhancing or suppressing immune responses.

[0077] "VEGF-A," also known as vascular endothelial growth factor A, vascular permeability factor, VEGF, VPF, or MVCD1, refers to the gene or its expressed polypeptide, which is a member of the PDGF / VEGF growth factor family. VEGF-A encodes a heparin-binding protein. It is a growth factor that induces the proliferation and migration of vascular endothelial cells and is essential for both physiological and pathological angiogenesis. Disruption of this gene in mice resulted in aberrant embryonic angiogenesis. This gene is upregulated in many known tumors, and its expression correlates with tumor stage and progression. Variants of this gene have been reported, including, but not limited to, allelic variants associated with microvascular complications of diabetes 1 (MVCD1) and atherosclerosis, alternatively spliced ​​transcript variants encoding different isoforms, alternative translation initiation from upstream non-AUG (CUG) codons (resulting in additional isoforms), and C-terminally extended isoforms generated by the use of alternative in-frame translation termination codons via a stop codon read-through mechanism (this isoform is antiangiogenic). In some embodiments, the term VEGF-A encompasses the sequences of Unipro Acc. No. P15692, NCBI Gene ID 7422, and its homologs and isoforms.

[0078] "Programmed death 1" (PD-1) refers to an immunosuppressive receptor belonging to the CD28 family. PD-1 is expressed primarily in vivo on previously activated T cells and binds to two ligands, PD-L1 and PD-L2. As used herein, the term "PD-1" includes human PD-1 (hPD-1), variants, isoforms, and species homologs of hPD-1, as well as analogs that share at least one epitope with hPD-1. The complete hPD-1 sequence can be found under GenBank accession number U64863.

[0079] "Programmed death-ligand-1" (PD-L1) is one of two cell surface glycoprotein ligands for PD-1 (the other is PD-L2) that downregulates T cell activation and cytokine secretion upon binding to PD-1. As used herein, the term "PD-L1" includes human PD-L1 (hPD-L1), variants, isoforms, and species homologs of hPD-L1, and analogs that share at least one epitope with hPD-L1. The complete hPD-L1 sequence can be found in GenBank accession number Q9NZQ7. The human PD-L1 protein is encoded by the human CD274 gene (NCBI Gene ID: 29126).

[0080] As used herein, the term "subject" includes any human or non-human animal. The terms "subject" and "patient" are used interchangeably herein. The term "non-human animal" includes, but is not limited to, vertebrates such as dogs, cats, horses, cows, pigs, rams, goats, buffalo, bison, llamas, deer, other large animals, and their young, including calves and lambs, as well as primates, such as mice, rats, rabbits, guinea pigs, monkeys, and other laboratory animals. Among animals, mammals are preferred, and most preferred are valuable and useful animals, such as domestic pets, racehorses, and animals used to produce food directly (e.g., meat) or indirectly (e.g., milk) for human consumption, although laboratory animals are also included. In certain embodiments, the subject is a human. Thus, the present disclosure is applicable to clinical, veterinary, and research applications.

[0081] "Treat," "treating," and "treatment" of a subject refer to any type of intervention or procedure performed on a subject, or the administration of an active agent to a subject, for the purpose of ameliorating, alleviating, ameliorating, suppressing, slowing, or preventing the progression, development, severity or recurrence of symptoms, complications, conditions, or biochemical markers associated with a disease, or improving overall survival. Treatment can be of a subject with a disease or a subject without a disease (e.g., for prophylaxis). As used herein, the terms "treat," "treating," and "treatment" refer to the administration of an effective amount or dosage.

[0082] The term "effective amount" or "effective dosage" is defined as an amount sufficient to achieve or at least partially achieve a desired effect.

[0083] A "therapeutically effective amount" or "therapeutically effective dose" of a drug or therapeutic agent is any amount of drug that, when used alone or in combination with another therapeutic agent, protects a subject against the onset of the disease or promotes regression of the disease as manifested by a reduction in the severity of disease symptoms, an increase in the frequency and duration of symptom-free periods of the disease, or prevention of impairment or disability due to the affliction of the disease.

[0084] A therapeutically effective amount or dosage of a drug includes a "prophylactically effective amount" or "prophylactically effective dosage," which is any amount of drug that, alone or in combination with another therapeutic agent, when administered to a subject at risk of developing a disease or at risk of having a recurrence of a disease, prevents the development or recurrence of the disease.

[0085] Additionally, the terms "effective" and "effectiveness" with respect to the treatments disclosed herein include both pharmacological effectiveness and physiological safety. Pharmacological effectiveness refers to the ability of a drug to promote cancer regression in a patient. Physiological safety refers to the level of toxicity or other adverse physiological effects at the cellular, organ, and / or organism level resulting from the administration of a drug.

[0086] The ability of a therapeutic agent to promote disease regression, e.g., cancer regression, can be assessed using a variety of methods known to those of skill in the art, such as as human subjects in clinical trials, in animal model systems predictive of efficacy in humans, or by assaying the activity of the agent in vitro.

[0087] For example, an "anti-cancer agent" or combination thereof promotes regression of a cancer in a subject. In some embodiments, a therapeutically effective amount of a therapeutic agent promotes regression of a cancer to the point of removal of the cancer.

[0088] In some embodiments of the present disclosure, the anti-cancer agent is administered as a combination of therapies, for example, therapies including (i) anti-angiogenic therapy, e.g., an anti-VEGF-A antibody such as bevacizumab, and (ii) checkpoint inhibitor therapy, e.g., administration of an antibody against PD-1 or PD-L1.

[0089] "Promoting cancer regression" means that administration of an effective amount of a drug or combination thereof (administered together as a single treatment composition or in separate treatments as separate compositions, as discussed above) results in a reduction in cancer burden, e.g., a decrease in tumor growth or size, tumor necrosis, a decrease in the severity of at least one symptom, an increase in the frequency and duration of disease symptom-free periods, or prevention of disability or disability due to the affliction of the disease.

[0090] Regardless of these ultimate therapeutic efficacy measurements, the evaluation of immunotherapeutic drugs must also take into account immune-related response patterns. The ability of a therapeutic agent to inhibit cancer growth, e.g., tumor growth, can be evaluated using the assays described herein and other assays known in the art. Alternatively, this property of a composition can be evaluated by testing the ability of the compound to inhibit cell proliferation, and such inhibition can be measured in vitro using assays known to those skilled in the art.

[0091] As used herein, "tumor" refers to all neoplastic cell growth and proliferation, and all pre-cancerous and cancerous cells and tissues. In some embodiments, a cancer, such as colorectal cancer, gastric cancer, breast cancer, prostate cancer, liver cancer, head and neck cancer, melanoma, or ovarian cancer, is recurrent. The term "recurrent" refers to a situation in which cancer cells recur in a subject who has experienced remission of cancer (e.g., colorectal cancer) after treatment. In some embodiments, a cancer, such as colorectal cancer, gastric cancer, breast cancer, prostate cancer, liver cancer, head and neck cancer, melanoma, or ovarian cancer, glioma, glioblastoma, or lung cancer, is refractory. As used herein, the terms "refractory" or "resistant" refer to a situation in which a subject has residual cancer cells in their body even after intensive treatment. In some embodiments, cancer, such as colorectal cancer, gastric cancer, breast cancer, prostate cancer, liver cancer, head and neck cancer, melanoma, ovarian cancer, glioma, glioblastoma, or lung cancer, is refractory after at least one prior treatment, including the administration of at least one anticancer agent.In some embodiments, cancer, such as colorectal cancer, gastric cancer, breast cancer, prostate cancer, liver cancer, head and neck cancer, melanoma, ovarian cancer, glioma, glioblastoma, or lung cancer, is metastatic.

[0092] "Cancer" or "cancerous tissue" can include tumors of various stages. In certain embodiments, a cancer or tumor is stage 0, whereby, for example, the cancer or tumor is very early in development and has not metastasized. In some embodiments, a cancer or tumor is stage I, whereby, for example, the cancer or tumor is relatively small in size, has not spread into nearby tissues, and has not metastasized. In other embodiments, a cancer or tumor is stage II, whereby, for example, the cancer or tumor is larger than stage 0 or stage I and has grown into nearby tissues, but has not metastasized, except possibly to the lymph nodes. In other embodiments, a cancer or tumor is stage IV, whereby, for example, the cancer or tumor has metastasized. Stage IV is also referred to as advanced or metastatic cancer.

[0093] As used herein, the term "biological sample" or "sample" means biological material isolated from a subject. A biological sample can include any biological material suitable for measuring gene expression, for example, by nucleic acid sequencing.

[0094] The biological sample can be any suitable biological tissue, such as, for example, cancer tissue. In one embodiment, the sample is a tumor tissue biopsy, such as formalin-fixed, paraffin-embedded (FFPE) tumor tissue or freshly frozen tumor tissue. In another embodiment, an intratumoral sample is used. In another embodiment, a biological fluid may be present in a tumor tissue biopsy, but the biological sample is not itself a biological fluid.

[0095] In some embodiments, the sample, e.g., a biopsy (e.g., a tumor biopsy, a peritumoral biopsy, or a combination thereof), a tissue section, can be obtained from a primary tumor. In some embodiments, the sample, e.g., a biopsy (e.g., a tumor biopsy, a peritumoral biopsy, or a combination thereof), a tissue section, or a tissue sample can be obtained from a metastatic response or a metastatic tumor. In some embodiments, the sample, e.g., a biopsy (e.g., a tumor biopsy, a peritumoral biopsy, or a combination thereof), a tissue section, or a tissue sample can be obtained from any other site other than the original diagnostic site.

[0096] The singular forms "a," "an," and "the" include plural references unless the context clearly dictates otherwise. The terms "a" (or "an"), as well as the terms "one or more" and "at least one," can be used interchangeably herein. In certain embodiments, the terms "a" or "an" mean "single." In certain embodiments, the terms "a" or "an" mean "two or more" or "plural."

[0097] Furthermore, as used herein, "and / or" should be interpreted as a specific disclosure of each of the two specified features or components, with or without the other. Thus, as used herein, the term "and / or" in phrases such as "A and / or B" is intended to include "A and B," "A or B," "A (alone) and B (alone)." Similarly, the term "and / or" in phrases such as "A, B and / or C" encompasses each of the following aspects: A, B, and C; A, B, or C; A or C; A or B; B or C; A and C; A and B; B and C; A (alone); B (alone); and C (alone).

[0098] The terms "about," "comprising essentially," or "consisting essentially of" mean within an acceptable error range for a particular value or composition as determined by one of ordinary skill in the art, which will depend in part on how the value or composition is measured or determined, i.e., the limitations of the measurement system. For example, "about," "comprising essentially," or "consisting essentially of" can mean within one or more standard deviations, in accordance with practice in the art. Alternatively, "about," "comprising essentially," or "consisting essentially of" can mean a range of up to 10%. Moreover, particularly with respect to biological systems or processes, the term can mean ten times the value, or up to five times the value. Where specific values ​​are provided in the specification and claims, unless otherwise specified, "about," "comprising essentially," or "consisting essentially of" should be assumed to be within an acceptable range of error for that particular value or composition.

[0099] As used herein, the term "approximately" as applied to one or more values ​​of interest means a value that is similar to the indicated reference value. In certain embodiments, the term "approximately" refers to a range of values ​​that falls within 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, or more in either direction (greater or less) of the indicated reference value, unless otherwise specified or apparent from the context (except where such number is deemed to exceed 100% of possible values).

[0100] As described herein, any concentration range, percentage range, ratio range, or integer range should be understood to include any integer value within the stated range and, where appropriate, fractions thereof (such as integer tenths and hundredths), unless otherwise specified.

[0101] Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this disclosure pertains.For example, the Concise Dictionary of Biomedicine and Molecular Biology, Juo, Pei-Show, 2nd ed., 2002, CRC Press; The Dictionary of Cell and Molecular Biology, 3rd ed., 1999, Academic Press; and the Oxford Dictionary of Biochemistry and Molecular Biology, Revised, 2000, Oxford University Press provide those skilled in the art with a general dictionary of many terms used in this disclosure.

[0102] Whenever an embodiment is described herein with the term "comprising," it is understood that other similar embodiments described with the terms "consisting of" and / or "consisting essentially of" are also provided.

[0103] Units, prefixes, and symbols are denoted in their Systeme International de Unites (SI) recognized form. The headings provided herein are not limiting of the various aspects of this disclosure but may be used as a reference herein generally. Accordingly, defined terms are more fully defined by reference to the specification.

[0104] Abbreviations used herein are defined throughout this disclosure. Various aspects of the disclosure are described in further detail in the following subsections.

[0105] I. TME Phenotype Class-Specific Cancer Therapy The present disclosure provides methods for classifying cancer patients or cancer tumors into specific tumor microenvironment (TME) phenotype classes, which can be used to guide therapy selection, to determine the cancer patient's eligibility for a particular treatment, or to predict the therapeutic response to a particular treatment, wherein the cancer is selected from the group consisting of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC).

[0106] The TME, also known as the stroma, encompasses the non-malignant components of a tumor, including endothelial cells, smooth muscle cells, pericytes, fibroblasts, immune cells (including lymphoid and myeloid cell types), and supporting and / or connective tissue that characterize the tissue within which a tumor is located, and / or the connective and / or supportive tissue located in or around tumor tissue or entire tumors found in vivo.

[0107] The TME panel 1 classifier of the present disclosure can classify patients with, for example, gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) into TME phenotypic classes that reflect recognized molecular biological characteristics of the disease, as discussed below. The disclosed TME Panel 1 classifier successfully classifies colorectal cancer patients into TME phenotype classes that reflect highly valued molecular biological characteristics of the disease, namely enrichment for angiogenesis and immune processes across disease stage and tumor size. TME Panel 1 identifies similar incidences of TME phenotype classes in colorectal cancer as in gastric cancer, with similar implications for survival. Thus, TME Panel 1 is prognostic, predicting disease-free and overall survival in cancer and the outcome of targeted cancer therapy. When applied to colorectal cancer, TME Panel 1 is consistent with the general annotations of the Consensus Molecular Subtype (CMS) model, with CMS1 as immune enrichment and CMS4 as angiogenesis. However, the CMS group designation does not fully capture these biologies or their interactions. TME Panel 1 was specifically developed to capture these dominant biological processes and obtain more granular predictions regarding the appropriate pairing of targeted therapies to patients.

[0108] The classifiers of the present disclosure, such as the TME Panel 1 classifier, are based on the application of a predictive model generated by machine learning using an artificial neural network (ANN). In some embodiments, the classifiers, such as the TME Panel 1 classifier, are generated using a training set that includes expression data (e.g., RNA expression data) preprocessed according to a population-based classifier as the training set. See U.S. Patent Application No. 17 / 089,234, which is incorporated herein by reference in its entirety.

[0109] Providing an ANN classifier of the present disclosure, e.g., the TME Panel 1 classifier, includes measuring the expression levels (e.g., mRNA expression levels) of a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene panel including a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or any of the gene panels (gene sets) disclosed in Figures 9A-G) in a sample obtained from a cancer patient; and applying a classifier to the measured expression levels. A classifier, e.g., the TME Panel 1 Classifier, assigns a patient's cancer, e.g., gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), to a particular TME phenotype class or combinations thereof.

[0110] The output of an ANN classifier of the present disclosure, e.g., the TME Panel 1 classifier, which subsequently assigns a subject's cancer to a particular TME phenotype class, or combination thereof, should guide the selection and administration of a particular treatment or treatments that have been identified in other subjects to be effective in treating cancers assigned to the same TME phenotype class, i.e., a TME phenotype class-specific therapy or combination thereof, as disclosed below.

[0111] As used herein, the terms "tumor microenvironment" and "TME" refer to the environment surrounding tumor cells, including, for example, blood vessels, immune cells, endothelial cells, fibroblasts, other stromal cells, signaling molecules, and extracellular matrix. In some embodiments, the terms "stromal subtype," "stromal phenotype," and grammatical variants thereof are used synonymously with the term "TME phenotype." As used herein, the term "TME phenotype class" refers to the output of a classifier of the present disclosure, e.g., the TME Panel 1 classifier, that assigns a subject's cancer to a particular TME phenotype.

[0112] Tumor cells and the surrounding microenvironment are closely related and constantly interact. Generally, the tumor microenvironment (also known as the stromal phenotype) encompasses any structural and / or functional characteristics of the tumor and the stroma of the tumor environment. Numerous non-tumor cell types, such as fibroblast-associated carcinoma cells, myeloid-derived suppressor cells, tumor-associated macrophages, neutrophils, or tumor-infiltrating lymphocytes, may exist in the TME. In some embodiments, the classification of a particular TME may include analyzing the cell types present in the stroma. The TME may also be characterized by specific functional properties, such as abnormal oxygenation levels, abnormal vascular permeability, or abnormal levels of specific proteins, such as collagen, elastin, glycosaminoglycans, proteoglycans, or glycoproteins.

[0113] The output of the disclosed ANN classifier, for example, the TME Panel 1 classifier, is a combined biomarker, i.e., it is a biomarker derived from separate biomarkers integrated into a combined signature score, i.e., a combined angiogenesis signature score and an immune signature score. The assignment, non-assignment, interruption, blocking or alteration of a TME phenotype class-specific therapy can be based on the presence, absence, significance or change of a particular TME phenotype class. For example, if a subject has a cancer tumor, e.g., a tumor from gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), classified into an IA TME phenotype class by an ANN classifier of the present disclosure, e.g., the TME panel 1 classifier, an IA TME phenotype class-specific therapy would be administered. In some embodiments, the assignment of a TME phenotype class-specific therapy is based on the absence of a particular TME phenotype; i.e., if a subject has a cancer tumor that is not classified into the IA TME phenotype class by an ANN classifier of the present disclosure, e.g., the TME panel 1 classifier, the IA TME phenotype class-specific therapy will not be administered. Thus, identification of a patient or tumor as belonging to a TME phenotype class can be used to discard possible treatment options. Similarly, identification of a patient or tumor from a patient as belonging to a TME phenotype class that is inconsistent with current therapy can be used to discontinue or remove therapy or modify therapy, e.g., by including or excluding additional therapeutic agents. For example, identification of a mismatch between a patient or tumor TME phenotype class and current therapy can be used to include adjuvant therapy, resulting in a TME phenotype class treatment that can match the patient's TME phenotype classification.

[0114] In some embodiments, the classification of a patient or cancer sample into a TME phenotype class and the assignment of a TME phenotype class-specific therapy to the patient or cancer are not binary. In other words, a patient or cancer sample can be classified as biomarker-positive and / or biomarker-negative for two or more TME phenotype classes, and two or more TME phenotype class therapies, or combinations thereof, can be used to treat the patient. For example, classification of a patient or cancer sample as biomarker-positive for two different TME phenotype classes can be used to select a treatment that includes a combination of pharmacological approaches in a TME phenotype class-specific therapy corresponding to the TME phenotype class for which the patient or cancer sample is biomarker-positive. Furthermore, if a patient or cancer sample is biomarker-negative for a particular TME phenotype class, such knowledge can be used to eliminate certain pharmacological approaches in a TME phenotype class therapy corresponding to the TME phenotype class for which the patient or cancer sample is biomarker-negative. Thus, drugs or combinations thereof, treatments or combinations thereof, and / or clinical regimens or combinations that are useful for treating a particular type of cancer classified as biomarker positive for a particular TME phenotype class can be combined to treat patients with more than one biomarker positive signal, i.e., patients with cancer samples classified as biomarker positive for more than one TME phenotype class.

[0115] In some embodiments, depending on the mechanism of action of a drug or clinical regimen, different classification parameters, such as different gene panel subsets, different thresholds, different ANN configurations, different activation functions, or different post-processing functions, can be used to obtain different TME phenotype classes, which can then be used to select appropriate TME phenotype class-specific therapies. Thus, depending on the mechanism of action of a drug or clinical regimen, different classification parameters, such as different gene panel subsets, different variants of the ANN classifiers disclosed herein, such as the TME Panel 1 classifier, can be developed. Thus, each drug or medication regimen can have a different diagnostic gene panel and a differently configured ANN classifier to provide a clinician, such as a physician, with information to determine whether a patient should be selected for treatment, whether treatment should be initiated, whether treatment should be stopped, or whether treatment should be changed.

[0116] In some embodiments, clinicians can consider covariates of a patient's biomarker status and combine the probabilities of TME phenotype classes with confounding variables such as MSI / MSS (microsatellite instability / microsatellite stability-high) status, EBV (Epstein-Barr virus) status, PD-1 / PD-L1 status (e.g., CPS, i.e., combined positive score), neutrophil-leukocyte ratio (NLR), dMMR status, the presence or absence of mutations in specific molecular biomarkers (e.g., KRAS, NRAS, BRAF), tumor location (e.g., left or right tumor in the case of colorectal cancer), tumor size, tumor shape, tumor surface to volume ratio, invasiveness (e.g., whether cancer cells are present in lymph nodes in the case of breast cancer), or prior treatment history.

[0117] In some aspects, the clinician is provided with a binary result from the classifier and a decision to treat or not treat is made as described herein. In one aspect, the clinician is provided with a plot of the patient's cancer classification results, for example, overlaid on a latent space, and interpreted with a probability threshold or linear or multinomial logistic regression analysis.

[0118] Classification of dMMR tumors into the IA TME phenotype class using an ANN classifier disclosed herein, e.g., TME Panel 1, is associated with improved clinical outcomes with treatment with checkpoint inhibitors. Accordingly, patients with dMMR cancers with the IA TME phenotype can be administered a therapy comprising a checkpoint inhibitor selected from the IA TME phenotype class-specific therapies disclosed below. The present disclosure provides methods of treating patients with dMMR cancers with an IA TME phenotype, the methods comprising administering an IA TME phenotype class-specific therapy to the patient. Also provided are methods of selecting patients for treatment with an IA TME phenotype class-specific therapy when the patient has a dMMR cancer with the IA TME phenotype.

[0119] Classification of dMMR tumors into IS TME phenotype classes using the ANN classifiers disclosed herein, such as TME Panel 1, is associated with improved clinical outcomes with combined treatment with a checkpoint inhibitor and a phosphatidylserine inhibitor. Thus, patients with dMMR cancers with IS TME phenotypes can be administered a combination therapy comprising a checkpoint inhibitor and a phosphatidylserine inhibitor selected from the IS TME phenotype class-specific therapies disclosed below. The present disclosure provides a method for treating patients with dMMR cancers with IS TME phenotypes, comprising administering a combined treatment with a checkpoint inhibitor and a phosphatidylserine inhibitor to the patient. Also provided is a method for selecting a patient for combined treatment with a checkpoint inhibitor and a phosphatidylserine inhibitor when the patient has dMMR cancer with IS TME phenotype.

[0120] Classification of MSI-H tumors into the IA TME phenotype class using the ANN classifiers disclosed herein, e.g., TME Panel 1, is associated with improved clinical outcomes with treatment with checkpoint inhibitors. Accordingly, patients with MSI-H cancers with the IA TME phenotype can be administered a therapy comprising a checkpoint inhibitor selected from the IA TME phenotype class-specific therapies disclosed below. The present disclosure provides methods of treating patients with MSI-H cancers with an IA TME phenotype, the methods comprising administering an IA TME phenotype class-specific therapy to the patient. Also provided are methods of selecting patients for treatment with an IA TME phenotype class-specific therapy when the patient has a dMMR cancer with the IA TME phenotype.

[0121] Classification of MSI-H tumors into the IS TME phenotype class using an ANN classifier disclosed herein, e.g., TME Panel 1, is associated with improved clinical outcomes with combined treatment with a checkpoint inhibitor and a phosphatidylserine inhibitor. Thus, patients with MSI-H cancers with the IS TME phenotype can be administered a combination therapy including a checkpoint inhibitor and a phosphatidylserine inhibitor selected from the IS TME phenotype class-specific therapies disclosed below. The present disclosure provides a method of treating patients with MSI-H cancers with the IS TME phenotype, comprising administering a combined treatment with a checkpoint inhibitor and a phosphatidylserine inhibitor to the patient. Also provided is a method of selecting a patient for combined treatment with a checkpoint inhibitor and a phosphatidylserine inhibitor when the patient has an MSI-H cancer with the IS TME phenotype.

[0122] The methods and components disclosed herein for cancer treatment can be used to treat multiple types of cancer, for example, to identify patients for treatment with specific therapies, predict disease-free probability and overall survival, or predict the outcome of targeted therapy.In some embodiments, the cancer is, for example, gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC).

[0123] In some aspects, the methods and components disclosed herein are used to reduce or shrink the size of a cancer tumor or to effectuate inhibition of cancer tumor growth in a subject in need thereof, such as gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC).

[0124] Classification of metastatic tumors, such as tumors from gastric cancer, breast cancer, prostate cancer, liver cancer, head and neck cancer, melanoma, colorectal cancer, ovarian, glioma, glioblastoma, or lung cancer, into an A or IS TME phenotype class using an ANN classifier disclosed herein, such as TME Panel 1, is associated with improved clinical outcomes with treatment using an angiogenesis inhibitor. Thus, patients with metastatic cancers with an A or IS TME phenotype can be administered a therapy comprising an angiogenesis inhibitor selected from the A TME phenotype class-specific therapies disclosed below. The present disclosure provides a method of treating a patient with metastatic cancer with an A or IS TME phenotype, the method comprising administering to the patient a therapy comprising an angiogenesis inhibitor. Also provided is a method of selecting a patient for treatment with an angiogenesis inhibitor when the patient has metastatic cancer with an A or IS TME phenotype.

[0125] Classification of a tumor, e.g., from gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), into an IA TME phenotype class using an ANN classifier disclosed herein, e.g., TME Panel 1, can be used to select a checkpoint inhibitor, e.g., pembrolizumab, as adjuvant therapy. The present disclosure provides methods of treating a patient having a cancer with an IA TME phenotype, e.g., gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), the method comprising administering a treatment comprising the checkpoint inhibitor pembrolizumab as adjuvant therapy. Also provided is a method of selecting a patient for treatment comprising a checkpoint inhibitor, e.g., pembrolizumab, as adjuvant therapy, where the patient has colorectal cancer with an IA TME phenotype.Classification of a tumor, e.g., a tumor from gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), into a TME phenotype class using an ANN classifier disclosed herein, e.g., TME Panel 1, can be used to select an antiangiogenic therapy, e.g., bevacizumab, as an adjuvant therapy. The present disclosure provides methods of treating a patient having a cancer having an A TME phenotype, e.g., gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), the method comprising administering a treatment comprising an anti-angiogenic therapy, e.g., bevacizumab, as adjuvant therapy. Also provided is a method of selecting a patient for treatment that includes anti-angiogenic therapy, e.g., pembrolizumab, as adjuvant therapy, where the patient has colorectal cancer with an A TME phenotype.

[0126] Classification of a tumor from gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) as having a dominant TME phenotype class using an ANN classifier disclosed herein, e.g., TME Panel 1, can be used, for example, to select a therapy disclosed below that is likely to fit the dominant TME phenotype class. The present disclosure provides methods of treating a patient with gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) having a particular TME phenotype, the method comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the particular TME phenotype class.Also provided is a method of selecting a patient for treatment with a TME phenotype class-specific therapy, where the patient has gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) with a particular TME phenotype, wherein the TME phenotype class-specific therapy is compatible with the particular TME phenotype class.

[0127] The present disclosure provides methods of treating patients with locally advanced / metastatic gastric cancer with a specific TME phenotype class, the methods comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with locally advanced / metastatic gastric cancer to the IA TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with locally advanced / metastatic gastric cancer to the A TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with locally advanced / metastatic gastric cancer to the IS TME phenotype class. In some embodiments, patients with locally advanced / metastatic gastric cancer assigned to the IA TME phenotype class can be treated with a therapy comprising or consisting of immune checkpoint inhibitor therapy (e.g., anti-PD(L)1 therapy). In some embodiments, patients with locally advanced / metastatic gastric cancer assigned to the IA TME phenotype class can be treated with a combination therapy comprising or consisting of chemotherapy, immune checkpoint inhibitor therapy (e.g., anti-PD(L)1 therapy), and nabicikizumab. In some embodiments, patients with locally advanced / metastatic gastric cancer assigned to the IA TME phenotype class can be treated with a combination therapy comprising or consisting of chemotherapy, immune checkpoint inhibitor therapy (e.g., anti-PD(L)1 therapy), and bavituximab. In some embodiments, patients with locally advanced / metastatic gastric cancer assigned to the A TME phenotype class can be treated with a combination therapy comprising or consisting of chemotherapy and antiangiogenic therapy. In some embodiments, patients with locally advanced / metastatic gastric cancer assigned to the A TME phenotype class can be treated with a combination therapy comprising or consisting of chemotherapy, immune checkpoint inhibitor therapy (e.g., anti-PD(L)1 therapy), and nabicikizumab. In some embodiments, patients with locally advanced / metastatic gastric cancer assigned to the IS TME phenotype class can be treated with a combination therapy comprising or consisting of chemotherapy, immune checkpoint inhibitor therapy (e.g., anti-PD(L)1 therapy), and bavituximab.In some embodiments, patients with locally advanced / metastatic gastric cancer assigned to the IS TME phenotype class can be treated with a combination therapy comprising or consisting of chemotherapy, immune checkpoint inhibitor therapy (e.g., anti-PD(L)1 therapy), and nabiquizumab.

[0128] The present disclosure provides methods of treating previously untreated gastric cancer patients with a specific TME phenotype class, the methods comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a previously untreated gastric cancer patient to the IS TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a previously untreated gastric cancer patient to the A TME phenotype class. In some embodiments, previously untreated gastric cancer patients assigned to the IS TME phenotype class can be treated with a combination therapy including or consisting of chemotherapy and antiangiogenic therapy. In some embodiments, previously untreated gastric cancer patients assigned to the A TME phenotype class can be treated with a combination therapy including or consisting of chemotherapy and antiangiogenic therapy.

[0129] The present disclosure provides methods of treating patients with breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer) with a particular TME phenotype class, the method comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the particular TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer) to the A TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer) to the IS TME phenotype class. In some embodiments, patients with breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer) assigned to the A TME phenotype class can be treated with a combination therapy including nabicikizumab. In some embodiments, patients with breast cancer assigned to the IS TME phenotype class (e.g., locally advanced or metastatic HER2-negative breast cancer) can be treated with a combination therapy comprising nabicikizumab. In some embodiments, the combination therapy comprising nabicikizumab comprises or consists of nabicikizumab plus a chemotherapeutic agent (e.g., docetaxel or cabazitaxel). In some embodiments, the combination therapy comprising nabicikizumab comprises or consists of nabicikizumab plus a PARP inhibitor therapy (e.g., rucaparib or olaparib).

[0130] The present disclosure provides methods of treating patients with prostate cancer (e.g., castration-resistant metastatic prostate cancer) with a specific TME phenotype class, the method comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with prostate cancer (e.g., castration-resistant metastatic prostate cancer) to the A TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with prostate cancer (e.g., castration-resistant metastatic prostate cancer) to the IS TME phenotype class. In some embodiments, patients with prostate cancer (e.g., castration-resistant metastatic prostate cancer) assigned to the A TME phenotype class can be treated with a combination therapy including nabi- kizumab. In some embodiments, patients with prostate cancer (e.g., castration-resistant metastatic prostate cancer) assigned to the IS TME phenotype class can be treated with a combination therapy including nabi- kizumab. In some embodiments, the combination therapy comprising nabicikizumab comprises or consists of nabicikizumab plus a chemotherapeutic agent (e.g., docetaxel or cabazitaxel). In some embodiments, the combination therapy comprising nabicikizumab comprises or consists of nabicikizumab plus a PARP inhibitor therapy (e.g., rucaparib or olaparib).

[0131] The present disclosure provides methods of treating patients with liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma) with a specific TME phenotype class, the method comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma) to the IA TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma) to the IS TME phenotype class. In some embodiments, patients with liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma) assigned to the IA TME phenotype class can be treated with a combination therapy comprising or consisting of bavituximab and immune checkpoint inhibitor therapy (e.g., anti-PD(L)1 therapy). In some embodiments, patients with liver cancer assigned to the IS TME phenotype class (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma) can be treated with a combination therapy comprising or consisting of bavituximab and immune checkpoint inhibitor therapy (e.g., anti-PD(L)1 therapy).

[0132] The present disclosure provides methods of treating patients with head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma) with a specific TME phenotype class, the method comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma) to the IA TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma) to the IS TME phenotype class. In some embodiments, a patient with head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma) assigned to the IA TME phenotype class can be treated with a combination therapy comprising or consisting of bavituximab and immune checkpoint inhibitor therapy (e.g., anti-PD(L)1 therapy). In some embodiments, patients with head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma) assigned to the IS TME phenotype class can be treated with a combination therapy comprising or consisting of bavituximab and immune checkpoint inhibitor therapy (e.g., anti-PD(L)1 therapy).

[0133] The present disclosure provides methods of treating a melanoma patient with a specific TME phenotype class, the method comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a melanoma patient to the IA TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a melanoma patient to the IS TME phenotype class. In some embodiments, a melanoma patient assigned to the IA TME phenotype class can be treated with a combination therapy comprising or consisting of bavituximab and radiation therapy. In some embodiments, a melanoma patient assigned to the IS TME phenotype class can be treated with a combination therapy comprising or consisting of bavituximab and radiation therapy.

[0134] The present disclosure provides methods of treating patients with colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver) with a specific TME phenotype class, the methods comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver) to an ID TME phenotype class. In some embodiments, a patient with colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver) assigned to an ID TME phenotype class can be treated with a combination therapy comprising or consisting of nabicizumab, anti-PD(L)1 therapy, and an innate immune stimulator, such as the dectin agonist imprime PGG, the STING agonist BMS-986301, or the NLR agonist BMS-986299.

[0135] The present disclosure provides methods of treating patients with ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer) with a specific TME phenotype class, the method comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer) to the A TME phenotype class. In some embodiments, patients with ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer) assigned to the IS or A TME phenotype class can be treated with a combination therapy comprising or consisting of a PARP inhibitor (olaparib, rucaparib, niraparib, etc.) + immune checkpoint inhibitor therapy (e.g., anti-PD(L)1, i.e., an inhibitor against PD-1 or PD-L1) + nabicikizumab, representing a non-chemotherapeutic treatment option for ovarian cancer.

[0136] The present disclosure provides methods of treating patients with breast cancer (e.g., platinum-resistant or platinum-sensitive recurrent triple-negative breast cancer) with a specific TME phenotype class, the methods comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with breast cancer (e.g., platinum-resistant or platinum-sensitive recurrent triple-negative breast cancer) to an IA, IS, or A TME phenotype class. In some embodiments, patients with breast cancer (e.g., platinum-resistant or platinum-sensitive recurrent triple-negative breast cancer) assigned to the IA, IS, or A TME phenotype class can be treated with a PARP inhibitor, an immune checkpoint inhibitor, and nabicikizumab.

[0137] The present disclosure provides methods of treating a melanoma patient with a specific TME phenotype class, the methods comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME Panel 1 classifier) ​​assigns a tumor sample from a melanoma patient to an IS TME phenotype class. In some embodiments, a melanoma patient assigned to the IS TME phenotype class can be treated with an immunomodulator (such as bidutolimod) and a CPI combination therapy. Examples of additional immunomodulatory agents in this class are ProMune CpG7909 (PF3512676), SD-101, 1018 ISS, IMO-2123, Litenimod, MIS416, Cobitolimod, Imprim PGG (Odetiglucan), Imiquimod, Fingolimod, Tirsotrimod, and BL-7040.

[0138] The present disclosure provides methods of treating patients with colorectal cancer (e.g., metastatic colorectal cancer) with a particular TME phenotype class, the methods comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the particular TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME panel 1 classifier) ​​assigns a tumor sample from a patient with colorectal cancer (e.g., metastatic colorectal cancer) to an A or IS TME phenotype class. In some embodiments, patients with colorectal cancer (e.g., metastatic colorectal cancer) assigned to the A or IS TME phenotype class can be treated with an anti-DLL4 / anti-VEGF antagonist in combination with a chemotherapeutic agent (e.g., forfox, forfiri, or irinotecan).

[0139] The present disclosure provides methods for treating patients with glioma or glioblastoma with a specific TME phenotype class, the methods comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME Panel 1 classifier) ​​assigns tumor samples from patients with glioma or glioblastoma to an IS or IA TME phenotype class. In some embodiments, patients with glioma or glioblastoma assigned to the IS TME phenotype class can be treated with bavituximab, a checkpoint inhibitor, and radiation.

[0140] The present disclosure provides methods for treating patients with non-small cell lung cancer with a specific TME phenotype class, the methods comprising administering a TME phenotype class-specific treatment to the patient, wherein the treatment is compatible with the specific TME phenotype class. In some embodiments, an ANN classifier (e.g., a TME Panel 1 classifier) ​​assigns tumor samples from patients with non-small cell lung cancer to an IS or IA TME phenotype class. In some embodiments, patients with glioma or glioblastoma assigned to an IS or IA TME phenotype class can be treated with a combination therapy of tislelizumab and chemotherapy.

[0141] The present disclosure provides methods of treating tumors that are biomarker positive and comprise an immunoactive (IA), immunosuppressive (IS) and angiogenic (A) phenotype with an anti-DLL4 / anti-VEGF antagonist such as nabicikizumab, ABT-165 or CTX-009, or an anti-VEGF antagonist such as bevacizumab, ramucirumab, or valisacumab, in combination with an anti-PD-1 or anti-PD-L1 checkpoint inhibitor (CPI), or with a bispecific immunoglobulin or modified immunoglobulin of an anti-VEGF antagonist and a CPI.

[0142] The present disclosure provides methods for monitoring disease progression, selecting a particular treatment, selecting patients for treatment, or deciding whether to continue or discontinue treatment, the methods comprising the steps of (i) treating immune desert (ID) patients with investigator-selected standard of care chemotherapeutic agents and / or vaccines, the latter being selected from, for example, AST-301 (pNGVL3-hICD), NeoVax, Proscavax, personalized vaccines, alpha-lactalbumin vaccine, P10s-PADRE, OncoVax, PVX-410, Galinpepimut-S, GRT-C903 / GRT -C904, KRAS peptide vaccine, pING-hHER3FL, GVAX, INCAGN01876, or non-genetically engineered, live immune cell immunotherapy (non-limiting example is AlloStim); (ii) taking a biopsy, for example, two months after treatment; and (iii) re-evaluating the patient's TME panel 1 status, wherein patients in a transition state from ID to IA are treated with and respond to immunotherapy, and patients remaining in the ID group are spared less effective therapies, such as immunotherapy or anti-angiogenic therapy, to which they are unlikely to respond.

[0143] The present disclosure also provides stratification strategies that include predetermined randomization ratios or prioritization biomarkers. In some embodiments, the predetermined randomization ratio uses an inverse incidence ratio, where patients with low-incidence biomarkers are more likely to be assigned to substudies for lower-incidence populations. In some embodiments, the biomarker prioritization approach involves ranking biomarker groups based on their predictive value and assigning patients to the treatment group whose biomarker profile has the highest predictive value. In some embodiments, TME phenotype or biomarker status (i.e., IA, IS, ID, A, A+IA, A+IS, or biomarker positive) is prioritized over other biomarkers or used in combination with other biomarkers, such as MSS status or PD-L1.

[0144] The present disclosure provides methods for classifying / stratifying cancer patients and / or cancer or tumor samples from those patients according to a TME phenotyping obtained from application of a combined biomarker, e.g., an ANN classifier derived from a gene expression dataset corresponding to a gene panel, wherein the cancer is, for example, gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC). In some embodiments, the ANN classifier is a TME Panel 1 classifier.

[0145] In one aspect, the present disclosure provides a method of treating a human subject afflicted with cancer, e.g., gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), the method comprising administering to a subject a therapeutically effective amount of IA-1000 or IA-10 ... The method includes administering a TME phenotype class-specific therapy to a subject, wherein prior to administration, an ANN classifier disclosed herein, e.g., TME Panel 1, is applied to a dataset including RNA expression levels of a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or a gene panel including any of the gene panels (gene sets) disclosed in Figures 9A-G) in a tumor sample obtained from the subject, and the ANN classifier assigns the tumor sample to an IA TME phenotype class.

[0146] The present disclosure also relates to the treatment of cancer, for example, gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and provides a method of treating a human subject suffering from lung cancer (e.g., NSCLC), the method comprising: (A) prior to administration, identifying the subject as exhibiting an IA TME phenotype as determined by measuring RNA expression levels of a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or a gene panel (gene set) including any of the gene panels (gene sets) disclosed in Figures 9A-G) via an ANN disclosed herein, e.g., TME Panel 1; and (B) administering an IA TME phenotype class-specific therapy to the subject.

[0147] In some embodiments, an IA TME phenotype class-specific therapy can be administered in combination with an additional TME phenotype class-specific therapy disclosed herein if the subject is biomarker positive for the additional TME phenotype.

[0148] Also provided are cancers suitable for treatment with IA TME phenotype class-specific therapies, such as gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), and the like. and a gene panel comprising at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3, and a gene set from Table 4, a gene panel from Table 5, or any of the gene panels (gene sets) disclosed in Figures 9A-G in the disclosure of a method for identifying a human subject suffering from IA TME phenotype class, glioblastoma, or lung cancer (e.g., NSCLC), the method comprising applying an ANN classifier disclosed herein, e.g., TME Panel 1, to RNA expression levels of a gene panel (e.g., a gene panel including ...

[0149] In some embodiments, the IA TME phenotype class-specific therapy comprises a checkpoint modulator therapy.

[0150] In some embodiments, checkpoint modulator therapy comprises administering an activator of an inhibitory immune checkpoint molecule. In some embodiments, the activator of a stimulatory immune checkpoint molecule is, for example, an antibody molecule against GITR (glucocorticoid-inducible tumor necrosis factor receptor, TNFRSF18), OX-40 (TNFRSF4, ACT35, CD134, IMD16, TXGP1L, tumor necrosis factor receptor superfamily member 4, TNF receptor superfamily member 4), ICOS (inducible T cell costimulatory molecule), 4-1BB (TNFRSF9, CD137, CDw137, ILA, tumor necrosis factor receptor superfamily member 9, TNF receptor superfamily member 9), or a combination thereof. In some embodiments, checkpoint modulator therapy comprises administering an RORy (RORC, NR1F3, RORG, RZR-GAMMA, RZRG, TOR, RAR-related orphan receptor gamma, IMD42, RAR-related orphan receptor C) agonist.

[0151] In some embodiments, checkpoint modulator therapy involves the administration of an inhibitor, modulator, agonist, or antagonist of an inhibitory immune checkpoint molecule. As used herein, the term "modulator" refers to a molecule that directly or indirectly interacts with a target and affects a biological or chemical process or mechanism. For example, a modulator can increase, promote, upregulate, activate, suppress, reduce, block, prevent, slow, desensitize, inactivate, downregulate, etc. a biological or chemical process or mechanism. Thus, a modulator can be an "agonist" or "antagonist" of a target. The term "agonist" refers to a compound that enhances at least some of the effect of an endogenous ligand, such as a protein, receptor, enzyme, etc. The term "antagonist" refers to a compound that suppresses at least some of the effect of an endogenous ligand, such as a protein, receptor, enzyme, etc.

[0152] In some embodiments, the inhibitor of an inhibitory immune checkpoint molecule is, for example, an antibody against PD-1 (PDCD1, CD279, SLEB2, hPD-1, hPD-l, hSLE1, programmed cell death 1), e.g., sintilimab tislelizumab, pembrolizumab, or an antigen-binding portion thereof, a ligand against PD-L1 (CD274, B7-H, B7H1, PDCD1L1, PDCD1LG1, PDL1, CD274 molecule, programmed cell death ligand 1, hPD-L1), Antibodies against PD-L2 (PDCD1LG2, B7DC, Btdc, CD273, PDCD1L2, PDL2, bA574F11.2, programmed cell death 1 ligand 2), antibodies against CTLA-4 (CTLA4, ALPS5, CD, CD152, CELIAC3, GRD4, GSE, IDDM12, cytotoxic T-lymphocyte-associated protein 4), bispecific antibodies containing at least binding specificity for PD-L1, PD-L2, or CTLA-4, either alone or in combination.

[0153] In some embodiments, the inhibitor of an inhibitory immune checkpoint molecule is selected from the group consisting of, for example, TIM-3 (T-cell immunoglobulin and mucin domain containing-3), LAG-3 (lymphocyte activation gene 3), BTLA (B- and T-lymphocyte attenuator), TIGIT (T-cell immunoreceptor with Ig and ITIM domains), VISTA (V-domain Ig inhibitor of T-cell activation), TGFβ (transforming growth factor beta) or its receptor, CD86 (cluster of differentiation 86) agonist, LAIR1 (leukocyte-associated immunoglobulin-like receptor 1), CD160 (cluster of differentiation 160), 2B4 (natural killer cell receptor 2B4; cluster of differentiation 244), GITR, OX40, 4-1BB (CD137), CD2 (cluster of differentiation 2), CD27 (cluster of differentiation 27), CDS (CDP-diacylglycerol synthase 1), ICA and any of the above disclosed antibodies in combination with an inhibitor, modulator, antagonist, or agonist of M-1 (intercellular adhesion molecule 1), LFA-1 (lymphocyte function-associated antigen 1: CD11a / CD18), ICOS (inducible T-cell costimulatory molecule; CD278), CD30 (cluster of differentiation 30), CD40 (cluster of differentiation 40), BAFFR (B-cell activating factor receptor), HVEM (herpes virus entry mediator), CD7 (cluster of differentiation 7), LIGHT (tumor necrosis factor superfamily member 14; TNFSF14), NKG2C (killer cell lectin-like receptor C2; KLRC2, CD159c), SLAMF7 (SLAM family member 7), NKp80 (activating co-receptor NKp80; lectin-like receptor F1; KLRF1; killer cell lectin-like receptor F1), or any combination thereof.

[0154] In some embodiments, the anti-PD-1 antibody comprises nivolumab, pembrolizumab, cemiplimab, sintilimab, tislelizumab, or an antigen-binding portion thereof. In some embodiments, the anti-PD-1 antibody cross-competes with nivolumab, pembrolizumab, cemiplimab, sintilimab, or tislelizumab for binding to human PD-1, or binds to the same epitope as nivolumab, pembrolizumab, cemiplimab, sintilimab, or tislelizumab.

[0155] In some embodiments, the anti-PD-L1 antibody comprises avelumab, atezolizumab, durvalumab, or an antigen-binding portion thereof. In some embodiments, the anti-PD-1 antibody cross-competes with avelumab, atezolizumab, or durvalumab for binding to human PD-1, or binds to the same epitope as avelumab, atezolizumab, or durvalumab.

[0156] In some embodiments, the checkpoint modulator therapy comprises administration of: (i) an anti-PD-1 antibody, e.g., an antibody selected from the group consisting of nivolumab, pembrolizumab, sintilimab, tislelizumab, and cemiplimab; (ii) an anti-PD-L1 antibody, e.g., an antibody selected from the group consisting of avelumab, atezolizumab, and durvalumab; or (iii) a combination thereof.

[0157] In one aspect, the present disclosure provides a method of treating a human subject afflicted with cancer, e.g., gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), the method comprising administering to a human subject an ischemic heart failure or other life-threatening condition. The method includes administering a TME phenotype class-specific therapy to a subject, wherein prior to administration, an ANN classifier disclosed herein, e.g., TME Panel 1, is applied to a dataset including RNA expression levels of a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or a gene panel including any of the gene panels (gene sets) disclosed in Figures 9A-G) in a tumor sample obtained from the subject, and the ANN classifier assigns the tumor sample to an IS TME phenotype class.

[0158] In one aspect, the present disclosure provides a method of treating a human subject afflicted with cancer, e.g., gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), the method comprising administering to a human subject an ischemic heart failure or other life-threatening condition. The method includes administering a TME phenotype class-specific therapy to a subject, wherein prior to administration, an ANN classifier disclosed herein, e.g., TME Panel 1, is applied to a dataset including RNA expression levels of a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or a gene panel including any of the gene panels (gene sets) disclosed in Figures 9A-G) in a tumor sample obtained from the subject, and the ANN classifier assigns the tumor sample to an IS TME phenotype class.

[0159] The present disclosure also relates to the treatment of cancer, for example, gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and provides a method of treating a human subject suffering from lung cancer (e.g., NSCLC), the method comprising: (A) prior to administration, identifying the subject as exhibiting an IS TME phenotype as determined by measuring the RNA expression levels of a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or a gene panel (gene set) including any of the gene panels (gene sets) disclosed in Figures 9A-G) via an ANN disclosed herein, e.g., TME Panel 1; and (B) administering an IS TME phenotype class-specific therapy to the subject.

[0160] In some embodiments, an IS TME phenotype class-specific therapy can be administered in combination with an additional TME phenotype class-specific therapy disclosed herein if the subject is biomarker positive for the additional TME phenotype.

[0161] Also provided are gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and other cancers suitable for treatment with IS TME phenotype class-specific therapy. and lung cancer (e.g., NSCLC), the method comprising applying an ANN classifier disclosed herein, e.g., TME panel 1, to RNA expression levels of a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or a gene panel including any of the gene panels (gene sets) disclosed in Figures 9A-G) in a sample obtained from a tumor from the subject, wherein classification of the tumor into an IS TME phenotype class indicates that an IS TME phenotype class-specific therapy can be administered to the subject to treat the cancer.

[0162] In some embodiments, the IS TME phenotype class-specific therapy includes, for example, (1) the administration of a checkpoint modulator therapy and an anti-immunosuppressive therapy (e.g., a combination therapy including the administration of pembrolizumab and bavituximab), and / or (2) an anti-angiogenic therapy. In some embodiments, the checkpoint modulator therapy includes, for example, the administration of an inhibitor of an inhibitory immune checkpoint molecule. In some embodiments, the inhibitor of an inhibitory immune checkpoint molecule is, for example, an antibody against PD-1 (e.g., sintilimab, tislelizumab, pembrolizumab, or an antigen-binding portion thereof), an antibody against PD-L1, PD-L2, CTLA-4, or a combination thereof.

[0163] In some embodiments, the anti-PD-1 antibody includes, for example, nivolumab, pembrolizumab, cemiplimab, spartalizumab (PDR001), sintilimab, tislelizumab, or geptanolimab (CBT-501), or an antigen-binding portion thereof. In some embodiments, the anti-PD-1 antibody cross-competes with nivolumab, pembrolizumab, cemiplimab, PDR001, sintilimab, tislelizumab, or CBT-501 for binding to human PD-1, or binds to the same epitope as nivolumab, pembrolizumab, cemiplimab, sintilimab, tislelizumab, PDR001, or CBT-501.

[0164] In some embodiments, the anti-PD-L1 antibody includes, for example, avelumab, atezolizumab, durvalumab, or an antigen-binding portion thereof. In some embodiments, the anti-PD-L1 antibody cross-competes with avelumab, atezolizumab, or durvalumab for binding to human PD-L1, or binds to the same epitope as avelumab, atezolizumab, or durvalumab.

[0165] In some embodiments, the anti-CTLA-4 antibody comprises ipilimumab or an antigen-binding portion thereof. In some embodiments, the anti-CTLA-4 antibody cross-competes with ipilimumab for binding to CTLA-4 or binds to the same epitope as ipilimumab.

[0166] In some embodiments, the checkpoint modulator therapy comprises, for example, administration of (i) an anti-PD-1 antibody, e.g., selected from the group consisting of nivolumab, pembrolizumab, sintilimab, tislelizumab, and cemiplimab; (ii) an anti-PD-L1 antibody, e.g., selected from the group consisting of avelumab, atezolizumab, and durvalumab; or (iii) an anti-CTLA-4 antibody, e.g., ipilimumab, or (iv) a combination thereof.

[0167] In some embodiments, the anti-angiogenic therapy includes, for example, an anti-VEGF (vascular endothelial growth factor) antibody selected from the group consisting of valisacumab, bevacizumab, nabicikizumab (anti-DLL4 / anti-VEGF bispecific antibody), and combinations thereof. In some embodiments, the anti-angiogenic therapy includes, for example, administration of an anti-VEGFR antibody. In some embodiments, the anti-VEGFR antibody is an anti-VEGFR2 vascular endothelial growth factor receptor 2 antibody. In some embodiments, the anti-VEGFR2 antibody includes ramucirumab. In some embodiments, the anti-angiogenic therapy includes, for example, nabicikizumab, ABL101 (NOV1501), or zilpakimab (ABT165).

[0168] In some embodiments, the anti-immunosuppressive therapy comprises administering an anti-PS (phosphatidylserine) antibody, an anti-PS targeting antibody, an antibody that binds to β2 glycoprotein 1, an inhibitor of PI3Kγ (phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit gamma isoform), an adenosine pathway inhibitor, an inhibitor of IDO, an inhibitor of TIM, an inhibitor of LAG3, an inhibitor of TGFβ, a CD47 inhibitor, or a combination thereof.

[0169] In some embodiments, the anti-PS targeting antibody is, for example, bavituximab, 1N11, or an antibody that binds to beta2 glycoprotein 1 (β2GP1, or Apo-H). In some embodiments, the anti-PS targeting antibody is bavituximab. In some embodiments, the anti-PS targeting antibody is an antibody that binds to beta2 glycoprotein 1 (β2GP1, or Apo-H). In some embodiments, the anti-PS targeting antibody is 1N11. See, e.g., Schad et al. (2020) J. Immunol. 201(S1):170.5; Yin et al. (2009) Cancer Res. 69(S9):5463; Zohar & Shoenfeld (2018) Immunotargets Ther. 7:51-53, all of which are incorporated herein by reference in their entireties.

[0170] In some embodiments, the PI3Kγ inhibitor is, for example, LY3023414 (samotricisib) or IPI-549 (eganelisib). In some embodiments, the adenosine pathway inhibitor is, for example, AB-928. In some embodiments, the TGFβ inhibitor is, for example, LY2157299 (galunisertib) or the TGFβR1 inhibitor LY3200882. In some embodiments, the CD47 inhibitor is, for example, magrolimab (5F9). In some embodiments, the CD47 inhibitor targets SIRPα.

[0171] In some embodiments, the anti-immunosuppressive therapy comprises administration of an inhibitor, modulator, agonist or antagonist of TIM-3, LAG-3, BTLA, TIGIT, VISTA, TGF-β or its receptor, CD86, LAIR1, CD160, 2B4, GITR, OX40, 4-1BB (CD137), CD2, CD27, CDS, ICAM-1, LFA-1 (CD11a / CD18), ICOS (CD278), CD30, CD40, BAFFR, HVEM, CD7, LIGHT, NKG2C, SLAMF7, inhibitor of NKp80, or a combination thereof.

[0172] In one aspect, the present disclosure provides a method of treating a human subject afflicted with cancer selected from the group consisting of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and lung cancer (e.g., NSCLC), the method comprising: The method includes administering a TME phenotype class-specific therapy to a subject, wherein prior to administration, an ANN classifier disclosed herein, e.g., TME Panel 1, is applied to a dataset including RNA expression levels of a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or a gene panel including any of the gene panels (gene sets) disclosed in Figures 9A-G) in a tumor sample obtained from the subject, and the ANN classifier assigns the tumor sample to a TME phenotype class A.

[0173] The present disclosure also provides methods for treating gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and lung cancer (e.g., NSCLC). The present invention provides a method for treating a human subject suffering from a cancer selected from the group consisting of: SCLC (Cancer A, Cancer B, Cancer C, Cancer D, Cancer E, Cancer F, Cancer I, Cancer L, Cancer L) comprising: (A) identifying the subject exhibiting an A TME phenotype as determined by measuring RNA expression levels of a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or a gene panel (gene set) including any of the gene panels (gene sets) disclosed in Figures 9A-G) in a sample obtained from the subject via an ANN disclosed herein, e.g., TME Panel 1, prior to administration; and (B) administering an A TME phenotype class-specific therapy to the subject.

[0174] In some embodiments, a TME phenotype class-specific therapy can be administered in combination with an additional TME phenotype class-specific therapy disclosed herein if the subject is biomarker positive for the additional TME phenotype.

[0175] Also provided are gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and the like, which are suitable for treatment with a TME phenotype class-specific therapy. and lung cancer (e.g., NSCLC), the method comprising applying an ANN classifier disclosed herein, e.g., TME Panel 1, to RNA expression levels of a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or a gene panel including any of the gene panels (gene sets) disclosed in Figures 9A-G) in a sample obtained from a tumor from the subject, wherein classification of the tumor into an A TME phenotype class indicates that an A TME phenotype class-specific therapy can be administered to the subject to treat the cancer.

[0176] In some embodiments, A TME phenotype class-specific therapies include VEGF-targeted therapies and other anti-angiogenic agents, angiopoietin 1 and 2 (Ang1 and Ang2), DLL4 (Delta-like canonical Notch ligand 4), anti-VEGF and anti-DLL4 bispecifics, TKIs (tyrosine kinase inhibitors) such as fruquintinib, anti-FGF (fibroblast growth factor) antibodies, and antibodies or small molecules that inhibit the FGF receptor family (FGFR1 and FGFR2); anti-PLGF (placental growth factor) antibodies and small molecules and antibodies against PLGF receptors, anti-VEGF-B (vascular endothelial growth factor B) antibodies, anti-VEGF-C (vascular endothelial growth factor C) antibodies, anti-VEGF-D (vascular endothelial growth factor D); antibodies against VEGF / PLGF trapping molecules such as aflibercept, or ziv-aflibercept; anti-Notch therapies such as anti-DLL4 antibodies or inhibitors of gamma-secretase. In some embodiments, the anti-angiogenic therapy comprises administration of an antagonist to endoglin, such as carotuximab (TRC105).

[0177] As used herein, the term "VEGF targeted therapy" means targeting a ligand, i.e., VEGF-A (vascular endothelial growth factor A), VEGF-B (vascular endothelial growth factor B), VEGF-C (vascular endothelial growth factor C), VEGF-D (vascular endothelial growth factor D), or PLGF (placental growth factor); a receptor, e.g., VEGFR1 (vascular endothelial growth factor receptor 1), VEGFR2 (vascular endothelial growth factor receptor 2), or VEGFR3 (vascular endothelial growth factor 3); or any combination thereof.

[0178] In some embodiments, the VEGF-targeted therapy comprises administering an anti-VEGF antibody or an antigen-binding portion thereof. In some embodiments, the anti-VEGF antibody comprises, for example, valisacumab, bevacizumab, or an antigen-binding portion thereof. In some embodiments, the anti-VEGF antibody cross-competes with valisacumab or bevacizumab for binding to human VEGF-A, or binds to the same epitope as valisacumab or bevacizumab.

[0179] In some embodiments, the VEGF-targeted therapy comprises administration of an anti-VEGFR antibody. In some embodiments, the anti-VEGFR antibody is an anti-VEGFR2 antibody. In some embodiments, the anti-VEGFR2 antibody comprises ramucirumab or an antigen-binding portion thereof.

[0180] In some embodiments, the A TME phenotype class-specific therapy comprises administration of angiopoietin / TIE2 (TEK receptor tyrosine kinase; CDC202B) targeted therapy. In some embodiments, the angiopoietin / TIE2 targeted therapy comprises administration of endoglin and / or angiopoietin. In some embodiments, the A TME phenotype class-specific therapy comprises administration of DLL4 targeted therapy. In some embodiments, the DLL4 targeted therapy comprises administration of nabicikizumab, ABL101 (NOV1501), or ABT165.

[0181] In all of the methods disclosed above, for example, in methods of treating a subject with gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), or in methods of treating a subject with gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic In a method of selecting a subject with a cancer selected from the group consisting of prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) for treatment with a TME phenotype class-specific therapy (the TME phenotype class-specific therapy is selected according to the classification of the cancer TME phenotype into one or more TME phenotype classes using an ANN classifier disclosed herein, e.g., TME panel 1), administration of the specific therapy, e.g., a TME phenotype class-specific therapy disclosed herein, or a combination thereof, can effectively treat the cancer.

[0182] In some embodiments, administration of a TME phenotype class-specific therapy disclosed herein, or a combination thereof, to a subject with a cancer selected from the group consisting of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) reduces the cancer burden.

[0183] In some embodiments, administration of a TME phenotype class-specific therapy disclosed herein or a combination thereof to a subject with colorectal cancer reduces the cancer burden by at least about 10%, at least about 15%, at least about 20%, at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 45%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, or about 100% compared to the cancer burden before administration of the therapy.

[0184] In some embodiments, administration of a TME phenotype class-specific therapy or combination thereof disclosed herein results in progression-free survival of at least about 1 month, at least about 2 months, at least about 3 months, at least about 4 months, at least about 5 months, at least about 6 months, at least about 7 months, at least about 8 months, at least about 9 months, at least about 10 months, at least about 11 months, at least about 1 year, at least about 18 months, at least about 2 years, at least about 3 years, at least about 4 years, or at least about 5 years after initial administration of the therapy.

[0185] In some embodiments, after administration of a TME phenotype class-specific therapy or a combination thereof, the subject exhibits stable disease. The term "stable disease" refers to a diagnosis of the presence of cancer, such as gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), but the cancer has been treated and remains stable, i.e., is not progressing, as determined, for example, by imaging data and / or best clinical judgment.

[0186] The term "progression" refers to a diagnosis of the presence of high activity of cancer, e.g., gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), i.e., cancer that is untreated, is not stable, or has been treated and has not responded to therapy, or has been treated and there is residual active disease, as determined by imaging data and / or best clinical judgment.

[0187] "Stable disease" can encompass (temporary) tumor shrinkage / reduction in tumor volume during the course of treatment compared to the initial tumor volume at the start of treatment (i.e., before treatment). In this context, "tumor shrinkage" can refer to a decrease in tumor volume during treatment compared to the initial volume at the start of treatment (i.e., before treatment). For example, a tumor volume of less than 100% (e.g., about 99% to about 66% of the initial volume at the start of treatment) can represent "stable disease."

[0188] Alternatively, "stable tumor" may encompass (temporary) tumor growth / increase in tumor volume during the course of treatment compared to the initial (i.e., pre-treatment) tumor volume at the start of treatment. In this case, "tumor growth" may refer to an increase in tumor volume at the time of treatment compared to the initial volume at the start of treatment (i.e., before treatment). For example, a tumor volume of greater than 100% (e.g., about 101% to about 135% of the initial volume at the start of treatment, preferably about 101% to about 110% of the initial volume) may represent "stable tumor."

[0189] The term "stable" can include the following aspects: for example, tumor volume does not shrink after treatment (i.e., tumor growth stops), or, for example, tumor shrinks at the start of treatment but does not continue to shrink until the tumor disappears, i.e., tumor growth initially reverses but then grows again before the tumor is, for example, less than 65% of its initial volume.

[0190] When used in reference to a patient or tumor to a TME phenotype class-specific therapy or combination thereof disclosed herein, the term "response" can be expressed as a "complete response" or a "partial response" of the patient or tumor. As used herein, the term "complete response" can refer to the disappearance of all signs of cancer in response to a TME phenotype class-specific therapy or combination thereof disclosed herein. The terms "complete response" and "complete remission" can be used interchangeably. For example, a "complete response" can be reflected by the continued shrinkage of a tumor until it disappears (as shown in the accompanying examples). For example, a tumor volume of 0% compared to the initial tumor volume (100%) at the start of treatment (i.e., before treatment) can represent a "complete response."

[0191] Treatment with a TME phenotype class-specific therapy or combination thereof disclosed herein may result in a "partial response" (or partial remission; e.g., a reduction in tumor size or the spread of cancer throughout the body). A "partial response" may encompass a (temporary) tumor shrinkage / reduction in tumor volume during the course of treatment compared to the initial tumor volume at the start of treatment (i.e., before treatment). Thus, in some embodiments, after administration of a TME phenotype class-specific therapy or combination thereof disclosed herein, the subject exhibits a partial response. In other embodiments, after administration of a TME phenotype class-specific therapy or combination thereof disclosed herein, the subject exhibits a complete response.

[0192] The term "response" can refer to "tumor shrinkage." Thus, administration of a TME phenotype class-specific therapy or combination thereof disclosed herein to a subject in need thereof can result in a decrease in volume or shrinkage of the tumor.

[0193] In some embodiments, after administration of a TME phenotype class-specific therapy or combination thereof disclosed herein, the tumor can be reduced in size by at least about 5%, at least about 10%, at least about 15%, at least about 20%, at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 45%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, or about 100% compared to the tumor volume before treatment.

[0194] In some embodiments, after administration of a TME phenotype class-specific therapy or combination thereof disclosed herein, the volume of the tumor is at least about 5%, at least about 10%, at least about 15%, at least about 20%, at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 45%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, or at least about 90% of the original volume of the tumor before treatment.

[0195] In some embodiments, administration of a TME phenotype class-specific therapy or combination thereof disclosed herein can reduce the growth rate of a tumor by at least about 5%, at least about 10%, at least about 15%, at least about 20%, at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 45%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, or about 100% compared to the growth rate of the tumor before treatment.

[0196] The term "response" can also mean a reduction in the number of tumors if the cancer has metastasized.

[0197] In some embodiments, administration of a TME phenotype class-specific therapy disclosed herein or a combination thereof increases the subject's probability of progression-free survival by at least about 10%, at least about 15%, at least about 20%, at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 45%, compared to the probability of progression-free survival of a subject not exhibiting the TME phenotype or a subject not treated with a specific therapy disclosed herein, e.g., a TME phenotype class-specific therapy disclosed herein or a combination thereof. , at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, at least about 100%, at least about 105%, at least about 110%, at least about 115%, at least about 120%, at least about 125%, at least about 130%, at least about 135%, at least about 140%, at least about 145%, or at least about 150% improvement.

[0198] In some embodiments, administration of a TME phenotype class-specific therapy or combination thereof disclosed herein increases the probability of overall survival by at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 45%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, at least about 100%, at least about 110%, at least about 120%, at least about 125%, at least about 130%, at least about 140%, or at least about 150%, at least about 160%, at least about 170%, at least about 175%, at least about 180%, at least about 190%, at least about 200%, at least about 210%, at least about 220%, at least about 225%, at least about 230%, at least about 240%, at least about 250%, at least about 260%, at least about 270%, at least about 275%, at least about 280%, at least about 290%, at least about 300%, at least about 310%, at least about 320%, at least about 325%, at least about 330%, at least about 340%, at least about 350%, at least about 360%, at least about 370%, at least about 375%, at least about 380%, at least about 390%, or at least about 400% improvement.

[0199] The present disclosure also provides a method for classifying a tumor in a cancer patient, e.g., a tumor of a cancer selected from the group consisting of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and lung cancer (e.g., NSCLC), as a specific TME via an ANN classifier disclosed herein, e.g., TME Panel 1. A gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or any of the gene panels (gene sets) disclosed in Figures 9A-G) is provided for use in assigning a tumor to a TME phenotype class, wherein the assignment or non-assignment of a tumor to a particular TME phenotype class, or combinations thereof, is used to (i) identify the patient as suitable for anti-cancer therapy; (ii) determine the prognosis of the patient receiving anti-cancer therapy; (iii) initiate, discontinue, or modify the administration of anti-cancer therapy to the patient; or (iv) any combination thereof. In some embodiments, the gene panel is used in accordance with the methods disclosed herein, for example, to classify a colorectal cancer from a patient (e.g., to determine whether the tumor is biomarker-positive or biomarker-negative for a TME phenotype class or combinations thereof disclosed herein) and administer a specific therapy (e.g., a TME phenotype class-specific therapy or combinations thereof) based on the classification.

[0200] The present disclosure also provides, via an ANN classifier, e.g., TME Panel 1, classification of cancers suitable for treatment with anti-cancer therapy, such as gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and lung cancer (e.g., The present invention provides a method for identifying a human subject suffering from a cancer selected from the group consisting of: NSCLC (e.g., NSCLC), wherein the TME phenotype class of the cancer is determined by measuring the expression levels, e.g., mRNA expression levels, of genes in a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or any of the gene panels (gene sets) disclosed in Figures 9A-G) in a sample obtained from the subject, and wherein (a) if the assigned TME phenotype class is IA, the therapy is an IA TME phenotype class-specific therapy; (b) if the assigned TME phenotype is IS, the therapy is an IS TME phenotype class-specific therapy; (c) if the assigned TME phenotype is ID, the therapy is an ID TME phenotype class-specific therapy; or (d) if the assigned TME phenotype is A, the therapy is an A TME phenotype class-specific therapy.

[0201] In some embodiments, for example, if a subject is identified as biomarker positive or biomarker negative for two or more TME phenotype classes disclosed herein via an ANN classifier disclosed herein, e.g., TME Panel 1, for example, if the subject is biomarker positive for IA and IS TME phenotype classes, the subject may be administered a combination therapy corresponding to the TME phenotype class-specific therapy corresponding to the TME phenotype classes for which the subject is biomarker positive, e.g., a combination therapy including an IA TME phenotype class-specific therapy and an IS TME phenotype class-specific therapy.

[0202] The present disclosure also provides a method for treating cancer in a human subject in need of treatment, for example, a cancer selected from the group consisting of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and lung cancer (e.g., NSCLC). The present invention provides an anti-cancer therapy comprising: a subject identified as exhibiting or not exhibiting a particular TME phenotype as determined via an ANN classifier disclosed herein, e.g., a TME panel 1 classifier, by measuring the expression levels, e.g., mRNA expression levels, of genes in a gene panel (e.g., a gene panel including at least one gene from Table 1 and one gene from Table 2, a gene set from Table 3 and a gene set from Table 4, a gene panel from Table 5, or a gene panel including any of the gene panels (gene sets) disclosed in Figures 9A-G) of a sample obtained from the subject; and (a) if the assigned TME phenotype class is IA, then the therapy is an IA TME phenotype class-specific therapy; (b) if the assigned TME phenotype class is IS, then the therapy is an IS TME phenotype class-specific therapy; or (c) if the assigned TME phenotype class is A, then the therapy is an A TME phenotype class-specific therapy. In some embodiments, if a patient is biomarker positive for more than one TME phenotype class, the patient can receive a combination of TME phenotype class-specific therapies corresponding to each of the TME phenotype classes for which the patient is biomarker positive.

[0203] In some embodiments, the term "administering" also includes initiating therapy, interrupting or discontinuing therapy, temporarily discontinuing therapy, or modifying therapy (e.g., increasing the amount or frequency of administration, or adding more of one of the therapeutic agents in a combination therapy).

[0204] In some embodiments, the sample can be obtained and / or processed by the same or a different healthcare provider (nurse, hospital), or clinical laboratory, as may be requested by, for example, a healthcare provider (e.g., a doctor) or healthcare benefit provider, and after processing, the results can be sent back to the original healthcare provider or to yet another healthcare provider, healthcare benefit provider, or patient. Similarly, quantification of expression levels of biomarkers disclosed herein; comparisons between biomarker scores or protein expression levels; assessment of the absence or presence of biomarkers; determination of biomarker levels relative to a particular threshold; medical judgment; or combinations thereof can be performed by one or more healthcare providers, healthcare benefit providers, and / or clinical laboratories.

[0205] As used herein, the term "healthcare provider" means an individual or institution that directly interacts with and administers to living organisms, e.g., human patients. Non-limiting examples of healthcare providers include doctors, nurses, technicians, therapists, pharmacists, counselors, alternative medicine practitioners, medical facilities, clinics, hospitals, emergency rooms, doctors' offices, urgent care centers, alternative medicine clinics / facilities, and any other entity that provides general and / or specialty care, evaluation, maintenance, treatment, medication, and / or advice related to all or any part of a patient's health condition, including general medical, specialty medical, surgical, and / or any other type of treatment, evaluation, maintenance, therapy, medication, and / or advice.

[0206] As used herein, the term "clinical laboratory" refers to a facility for the testing or processing of materials from a living organism, e.g., a human. Non-limiting examples of processing include, for example, biological, biochemical, serological, chemical, immunohematological, hematological, biophysical, cytological, pathological, genetic, or other testing of materials from the human body for the purpose of providing information for the diagnosis, prevention, or treatment of any disease or disorder, or evaluation of the health of a living organism, e.g., a human. These tests can also include procedures for collecting, otherwise obtaining, preparing samples, identifying, measuring, or describing the presence or absence of various substances in a living organism, e.g., a human body, or in samples obtained from a living organism, e.g., a human body.

[0207] As used herein, the term "healthcare benefits provider" includes any individual party, organization, or entity involved in applying for, offering, providing, paying in whole or in part for, or otherwise making available to a patient one or more healthcare benefits, benefit plans, health insurance, and / or healthcare accounting programs.

[0208] In some embodiments, a health care provider can administer, or instruct another health care provider to administer, a therapy disclosed herein to treat cancer, e.g., a cancer selected from the group consisting of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and lung cancer (e.g., NSCLC).

[0209] A healthcare provider may provide means or instructions to another healthcare provider or a patient to perform the following actions: obtain a sample, process a sample, submit a sample, receive a sample, transfer a sample, analyze or measure a sample, quantify a sample, provide results after analyzing / measuring / quantifying a sample, receive results after analyzing / measuring / quantifying a sample, compare / score results after analyzing / measuring / quantifying one or more samples, provide a comparison / score from one or more samples, obtain a comparison / score from one or more samples, administer a therapy, begin administering a therapy, discontinue administering a therapy, continue administering a therapy, temporarily interrupt administering a therapy, increase the amount of therapeutic agent administered, decrease the amount of therapeutic agent administered, continue administering a constant amount of therapeutic agent, increase the frequency of administration of a therapeutic agent, decrease the frequency of administration of a therapeutic agent, maintain the same administration frequency for a therapeutic agent, replace a therapy or therapeutic agent with at least another therapy or therapeutic agent, combine a therapy or therapeutic agent with at least another therapy or additional therapeutic agent.

[0210] In some embodiments, the health care provider can authorize or deny, for example, sample collection, sample processing, sample submission, sample receipt, sample transfer, sample analysis or measurement, sample quantification, providing results obtained after analyzing / measuring / quantifying samples, transferring results obtained after analyzing / measuring / quantifying samples, comparing / scoring results obtained after analyzing / measuring / quantifying one or more samples, transferring comparisons / scores from one or more samples, administering a therapy or treatment, initiating administration of a therapy or treatment, discontinuing administration of a therapy or treatment, continuing administration of a therapy or treatment, suspending administration of a therapy or treatment, increasing the amount of a treatment administered, decreasing the amount of a treatment administered, continuing administration of a constant amount of a treatment, increasing the frequency of administration of a treatment, decreasing the frequency of administration of a treatment, maintaining the same frequency of administration for a treatment, replacing a therapy or treatment with at least another therapy or treatment, or combining a therapy or treatment with at least another therapy or additional treatment.

[0211] Additionally, the health care provider may, for example, approve or deny prescription of a therapy, approve or deny coverage of a therapy, approve or deny reimbursement of the cost of a therapy, determine or deny eligibility for a therapy, and the like.

[0212] In some embodiments, a clinical laboratory may, for example, collect or obtain cancer tumor samples, process samples, submit samples, receive samples, transfer samples, analyze or measure samples, quantify samples, provide results obtained after analyzing / measuring / quantifying samples, receive results obtained after analyzing / measuring / quantifying samples, compare / score results obtained after analyzing / measuring / quantifying one or more samples, provide a comparison / score from one or more samples, obtain a comparison / score from one or more samples, or perform other related activities, and the samples may be for gastric cancer (e.g., locally advanced, metastatic gastric cancer), or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, and lung cancer (e.g., NSCLC).

[0213] The assignment of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) patients or cancer tumors to particular TME phenotypic classes or classes disclosed herein may be applicable to patient treatment or patient selection for treatment, as well as other therapeutic or diagnostic uses.

[0214] For example, it can be applied to devise new methods of treatment (e.g., by selecting patients as candidates for a particular therapy or for participation in a clinical trial), to monitor the effectiveness of therapeutic drugs, or to adjust treatment (e.g., formulation, dosing regimen, or route of administration).

[0215] The methods disclosed herein may further include additional steps, such as prescribing, initiating, and / or altering prophylaxis and / or treatment, based, at least in part, on the determination of the presence or absence of a particular TME phenotype in a subject tumor from gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) by application of a classifier disclosed herein, e.g., a TME panel 1 classifier.

[0216] The present disclosure also provides methods for determining whether to treat a colorectal cancer patient having a tumor of a particular TME phenotype identified by application of a classifier disclosed herein, e.g., the TME panel 1 classifier, with a particular TME phenotype class-specific therapy disclosed herein, or a combination thereof. Also provided are methods for selecting a patient diagnosed with a particular type of colorectal cancer (e.g., left-sided colorectal cancer, right-sided colorectal cancer, dMMR colorectal cancer, MSI-H colorectal cancer, or metastatic colorectal cancer) as a candidate for treatment with a particular TME phenotype class-specific therapy disclosed herein, or a combination thereof, based on the presence and / or absence of a particular TME phenotype by application of a classifier disclosed herein, e.g., the TME panel 1 classifier.

[0217] The present disclosure also provides methods of determining whether to treat a cancer patient, e.g., a patient with gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), having a particular TME phenotype identified by application of a classifier disclosed herein, e.g., the TME panel 1 classifier.

[0218] Also provided are methods of selecting patients diagnosed with a particular type of colorectal cancer (e.g., left-sided colorectal cancer, right-sided colorectal cancer, dMMR colorectal cancer, MSI-H colorectal cancer, or metastatic colorectal cancer) as candidates for treatment with a particular TME phenotype class-specific therapy disclosed herein, or a combination thereof, based on the presence and / or absence of a particular TME phenotype by application of a classifier disclosed herein, e.g., the TME Panel 1 classifier.

[0219] Also provided is a method of selecting a patient diagnosed with a particular type of cancer selected from the group consisting of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), as a candidate for treatment with a particular TME phenotype class-specific therapy disclosed herein, or a combination thereof, based on the presence and / or absence of a particular TME phenotype identified by application of a classifier disclosed herein, e.g., the TME Panel 1 classifier.

[0220] In one aspect, a method disclosed herein includes making a diagnosis, which may be a differential diagnosis, based at least in part on assignment of a cancer tumor in a subject to a particular TME phenotype class of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), based on application of a classifier, e.g., a TME panel 1 classifier, to mRNA expression levels for a gene panel disclosed herein obtained from a sample derived from the tumor. This diagnosis can be recorded in the patient medical record. For example, in various embodiments, the classification of the cancer into a TME phenotype class, the diagnosis of the patient as subsequently treatable with a particular TME phenotype class-specific therapy or combination thereof, or the selected treatment can be recorded in the medical record. The medical record can be in paper form and / or can be maintained in a computer-readable medium. The medical record can be maintained by a laboratory, a clinic, a hospital, a health maintenance organization, an insurance company, and / or a personal health record website.

[0221] In some embodiments, a diagnosis, TME classification, selected therapy, etc., based on application of a classifier disclosed herein, e.g., a TME Panel 1 classifier, can be recorded on or in a medical alert item, such as a card, a wearable item, and / or a radio frequency identification (RFID) tag. As used herein, the term "wearable item" refers to any item that can be worn on a subject's body, including, but not limited to, a tag, bracelet, necklace, or armband.

[0222] In some embodiments, a sample can be obtained by a medical professional treating or diagnosing (e.g., using particular assays described herein) a cancer patient with gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) to measure biomarker levels in the sample (e.g., mRNA levels corresponding to a gene panel disclosed herein) under the direction of the medical professional.

[0223] In some embodiments, the clinical laboratory performing the assay is specialized in the treatment of gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer), and the like. Application of the classifiers disclosed herein, e.g., the TME Panel 1 classifier, can advise a cancer patient's healthcare provider as to whether their cancer (e.g., follicular carcinoma or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) may benefit from treatment with a particular TME phenotype class-specific therapy or combination thereof disclosed herein, based on whether the patient's cancer is classified as belonging to a particular TME phenotype class.

[0224] In some embodiments, the results of the TME phenotype classification derived by applying the classifiers disclosed herein, e.g., the TME Panel 1 Classifier, can be submitted to a healthcare benefits provider for determination of whether a cancer patient's insurance will cover a particular TME phenotype class-specific therapy or combination thereof disclosed herein. In some embodiments, the clinical laboratory performing the assay can advise a cancer patient's healthcare provider as to whether gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) could benefit from treatment with a particular TME phenotype class-specific therapy disclosed herein or a combination thereof, based on the TME phenotype classification of the cancer, e.g., using the TME panel 1 classifier disclosed herein.

[0225] The checkpoint inhibitor treatments disclosed above and throughout the specification may be used in combination with other checkpoint inhibitors, including nivolumab (PD-1), pembrolizumab (PD-1), durvalumab (PD-L1), atezolizumab (PD-L1), ABBV-181 (PD-1), AMG404 (PD-1), BI754091 (PD-1), dostallimab (PD-1), TSR-075 (PD-1 / LAG-3 bispecific), cetrelimab (PD-1), spartalizumab (PD-1), camrelizumab (PD-1), ATA2271 (PD-1), CDX-527 (PD-L1 / CD27 bispecific), cosibelimab (PD-L1), CX-072 (PD-1 / PD-L1 probody), FS222 (PD-L1 / CD137 FS118 (PD-L1 / LAG-3 bi-specific), GEN1046 (PD-L1 / CD137 bi-specific), JTX-4014 (PD-1), KY1043 (PD-L1), IMC-001 (PD-L1), TG-1501 (PD-L1), XmAb20717 (PD-1 / CTLA-4 XmAb23104 (PD-1 / ICOS bispecific), Genolimuzumab (PD-1), APL-502 (PD-L1), Cadnilimab (PD-1 / CTLA-4 bispecific), AK112 (PD-1 / VEGF bispecific), Penplimab (PD-1), KN046 (PD-L1 / CTLA-4 SHR-1316(PD-L1), BI-1206(PD-1 / FcyRIIB), BI-1808(PD-1 / TNFR2), PM8001(PD-L1 bispecific), CDX-527(PD-L1 / CD27 bispecific), IBI315(PD-1 / HER2 bispecific), HBM9167(PD-L1), HLX10(PD-1), LAE005(PD-L1), LZM009(PD-1), YBL-013(PD-L1 / CD3 Bispecific), avelumab (PD-L1), cemiplimab (PD-1), sintilimab (PD-1), tislelizumab (PD-1), toripalimab (PD-1), balstilimab (PD-1), zimberelimab (PD-1), sugemalimab (PD-L1), CS1003 (PD-1), GS-4224 (PD-L1), retifanlimab (PD-1), tebotelimab (PD-1 / LAG-3)DART), MGD019 (PD-1 / CTLA-4 DART), M7824 (PD-L1 / TGFβ Bifunctionality), Semen リマブ(PD-1), Semen バフォリマブ(PD-L1), ABSK043, ACE1708(P D-L1), AN4005 (PD-1 / PD-L1 small molecule), ALPN-202 (PD-L1 / CTLA-4 w / CD28), AVA004(PD-L1), BN-101A(PD-L1), プロルゴリマブ(PD-1), BCD-217(PD-1 / CTLA-4 dual specific), CCX-559(PD-L1), CTX-8371(PD-1 / PD-L1 (Dual specific), CB213 (PD-1 / LAG-3 dual specific), CA-170 (PD-L1 / VISTA), CA-327 (PD-L1 / TIM-3), Aurigene's PD-L1 / TIGIT, GNR-051 (PD-1), GS19 (PD-L1 / TGF-βR2 dual specific), HX008, IGM-7354 (PD-L1 / IL-15 dual specific), IMGS-001 (PD-1), MVR-T3011 (IL-12), INCB-086550 (PD-L1), INCB106385 (PD-1 / A2A / A2B / CD73), INBRX-105 (PD-L1 / CD137) quadruple-specific), IBI322 (PD-L1 / CD46 combination), IO103 (PD-L1), JS201 (PD-1 / TGF-β bifunctional), KD033 (PD-L1 / IL-15 bifunctional), GT90008 (PD-L1 / TGF-β bifunctional), Soka Soli Mab (PD-L1), MCLA-145 (PD-L1 / CD137), MT-6402 (PD-L1), ND021 (PD-L1 / CD137 / HSA triple-specific), OSE-279 (PD-1 / IL-7 bifunctional), PH-762-ACT (PD-1), PH-762-TME (PD-1), PRS-344 (PD-L1 / CD137 (Dual specific), QL1604, RG6139 (PD-1 / LAG-3 dual specific), RG6279 (PD-1 IL-2 Valiant), RG7769 (PD-1 / TIM-3 dual specific), SL-279252 (PD-L1 / OX40), SCT-I10A, STI-A1014 (PD-L1), PSB205 (PD-1 / CTLA4)The checkpoint inhibitors may include any checkpoint inhibitor selected from the group consisting of: PD-1 / CTLA-4 bispecific (PD-1 / CTLA-4 bispecific), PD-1 / ICOS bispecific (PD-1 / ICOS bispecific), CTLA-4 ...

[0226] II. ANN Classifier: TME Panel 1 Classifier In some aspects, the present disclosure provides methods for generating artificial neural network (ANN) classifiers capable of stratifying (or classifying) gene expression samples obtained from tumors derived from gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) into several TME phenotypic classes. The tumor biology underlying the four TME phenotype classes (i.e., stromal subtypes or phenotypes): IA (immune activation), ID (immune desert), A (angiogenesis), and IS (immunosuppression), described above, can be revealed through the application of ANNs. In some embodiments, application of the methods disclosed herein can classify a tumor sample or patient into two or more of the TME phenotype classes disclosed herein, e.g., a patient or sample can be positive for biomarkers for two or more TME phenotype classes.

[0227] The ANN takes as inputs (i.e., features) the gene expression values ​​of the genes disclosed herein, or a subset thereof, and identifies patient samples (i.e., patients) based on expression patterns as having primarily angiogenic expression, primarily activated immune gene expression, a mixture of both, or neither of these expression patterns. These four phenotypic types are predictive of response to certain types of treatment.

[0228] The ANN classifiers disclosed herein can be trained with gene expression data corresponding to a gene panel, e.g., data corresponding to a set of samples for which mRNA expression data has been obtained. For example, the training set includes expression data from the genes listed in Table 1 and Table 2, and any combination thereof. In some embodiments, the gene panel includes 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, In some embodiments, the gene panel comprises more than 100 genes. In some embodiments, the gene panel includes about 10 to about 20, about 20 to about 30, about 30 to about 40, about 40 to about 50, about 50 to about 60, about 60 to about 70, about 70 to about 80, about 80 to about 90, or about 90 to about 100 genes selected from Table 1 and Table 2. In some embodiments, the gene panel includes a gene set from Table 3 and a gene set from Table 4. In some embodiments, the gene panel is a gene panel selected from Table 5. In some embodiments, the gene panel is a gene panel (gene set) disclosed in Figures 9A-G.

[0229] The classifiers of the present disclosure depend on the selection of a particular gene panel as the source of input data used by the classifier. In some embodiments, each gene in a gene panel of the present disclosure is referred to as a "biomarker." The terms "gene set" and "gene panel" are used interchangeably. In some embodiments, a biomarker is a nucleic acid biomarker. As used herein, the term "nucleic acid biomarker" refers to a nucleic acid (e.g., a gene in a gene panel disclosed herein) that can be detected (e.g., quantified) in a subject or sample from which it is derived, e.g., a sample from a tumor, including tissue, stroma, cell lysate, and / or components thereof. In some embodiments, the term nucleic acid biomarker refers to the presence or absence of a particular sequence of interest (e.g., a nucleic acid variant or single nucleotide polymorphism) in a nucleic acid (e.g., a gene in a gene panel disclosed herein) that can be detected (e.g., quantified) in a subject or sample from which it is derived, e.g., a sample from a tumor, including tissue, stroma, cell lysate, and / or components thereof.

[0230] The "level" of a nucleic acid biomarker, in some embodiments, refers to the "expression level" of the biomarker, e.g., the level of RNA or DNA encoded by the nucleic acid sequence of the nucleic acid biomarker in a sample. For example, in some embodiments, the expression level of a particular gene disclosed in Table 1 or Table 2 refers to the amount of mRNA encoding such gene present in a sample obtained from a subject.

[0231] In some embodiments, the "level" of a nucleic acid biomarker, e.g., an RNA biomarker, can be determined by measuring a downstream output (e.g., the activity level of a target molecule or the expression level of an effector molecule that is being modulated, e.g., activated or repressed, by the nucleic acid biomarker or expression product, e.g., their RNA or DNA).

[0232] In some embodiments, the nucleic acid biomarker is an RNA biomarker. As used herein, "RNA biomarker" refers to an RNA that comprises the nucleic acid sequence of a nucleic acid biomarker of interest, e.g., an RNA that encodes a gene disclosed in Table 1 or Table 2.

[0233] The "expression level" of an RNA biomarker refers to the detected amount of a nucleic acid sequence of interest present in the subject or sample from which it is derived, e.g., the amount of RNA molecules expressed from a DNA molecule containing the nucleic acid sequence (e.g., the genome of the subject or the subject's cancer).

[0234] In some embodiments, the expression level of the RNA biomarker is the amount of the RNA biomarker in the tumor stroma sample. In some embodiments, the RNA biomarker is quantified using PCR (e.g., real-time PCR), sequencing (e.g., deep sequencing or next-generation sequencing, e.g., RNA-Seq), or other techniques that combine amplification or amplification with novel quantification methods, such as RNA-Seq, or other methods, using microarray expression profiling or RNase protection.

[0235] The methods disclosed herein include measuring the expression levels of a selected gene panel from a sample, e.g., a biological sample from a subject. See U.S. Patent Application No. 17 / 089,234, which is incorporated herein by reference in its entirety. Biomarker levels (e.g., expression levels of a gene panel of the present disclosure) can be measured in any tissue sample or biopsy from a subject or patient, including, for example, cancerous tissue, tumor, and / or stroma from the subject, that contains or is believed to contain one or more biomarkers (e.g., RNA biomarkers) disclosed herein. The tissue sample source can be, for example, solid tissue from a fresh, frozen, and / or preserved organ, tissue sample, biopsy, or aspirate. In some embodiments, the sample is a cell-free sample, e.g., containing cell-free nucleic acid (e.g., DNA or RNA). In some embodiments, the sample can contain compounds, such as preservatives, anticoagulants, buffers, fixatives, nutrients, antibiotics, and the like, that are not naturally mixed with tissue in nature.

[0236] In some embodiments, a fresh sample is preferred to an archived sample. As used herein, the terms "fresh sample," "non-archived sample," and grammatical variations thereof refer to a sample (e.g., a tumor sample from colorectal cancer, gastric cancer, breast cancer, prostate cancer, liver cancer, head and neck cancer, melanoma, ovarian cancer, glioma, glioblastoma, or lung cancer) that has been processed (e.g., to measure mRNA expression) for a predetermined period of time, e.g., less than one week after extraction from a subject. In some embodiments, a fresh sample is not frozen. In some embodiments, a fresh sample is not fixed. In some embodiments, a fresh sample has been stored for less than about two weeks, less than about one week, or less than six, five, four, three, or two days before processing.

[0237] As used herein, the term "archival sample" and grammatical variations thereof refer to a sample (e.g., a tumor sample from colorectal cancer, gastric cancer, breast cancer, prostate cancer, liver cancer, head and neck cancer, melanoma, ovarian cancer, glioma, glioblastoma, or lung cancer) that has been processed (e.g., to measure RNA) for a predetermined period of time, e.g., one week after extraction from a subject. In some embodiments, an archival sample is not frozen. In some embodiments, an archival sample is not fixed. In some embodiments, an archival sample has a known diagnosis and / or treatment history. In some embodiments, an archival sample has been stored for at least one week, at least one month, at least six months, or at least one year prior to processing.

[0238] In some cases, the biomarker levels are derived from fixed tumor tissue. In some embodiments, the sample is stored as a frozen sample or as a formalin-, formaldehyde-, or paraformaldehyde-fixed paraffin-embedded (FFPE) tissue preparation. For example, the sample can be embedded in a matrix, such as an FFPE block or a frozen sample. In some embodiments, the sample can include, for example, a tissue biopsy specimen or a surgical specimen. In some embodiments, the sample is or includes cells obtained from a patient.

[0239] In some embodiments, the sample can be obtained, for example, from surgical material or biopsy material (e.g., a recent biopsy, a recent biopsy since the last progression, or a recent biopsy since the last failed therapy). In some embodiments, the biopsy can be archival tissue from a previous course of therapy. In some embodiments, the biopsy can be treatment-naive tissue. In some embodiments, a biological fluid is not used as the sample.

[0240] The expression levels of genes in the gene panels described herein can be determined using any method in the art. In some embodiments, RNA expression levels are determined using a sequencing method, such as next-generation sequencing (NGS). In some embodiments, NGS is RNA-Seq, EdgeSeq, PCR, NanoString, or a combination thereof, or any technique that measures RNA. In some embodiments, the RNA measurement method includes nuclease protection. Specific methods for determining the expression levels of genes in the gene panels described herein are described in detail in U.S. Patent Application No. 17 / 089,234, which is incorporated herein by reference in its entirety.

[0241] In the ANN classifiers disclosed herein, the expression levels of genes in a gene panel obtained from a population of samples (e.g., samples from a clinical trial) and their assignment to TME phenotype classes (or a combination thereof, i.e., a sample can be classified not only as biomarker positive for a single TME phenotype class, but also as biomarker positive for two or more TME phenotype classes) obtained according to the population classifiers disclosed herein can be used as a training set for the ANN. The machine learning process can result in a model, e.g., an ANN model. The expression levels of the genes in the gene panel obtained from samples from a test subject are then used as inputs to the model, which can classify the subject's tumor into a particular TME phenotype class (or a combination thereof, i.e., a sample can be classified not only as biomarker positive for a single TME phenotype class, but also as biomarker positive for two or more TME phenotype classes).

[0242] Protein standard names, aliases, etc., and genes designated by identifiers used throughout this disclosure can be identified, for example, via Genecards (www.genecards.org) or Uniprot (www.uniprot.org). [Table 1] TIFF2024511166000002.tif104158

Table 2

Table 3

Table 4

[0243] In some embodiments, the gene panel to be used as part of the training set or model input to the ANN classifiers disclosed herein includes: ABCC9, ADAMTS4, AFAP1L2, AGR2, BACE1, BGN, BMP5, C11ORF9, CAPG, CAVIN2, CCL2, CCL3, CCL4, CD19, CD274, CD3E, CD4, CD79A, CD8B, COL4A2, COL8A1, COL8A2, CPXM2, CTLA4, CTSB, CXCL10, CXCL11, CXCL12, CXCL9, DUSP4, EBF1, ECM2, EDNRA, EIF5A, ELN, EPHA3, ETV5, FBLN5, FOLR2, GAD1, GNAS, GNB 4, GUCY1A1, GZMB, HAVCR2, HEY2, HFE, HMOX1, HP, HSPB2, IDO1, IFNA2, IFNB1, IFNG, IGFBP3, IGLL5, IL1B, IQGAP3, ITGA9, IT PR1, JAM2, JAM3, KCNJ8, LAG3, LAMB2, LHFPL6, LTBP4, MEOX1, MEST, MGP, MMP12, MMP13, MST1, MT2A, MTA2, NAALAD2, NFATC1 , NOV, OLFML2A, PCDH17, PDCD1, PDCD1LG2, PDE5A, PDGFRB, PEG3, PLA2G4A, PLAU, PLSCR2, PLXDC2, RAC2, REG4, RGS4, RGS5, R NF144A, RNH1, RRAS, RUNX1T1, SELP, SERPINE1, SERPINE2, SGIP1, SMARCA1, SPON1, SRSF6, STAB2, STEAP4, STRN3, TBX2, TEK , TGFB1, TGFB2, TIGIT, TIMP1, TLR9, TMEM204, TNFRSF18, TNFRSF4, TNFSF18, TRIM7, TTC28, USF1, UTRN, VSIR, and ZIC2.In some embodiments, the gene panel to be used as part of the training set or model input to the ANN classifiers disclosed herein includes: ABCC9, ADAMTS4, AFAP1L2, AGR2, BACE1, BGN, BMP5, C11ORF9, CAPG, CAVIN2, CCL2, CCL3, CCL4, CD19, CD274, CD3E, CD4, CD79A, CD8B, COL4A2, COL8A1, COL8A2, CPXM2, CTLA4, CTSB, CXCL10, CXCL11, CXCL12, CXCL9, DUSP4, EBF1, ECM2, EDNRA, EIF5A, ELN, EPHA3, ETV5, FBLN5, FOLR2, GAD1, GNAS, GNB4 , GUCY1A1, GZMB, HAVCR2, HEY2, HFE, HMOX1, HP, HSPB2, IDO1, IFNA2, IFNB1, IFNG, IGFBP3, IGLL5, IL1B, IQGAP3, ITGA9, ITP R1, JAM2, JAM3, KCNJ8, LAG3, LAMB2, LHFPL6, LTBP4, MEOX1, MEST, MGP, MMP12, MMP13, MST1, MT2A, MTA2, NAALAD2, NFATC1, NOV, OLFML2A, PCDH17, PDCD1, PDCD1LG2, PDE5A, PDGFRB, PEG3, PLA2G4A, PLAU, PLSCR2, PLXDC2, RAC2, REG4, RGS4, RGS5, RN F144A, RNH1, RRAS, RUNX1T1, SELP, SERPINE1, SERPINE2, SGIP1, SMARCA1, SPON1, SRSF6, STAB2, STEAP4, STRN3, TBX2, TEK, Consists of TGFB1, TGFB2, TIGIT, TIMP1, TLR9, TMEM204, TNFRSF18, TNFRSF4, TNFSF18, TRIM7, TTC28, USF1, UTRN, VSIR, and ZIC2.

[0244] In some embodiments, the gene panel to be used as part of the training set or model input to the ANN classifiers disclosed herein includes: ABCC9, ADAMTS4, AFAP1L2, AGR2, BACE1, BGN, BMP5, C11ORF9, CAPG, CAVIN2, CCL2, CCL3, CCL4, CD19, CD274, CD3E, CD4, CD79A, CD8B, COL4A2, COL8A1, COL8A2, CPXM2, CTLA4, CTSB, CXCL10, CXCL11, CXCL12, CXCL9, DUSP4, EBF1, ECM2, ECM3, ECM4, ECM5, ECM6, ECM7, ECM8, ECM9, ECM10, ECM11, ECM12, ECM13, ECM14, ECM15, ECM16, ECM17, ECM18, ECM19, ECM20, ECM21, ECM22, ECM23, ECM24, ECM25, ECM26, ECM27, ECM28, ECM29, ECM30, ECM31, ECM32, ECM40, ECM41, ECM42, ECM43, ECM44, ECM45, ECM46, ECM47, ECM48, ECM49, ECM50, ECM51, ECM52, ECM53, ECM54, ECM55, ECM56, ECM57, ECM58, ECM59, ECM60, ECM61, ECM62, ECM63, ECM74, ECM75, ECM76, ECM77, ECM78, ECM80, ECM81, ECM82, ECM93, ECM94, ECM95, ECM96, ECM97, E DNRA, EIF5A, ELN, EPHA3, ETV5, FBLN5, FOLR2, GAD1, GNAS, GNB4, GUCY1A1, GZMB, HAVCR2, HEY2, HFE, HMOX1, HP, HSPB2, IDO1, IFNA2, IFNB1, IFNG, IGFBP3 , IGLL5, IL1B, IQGAP3, ITGA9, ITPR1, JAM2, JAM3, KCNJ8, LAG3, LAMB2, LHFPL6, LTBP4, MEOX1, MEST, MGP, MMP12, MMP13, MST1, MT2A, MTA2, NAALAD2, NFAT C1, NOV, OLFML2A, PCDH17, PDCD1, PDCD1LG2, PDE5A, PDGFRB, PEG3, PLA2G4A, PLAU, PLSCR2, PLXDC2, RAC2, REG4, RGS4, RGS5, RNF144A, RNH1, RRAS, RUNX1 T1, SELP, SERPINE1, SERPINE2, SGIP1, SMARCA1, SPON1, SRSF6, STAB2, STEAP4, STRN3, TBX2, TEK, TGFB1, TGFB2, TIGIT, TIMP1, TLR9, TMEM204, TNFRSF18 , TNFRSF4, TNFSF18, TRIM7, TTC28, USF1, UTRN, VSIR, and ZIC2, 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, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, or 124 genes.

[0245] In some embodiments, the gene panel to be used as part of the training set or model input to the ANN classifiers disclosed herein includes: ABCC9, ADAMTS4, AFAP1L2, AGR2, BACE1, BGN, BMP5, C11ORF9, CAPG, CAVIN2, CCL2, CCL3, CCL4, CD19, CD274, CD3E, CD4, CD79A, CD8B, COL4A2, COL8A1, COL8A2, CPXM2, CTLA4, CTSB, CXCL10, CXCL11, CXCL12, CXCL9, DUSP4, EBF1, ECM2, ECM3, ECM4, ECM5, ECM6, ECM7, ECM8, ECM9, ECM10, ECM11, ECM12, ECM13, ECM14, ECM15, ECM16, ECM17, ECM18, ECM19, ECM20, ECM21, ECM22, ECM23, ECM24, ECM25, ECM26, ECM27, ECM28, ECM29, ECM30, ECM31, ECM32, ECM40, ECM41, ECM42, ECM43, ECM44, ECM45, ECM46, ECM47, ECM48, ECM49, ECM50, ECM51, ECM52, ECM53, ECM54, ECM55, ECM56, ECM57, ECM58, ECM59, ECM60, ECM61, ECM62, ECM63, ECM74, ECM75, ECM76, ECM77, ECM78, ECM80, ECM81, ECM82, ECM93, ECM94, ECM95, ECM96, ECM97, E DNRA, EIF5A, ELN, EPHA3, ETV5, FBLN5, FOLR2, GAD1, GNAS, GNB4, GUCY1A1, GZMB, HAVCR2, HEY2, HFE, HMOX1, HP, HSPB2, IDO1, IFNA2, IFNB1, IFNG, IGFBP3 , IGLL5, IL1B, IQGAP3, ITGA9, ITPR1, JAM2, JAM3, KCNJ8, LAG3, LAMB2, LHFPL6, LTBP4, MEOX1, MEST, MGP, MMP12, MMP13, MST1, MT2A, MTA2, NAALAD2, NFAT C1, NOV, OLFML2A, PCDH17, PDCD1, PDCD1LG2, PDE5A, PDGFRB, PEG3, PLA2G4A, PLAU, PLSCR2, PLXDC2, RAC2, REG4, RGS4, RGS5, RNF144A, RNH1, RRAS, RUNX1 T1, SELP, SERPINE1, SERPINE2, SGIP1, SMARCA1, SPON1, SRSF6, STAB2, STEAP4, STRN3, TBX2, TEK, TGFB1, TGFB2, TIGIT, TIMP1, TLR9, TMEM204, TNFRSF18 , TNFRSF4, TNFSF18, TRIM7, TTC28, USF1, UTRN, VSIR, and ZIC2, 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, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, or 124 genes.

[0246] In some embodiments, the ANN classifiers disclosed herein, for example, the TME Panel 1 classifier, were trained using the gene sets provided in the table below. [Table 5] TIFF2024511166000008.tif238158TIFF2024511166000009.tif238158TIFF2024511166000010.tif100158

[0247] In some embodiments, the training dataset includes additional variables for each sample, such as sample classification using a population-based classifier as disclosed in U.S. Patent Application No. 17 / 089,234. In other embodiments, the training data includes data about the sample, such as the type of treatment administered to the subject, dosage, dosing regimen, route of administration, presence or absence of concomitant treatment, response to therapy (e.g., complete response, partial response, or lack of response), age, weight, sex, ethnicity, tumor size, tumor stage, presence or absence of biomarkers, etc.

[0248] In some embodiments, a gene panel (e.g., a gene panel that determines an angiogenesis signature score or an immune signature score in an ANN classifier disclosed herein, e.g., TME panel 1, or a gene panel to be used as part of a training set or model input to an ANN classifier herein) comprises genes present in the gene sets disclosed in Figures 9A-G. In some embodiments, a gene panel (e.g., a gene panel that determines an angiogenesis signature score or an immune signature score in an ANN classifier disclosed herein, e.g., TME panel 1, or a gene panel to be used as part of a training set or model input to an ANN classifier herein) consists of genes present in the gene sets disclosed in Figures 9A-G. In some embodiments, a gene panel (e.g., a gene panel that determines an angiogenesis signature score or an immune signature score in an ANN classifier disclosed herein, e.g., TME panel 1, or a gene panel to be used as part of a training set or model input to an ANN classifier herein) does not comprise genes present in the gene sets disclosed in Figures 9A-G. In some embodiments, a gene panel (e.g., a gene panel that determines an angiogenesis signature score or immune signature score in an ANN classifier disclosed herein, e.g., TME Panel 1, or a gene panel to be used as part of a training set or model input to an ANN classifier herein) is not composed of genes present in the gene sets disclosed in Figures 9A-G.

[0249] In some embodiments, as can be understood by those skilled in the art, it is beneficial to select genes for training a dataset based on a combination of factors, including p-value, fold change, and coefficient of variation. In some embodiments, the use of one or more selection criteria and subsequent ranking allows the selection of the top 2.5%, 5%, 7.5%, 10%, 12.5%, 15%, 17.5%, 20%, 30%, 40%, 50% or more of the ranked genes in the gene panel for input into the model. Thus, as can be understood, all the individual identified genes or a subset of genes in Table 1 and Table 2 can be selected, and all possible combinations of the selected genes can be tested to identify gene combinations that are useful for generating a predictive model. To determine the number of selected individual genes to be tested in combination, and to select the number of possible combinations of genes, the selection criteria depend on the data resources available for obtaining genetic data and / or the computer resources available for calculating and evaluating the classifiers resulting from the model.

[0250] In some embodiments, based on the results of training a machine learning model, a gene may appear to be a driver gene. As used herein, the term "driver gene" refers to a gene containing a driver gene mutation. In some embodiments, a driver gene is a gene that may be associated with cancer progression due to one or more acquired mutations, e.g., driver gene mutations. In some embodiments, a driver gene can regulate one or more cellular processes, including cell fate determination, cell survival, and genome maintenance. A driver gene can be associated with (e.g., can regulate) one or more signal transduction pathways, such as the TGFβ pathway, the MAPK pathway, the STAT pathway, the PI3K pathway, the RAS pathway, the cell cycle pathway, the apoptosis pathway, the Notch pathway, the Hedgehog (HH) pathway, the APC pathway, a chromatin modification pathway, a transcriptional regulation pathway, a DNA damage control pathway, or a combination thereof. Exemplary driver genes include oncogenes and tumor suppressors. In some embodiments, a driver gene provides a selective growth advantage to cells in which it occurs. In some embodiments, the driver gene provides proliferation capabilities to the cell in which it occurs, e.g., allows cell proliferation, e.g., clonal expansion. In some embodiments, the driver gene is an oncogene. In some embodiments, the driver gene is a tumor suppressor gene (TSG).

[0251] The presence of noisy, low-expression genes in a gene set can reduce the sensitivity of the model. Thus, in some embodiments, low-expression genes can be reduced or filtered out (removed) from the machine learning model. In some embodiments, low-expression gene filtering is based on statistics calculated from gene expression (e.g., RNA levels). In some embodiments, low-expression gene filtering is based on, for example, the minimum (min), maximum (max), mean (mean), variance (sd) of the raw read count for each gene in the gene set, or a mixture thereof. For each gene set, an optimal filtering threshold can be determined. In some embodiments, the filtering threshold is optimized to maximize the number of differentially expressed genes in the gene set.

[0252] ANN classifier is then evaluated by determining the ability of classifier to accurately determine each test subject.In some embodiments, the subjects of the training group used to derive model are different from the subjects of the test group used to use model.As those skilled in the art will understand, this allows predicting the ability of the gene set used to train classifier with respect to their ability to properly characterize subjects whose stromal phenotypic trait characterization (for example, TME phenotype class) is unknown.

[0253] The data that is input to the mathematical model (ANN) can be any data, such as the expression level specific to the gene being evaluated, for example, the mRNA product. The mathematical model useful in the present disclosure includes those that use supervised and / or unsupervised learning techniques. In some embodiments of the present disclosure, the selected mathematical model uses supervised learning in conjunction with a "training population" to evaluate each possible biomarker combination.

[0254] Classifiers (e.g., ANN models) generated by the methods disclosed herein can be used to test unknown or test subjects. In one embodiment, a model generated by the ANN identified herein can detect whether a subject or cancer sample belongs to a particular TME phenotype class. In some embodiments, an ANN model can predict whether a subject will respond to a particular therapy. In other embodiments, an ANN model can select, or be used to select, a subject for administration of a particular therapy.

[0255] In one embodiment of the present disclosure, each ANN classifier is evaluated for its ability to adequately characterize each subject in the training population using methods known to those skilled in the art. For example, ANN classifiers can be evaluated using cross-validation, leave-one-out cross-validation (LOOCV), n-fold cross-validation, or jackknife analysis using standard statistical methods. In another embodiment, each ANN classifier is evaluated for its ability to adequately characterize those subjects in the training population that were not used to generate the classifier.

[0256] In some embodiments, an ANN classifier can be trained using one data set and then evaluated against another, different data set, thus eliminating the need for cross-validation since the test data set is different from the training data set.

[0257] In one embodiment, the method used to evaluate the ability of a classifier to adequately characterize each subject in a training population is a method that evaluates the sensitivity (TPF, true positive rate) and 1-specificity (FPF, false positive rate) of the classifier. In one embodiment, the method used to test a classifier is the receiver operating characteristic ("ROC"), which provides several parameters for evaluating both the sensitivity and specificity of the results of a generated model, e.g., a model obtained from the application of an ANN.

[0258] In some embodiments, the metrics used to evaluate a classifier's ability to adequately characterize each subject in the training population include classification accuracy (ACC), area under the receiver operating characteristic curve (AUC ROC), sensitivity (true positive rate, TFP), specificity (true negative rate, TNF), positive predictive value (PPV), negative predictive value (NPV), or any combination thereof. In a particular embodiment, the metrics used to evaluate a classifier's ability to adequately characterize each subject in the training population are classification accuracy (ACC), area under the receiver operating characteristic curve (AUC ROC), sensitivity (true positive rate, TFP), specificity (true negative rate, TNF), positive predictive value (PPV), and negative predictive value (NPV).

[0259] In some embodiments, the training set comprises a reference population of at least about 10, at least about 20, at least about 30, at least about 40, at least about 50, at least about 60, at least about 70, at least about 80, at least about 90, at least about 100, at least about 110, at least about 120, at least about 130, at least about 140, at least about 150, at least about 160, at least about 170, at least about 180, at least about 190, at least about 200, at least about 250, at least about 300, at least about 350, at least about 400, at least about 450, at least about 500, at least about 600, at least about 700, at least about 800, at least about 900, or at least about 1000 individuals (or animals).

[0260] In some embodiments, the measured expression, e.g., mRNA levels, for each biomarker in the gene panel of the present disclosure can be used to train a neural network. The neural network is a two-stage regression or classification model. The neural network can be binary or non-binary. The neural network has a layered structure including an input unit layer (and bias) connected to a weight layer of an output unit layer. For regression, the output unit layer typically contains only one output unit. However, neural networks can seamlessly handle multiple quantitative responses. Thus, neural networks can be applied to enable the identification of biomarkers that distinguish between three or more populations (i.e., three or more phenotypic traits), for example, the four TME phenotypic classes disclosed herein.

[0261] In one specific example, a neural network can be trained using expression data from the products, e.g., mRNA, of the biomarker genes disclosed in Tables 1 and 2 for a sample set from a subject population to identify combinations of these biomarkers that are specific to a particular TME phenotype. Neural networks are described in Duda et al., 2001, Pattern Classification, Second Edition, John Wiley & Sons, Inc., New York; and Hastie et al., 2001, The Elements of Statistical Learning, Springer-Verlag, New York.

[0262] In some embodiments, the ANN classifiers disclosed herein, e.g., TME Panel 1, comprise a backpropagation neural network with a single input layer having, for example, 98 or 87 genes from Tables 1 and 2, a single hidden layer of two neurons, and four outputs in a single output layer (see, e.g., Abdi, 1994, "A neural network primer," J. Biol. System. 2, 247-283). ANN classifiers such as TME Panel 1 disclosed herein can be incorporated using the EasyNN-Plus version 4.0g software package (Neural Planner Software Inc.), Scikit-Learn (scikit-learn.org), or any other machine learning package or program known in the art.

[0263] In some embodiments, the ANN classifier is a feedforward neural network. A feedforward neural network is an artificial network in which the connections between input and output nodes do not form a cycle. In the case of an ANN, the terms "node" and "neuron" are used interchangeably herein. Therefore, it is different from a recurrent neural network. In this network, information moves from the input node through hidden nodes (if any) in one direction: forward to the output node. In this network, there are no cycles or loops. Except for the input node, each node is a neuron that uses a nonlinear activation function, which was developed to model the action potential, or firing frequency, of biological neurons.

[0264] In some embodiments, the ANN classifier is a single-layer perceptron network, which consists of a single layer of output nodes, where the input is fed directly to the output via a set of weights. The sum of the products of the weights and the inputs is calculated at each node, and if the value exceeds some threshold (usually 0), the neuron fires and takes on an activation value (usually 1).

[0265] In some embodiments, the ANN is a multilayer perceptron (MLP). This class of network typically consists of multiple layers of computational units interconnected in a feedforward manner. Each neuron in one layer has a connection directed to neurons in the subsequent layer. In many applications, the units of these networks utilize activation functions, e.g., sigmoid functions. An MLP includes at least three layers of nodes: an input layer, a hidden layer, and an output layer.

[0266] In some embodiments, the activation function is of the form y(v i )=tanh(v i ), i.e., a sigmoid function described by a hyperbolic tangent ranging from −1 to +1. In some embodiments, the activation function is of the form y(v i )=(1+e -vi ) -1, i.e., a logistic function similar in shape to the tanh function but ranging from 0 to +1. In these equations, y i is the output of the i-th node (neuron), and v i is the weighted sum of the input connections.

[0267] In some aspects, the activation function is a rectified linear unit (ReLU) or a variant thereof, such as a noisy ReLU, a leaky ReLU, a parametric ReLU, or an exponential LU. In some aspects, a ReLU is a function of the formula f(x)=X + =max(0,x), where x is the input to the neuron. The ReLU activation function allows for better training of deep neural networks (DNNs) compared to hyperbolic tangent or logistic sigmoid. A DNN is an ANN with multiple layers between the input and output layers. DNNs are typically feedforward networks, where data flows from the input layer to the output layer without folding back. DNNs tend to overfit due to the added layer of abstraction, which allows them to model rare dependencies in the training data. In some embodiments, the activation function is a soft plus or smooth ReLU function, which is a smooth approximation of ReLU, and has the formula f(x)=ln(1+e x ) The derivative of the softplus is the logistic function.

[0268] In some embodiments, the ANN is an MLP that includes three or more layers of nonlinearly activated nodes (input and output layers with one or more hidden layers). Its multiple layers and nonlinear activation distinguish it from a linear perceptron, which can distinguish data that is not linearly separable. Because the ANN is fully connected, each node in a layer has a certain weight w ij It connects to all the nodes in the next layer with a weight. Learning occurs in a perceptron by changing the connection weights after each piece of data is processed, based on the amount of error in the output compared to the predicted result. This is an example of supervised learning, and is performed via backpropagation.

[0269] In some embodiments, the ANN has three layers. In other embodiments, the ANN has more than three layers. In some embodiments, the ANN has a single hidden layer. In other embodiments, the ANN has two or more hidden layers.

[0270] In some aspects, the input layer is 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, 110, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 20, 21, 22 2, 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, 110, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121 , 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, or 150 neurons.

[0271] In some embodiments, the input layer comprises 70 to 100 neurons. In some embodiments, the input layer comprises 70 to 80 neurons. In some embodiments, the input layer comprises 80 to 90 neurons. In some embodiments, the input layer comprises 90 to 100 neurons. In some embodiments, the input layer comprises 70 to 75 neurons. In some embodiments, the input layer comprises 75 to 80 neurons. In some embodiments, the input layer comprises 80 to 85 neurons. In some embodiments, the input layer comprises 85 to 90 neurons. In some embodiments, the input layer comprises 90 to 95 neurons. In some embodiments, the input layer comprises 95 to 100 neurons.

[0272] In some embodiments, the input layer comprises at least about 1 to at least about 5, at least about 5 to at least about 10, at least about 10 to at least about 15, at least about 15 to at least about 20, at least about 20 to at least about 25, at least about 25 to at least about 30, at least about 30 to at least about 35, at least about 35 to at least about 40, at least about 40 to at least about 45, at least about 45 to at least about 50, at least about 50 to at least about 55, at least about 55 to at least about 60, at least about 60 to at least about 65, at least about 65 to at least about 70, at least about 70 to at least about 75, at least about 75 to at least about 80, At least about 80, at least about 80 to at least about 85, at least about 85 to at least about 90, at least about 90 to at least about 95, at least about 95 to at least about 100, at least about 100 to at least about 105, at least about 105 to at least about 110, at least about 110 to at least about 115, at least about 115 to at least about 120, at least about 120 to at least about 125, at least about 125 to at least about 130, at least about 130 to at least about 135, at least about 135 to at least about 140, at least about 140 to at least about 145, or at least about 145 to at least about 150 neurons.

[0273] In some embodiments, the input layer comprises at least about 1 to at least about 10, at least about 10 to at least about 20, at least about 20 to at least about 30, at least about 30 to at least about 40, at least about 40 to at least about 50, at least about 50 to at least about 60, at least about 60 to at least about 70, at least about 70 to at least about 80, at least about 80 to at least about 90, at least about 90 to at least about 100, at least about 100 to at least about 110, at least about 110 to at least about 120, at least about 120 to at least about 130, at least about 130 to at least about 140, or at least about 140 to at least about 150 neurons.

[0274] In some embodiments, the input layer comprises at least about 1 to at least about 20, at least about 20 to at least about 40, at least about 40 to at least about 60, at least about 60 to at least about 80, at least about 80 to at least about 100, at least about 100 to at least about 120, at least about 120 to at least about 140, at least about 10 to at least about 30, at least about 30 to at least about 50, at least about 50 to at least about 70, at least about 70 to at least about 90, at least about 90 to at least about 110, at least about 110 to at least about 130, or at least about 130 to at least about 150 neurons.

[0275] In some embodiments, the input layer comprises more than about 1, more than about 5, more than about 10, more than about 15, more than about 20, more than about 25, more than about 30, more than about 35, more than about 40, more than about 45, more than about 50, more than about 55, more than about 60, more than about 65, more than about 70, more than about 75, more than about 80, more than about 85, more than about 90, more than about 95, more than about 100, more than about 105, more than about 110, more than about 115, more than about 120, more than about 125, more than about 130, more than about 135, more than about 140, more than about 145, or more than about 150 neurons.

[0276] In some embodiments, the input layer comprises less than about 1, less than about 5, less than about 10, less than about 15, less than about 20, less than about 25, less than about 30, less than about 35, less than about 40, less than about 45, less than about 50, less than about 55, less than about 60, less than about 65, less than about 70, less than about 75, less than about 80, less than about 85, less than about 90, less than about 95, less than about 100, less than about 105, less than about 110, less than about 115, less than about 120, less than about 125, less than about 130, less than about 135, less than about 140, less than about 145, or less than about 150 neurons. In some embodiments, a weight is applied to the input of each neuron in the input layer.

[0277] In some aspects, the ANN includes a single hidden layer. In some embodiments, the ANN includes 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 hidden layers. In some embodiments, the single hidden layer includes 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 neurons. In some aspects, the single hidden layer includes at least 1, at least 2, at least 3, at least 4, at least 5, at least 6, at least 7, at least 8, at least 9, or at least 10 neurons. In some aspects, the single hidden layer includes fewer than 10, fewer than 9, fewer than 8, fewer than 7, fewer than 6, fewer than 5, fewer than 4, or fewer than 3 neurons. In some embodiments, the single hidden layer includes two neurons. In some embodiments, the single hidden layer includes three neurons. In some embodiments, the single hidden layer includes four neurons. In some embodiments, the single hidden layer includes five neurons. In some aspects, a bias is applied to the neurons in the hidden layer.

[0278] In some embodiments, the ANN comprises four neurons in the output layer corresponding to different TME phenotypes, hi some embodiments, the four neurons in the output layer correspond to the four TME phenotype classes disclosed above: IA (immune activation), IS (immune suppression), ID (immune desert), and A (angiogenesis).

[0279] In some embodiments, the output layer classifications are normalized to a probability distribution for the predicted output class, where the components sum to one and so can be interpreted as probabilities.

[0280] In some embodiments, multi-class classification of output layer values ​​into four TME phenotype classes (IA, ID, A, and IS) is aided by applying a logistic regression function. In some embodiments, multi-class TME classification of output layer values ​​into four TME phenotype classes (IA, ID, A, and IS) is aided by applying a logistic regression classifier, e.g., a softmax function. Softmax assigns decimal probabilities to each class that sum to 1.0. In some embodiments, the use of a logistic regression classifier such as a softmax function aids in more rapid training convergence. In some embodiments, a logistic regression classifier including a softmax function is incorporated via a neural network layer immediately prior to the output layer. In some embodiments, such a neural network layer immediately prior to the output layer has the same number of nodes as the output layer.

[0281] In some embodiments, depending on the particular dataset used, different cutoffs are applied to the results of the logistic regression classifier (e.g., a softmax function) (e.g., the cutoffs are applied to select a particular subject population, e.g., a population responsive to a particular therapy). Thus, applying different cutoffs not only classifies a cancer or patient into only one of the four TME phenotype classes disclosed above, IA (immune activation), IS (immune suppression), ID (immune desert), or A (angiogenesis), but also classifies a cancer or patient into two or more of the TME phenotype classes disclosed above. Thus, in some embodiments, a cancer or patient can be classified as biomarker-positive for the IA, IS, ID, or A TME phenotype classes, or a combination thereof. Conversely, in some embodiments, a cancer or patient can be classified as biomarker-negative for the IA, IS, ID, or A TME phenotype classes, or a combination thereof.

[0282] In some embodiments, all of the genes, or a subset of the genes, in the angiogenesis signature and all of the genes, or a subset of the genes, in the immune signature have positive or negative gene weights in the ANN model for each hidden layer.

[0283] The actual behavior of the ANN model of the present disclosure is to present high-dimensional data in a compressed form. The compressed data can be visually represented in a so-called latent space. A common example of this is a two-dimensional graph (X & Y axis), where each patient is plotted as a number of vector X and vector Y values. Thus, the latent space is a projection of the signature generated by the method of the present disclosure, whether it is a projection of Z-scores or hidden neuron values. In some embodiments, the latent space can be plotted in three dimensions.

[0284] The disease score value (i.e., the probability result of the ANN model) for each patient can be plotted in the latent space. Over time, patient data can be accumulated, or the results of a retrospective analysis of patient data including disease scoring can be used as a reference plot onto which the ANN probability result for that patient is plotted.

[0285] In some embodiments, the latent space is a plot of the hidden neurons of the ANN model and can include all two-way combinations of these neurons. In some embodiments, the ANN model predicts four TME phenotype classes based on data compressed into two hidden neurons, and the plot of these neurons in the latent space also serves as a projection of the four output TME phenotype classes. In some embodiments, the TME phenotype class assignments for each patient are visualized in the Neuron 1 vs. Neuron 2 latent space.

[0286] The latent space projection can be enhanced by displaying probability contours of the output TME phenotype assignments. In this way, the projection can show not only where the subject falls in the latent space, but also the confidence of each TME phenotype classification. In some embodiments, clinical reports can use the TME phenotype classes as the biomarker logic, i.e., IA = positive, or IA + IS = positive, and then report the probability of TME phenotype assignment to the clinician, which is now the output of the model. The latent space plot can also be used to visualize the patient's distance from the decision boundary and assist clinical decision makers in evaluating edge cases and exceptions.

[0287] In some embodiments, the boundaries between TME phenotype classes are not on the Cartesian axes (x=0, y=0), but elsewhere on the plot.

[0288] In some embodiments, a second model can learn biomarker boundaries from the ANN model latent space. In some embodiments, the second model can be a logistic regression model. In some embodiments, it can be any other type of regression or machine learning algorithm. In some embodiments, a logistic regression function can be applied to the latent space. In some embodiments, when TME phenotypes are combined to define a biomarker-positive class, i.e., IA+IS, the confidence of the individual phenotype assignments is not the same as the confidence of the combined class assignment. The logistic regression function is used to learn what being biomarker-positive means and directly report statistics regarding being biomarker-positive. The logistic regression function can be used to fine-tune the biomarker positive / negative decision boundaries based on actual patient outcome data. In some embodiments, the accuracy of the ANN model can be improved by slicing the latent space according to a secondary model.

[0289] In some embodiments, the probability function can be plotted two-dimensionally, with one axis representing the probability that the signal is dominated by genes in the angiogenesis signature and the other axis representing the probability that the signal is dominated by genes in the immune signature. In some embodiments, genes involved in angiogenesis and immune function contribute to the respective probability functions. Each quadrant of the latent space plot represents a stromal phenotype. In further embodiments, a threshold is applied using logistic regression. In some embodiments, the logistic regression can be linear or polynomial. After setting the threshold, individual patient outcomes can be analyzed using the methods described herein.

[0290] III. TME Phenotype Class-Specific Therapy Analysis of RNA expression data from gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC) indicates that certain cancer patient subpopulations can be effectively treated with IA (immunoactive), IS (immunosuppressive), A (angiogenic), or ID (immune desert) TME phenotype class-specific therapy by the methods disclosed above. Assignment of a patient or cancer tumor to one of four TME phenotype classes using a classifier disclosed herein, e.g., the TME Panel 1 classifier, can be used to predict which therapy will be more effective for treating a particular patient, e.g., a subpopulation of patients with left- or right-sided colorectal cancer, mismatch repair deficient (dMMR), or MSI-H colorectal cancer. See, e.g., Figure 8.

[0291] Classification of dMMR or MSI-H tumors, e.g., colorectal cancer tumors, into the IA TME phenotype class is associated with improved clinical outcomes with treatment with checkpoint inhibitors. Accordingly, patients with dMMR or MSI-H tumors, e.g., colorectal cancer tumors, having an IA TME phenotype will be administered a therapy comprising a checkpoint inhibitor selected from the IA TME phenotype class-specific therapies disclosed below. Classification of dMMR or MSI-H tumors, e.g., colorectal cancer tumors, into the IS TME phenotype class is associated with improved clinical outcomes with combined treatment with a checkpoint inhibitor and a phosphatidylserine inhibitor. Accordingly, patients with dMMR or MSI-H tumors, e.g., colorectal cancer tumors, having an IS TME phenotype will be administered a combination therapy comprising a checkpoint inhibitor and a phosphatidylserine inhibitor selected from the IS TME phenotype class-specific therapies disclosed below. Classification of metastatic tumors, e.g., metastatic colorectal tumors, into the A or IS TME phenotype class is associated with improved clinical outcomes with treatment with angiogenesis inhibitors. Thus, patients with metastatic cancers with the A or IS TME phenotype will be administered a therapy comprising an angiogenesis inhibitor selected from the A TME phenotype class-specific therapies disclosed below. Classification of tumors into the IA TME phenotype class can be used to select a checkpoint inhibitor, e.g., pembrolizumab, as an adjuvant therapy. Classification of tumors into the A TME phenotype class can be used to select an antiangiogenic therapy, e.g., with bevacizumab as an adjuvant therapy. Furthermore, in the case of colorectal cancer, classification of tumors of the left or right colorectum as having a predominant TME phenotype class can be used, for example, to select a therapy described below that would fit the predominant TME phenotype class.

[0292] III.A:IA TME Phenotype Class-Specific Therapy Gastric cancer (e.g., locally advanced, metastatic, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., metastatic to the liver), which are classified into the IA (immune activity) TME phenotype class by a classifier of the present disclosure, e.g., a TME Panel 1 classifier (i.e., an IA biomarker-positive patient), Immune-dominated TMEs, such as those of tumors from ovarian cancer (e.g., recurrent, advanced colorectal cancer), ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), are likely to respond to immune checkpoint inhibitors (CPIs) such as anti-PD-1 (e.g., sintilimab, tislelizumab, pembrolizumab, or antigen-binding portions thereof), anti-PD-L1, or anti-CTLA-4, or RORy agonist therapeutics.

[0293] Checkpoint inhibitors: In some embodiments, the immune checkpoint inhibitor is a blocking antibody that binds to PD-1, e.g., nivolumab, cemiplimab (REGN2810), geptanolimab (CBT-501), pacmilimab (CX-072), dostallimab (TSR-042), sintilimab, tislelizumab, and pembrolizumab; a blocking antibody that binds to PD-L1, e.g., durvalumab (MEDI4736), avelumab, lodapolimab (LY-3300054), CX-188, and atezolizumab; or a blocking antibody that binds to CTLA-4, e.g., ipilimumab and tremelimumab. In some embodiments, a combination of one or more such antibodies can also be used.

[0294] Tremelimumab, nivolumab, durvalumab, and atezolizumab are described, for example, in U.S. Patent Nos. 6,682,736, 8,008,449, 8,779,108, and 8,217,149, respectively. In some embodiments, atezolizumab can be replaced with another immune checkpoint antibody, such as another blocking antibody that binds to CTLA-4, PD-1 (e.g., sintilimab, tislelizumab, pembrolizumab, or an antigen-binding portion thereof), PD-L1, or a bispecific blocking antibody that binds to any checkpoint inhibitor. When selecting a different blocking antibody, those skilled in the art will know the appropriate dosage and administration schedule from the literature. Suitable examples of anti-CTLA-4 antibodies include those described in U.S. Patent No. 6,207,156. Other suitable examples of anti-PD-L1 antibodies are those described in the following patents: U.S. Patent No. 8,168,179, which relates particularly to the treatment of PD-L1-overexpressing cancers with human anti-PD-L1 antibodies, including in combination with chemotherapy; U.S. Patent No. 9,402,899, which relates particularly to the treatment of tumors with antibodies against PD-L1, including chimeric, humanized, and human antibodies; and U.S. Patent No. 9,439,962, which relates particularly to the treatment of cancer with anti-PD-L1 antibodies and chemotherapy.

[0295] Additional suitable antibodies against PD-L1 are those described in U.S. Patent Nos. 7,943,743, 9,580,505, and 9,580,507, kits thereof (U.S. Patent No. 9,580,507), and nucleic acids encoding the antibodies (U.S. Patent No. 8,383,796). Such antibodies bind to PD-L1 and compete for binding with a reference antibody; are defined by VH and VL genes; or are defined by predetermined sequences or conservative modifications of the heavy and light chain CDR3s (U.S. Patent No. 7,943,743) or heavy chain CDR3s (U.S. Patent No. 8,383,796); or have 90% or 95% sequence identity to the reference antibody. These anti-PD-L1 antibodies also include antibodies, immunoconjugates, and bispecific antibodies with predetermined quantitative (including binding affinity) and qualitative properties. Further included are methods of using such antibodies, and antibodies with predetermined quantitative (including binding affinity) and qualitative properties, including antibodies in single chain format and isolated CDR format, to enhance an immune response. Enhanced immune responses, such as those in U.S. Patent No. 9,102,725, can be used to treat cancer.

[0296] Further suitable antibodies against PD-L1 are those described in U.S. Patent Application No. 2016 / 0009805, which relates to antibodies against specific epitopes on PD-L1, including antibodies of defined CDR sequences and competing antibodies; nucleic acids, vectors, host cells, immunoconjugates; detection, diagnostic, prognostic and biomarker methods; and therapeutic methods.

[0297] Specific therapeutic agents including ipilimumab are disclosed in U.S. Patent Nos. 7,605,238; 8,318,916; 8,784,815; and 8,017,114. Therapeutic agents including tremelimumab are disclosed, for example, in U.S. Patent Nos. 6,682,736, 7,109,003, 7,132,281, 7,411,057, 7,807,797, 7,824,679, 8,143,379, 8,491,895, and 8,883,984. Treatment with nivolumab is disclosed, for example, in U.S. Patent Nos. 8,008,449, 8,779,105, 9,387,247, 9,492,539, 9,492,540, 8,728,474, 9,067,999, 9,073,994, and 7,595,048. Treatment with pembrolizumab is disclosed, for example, in U.S. Patent Nos. 8,354,509, 8,900,587, and 8,952,136. Treatment with cemiplimab is disclosed, for example, in U.S. Patent Application Publication No. 20150203579A1. Treatment with durvalumab is disclosed, for example, in U.S. Patent Nos. 8,779,108 and 9,493,565. Therapeutic agents including atezolizumab are disclosed, for example, in U.S. Patent No. 8,217,149. Treatment with CX-072 is disclosed, for example, in U.S. Patent Application No. 15 / 069,622. Treatment with LY30054 is disclosed, for example, in U.S. Patent No. 10,214,586B2. Treatment of tumors with a combination of antibodies against PD-1 and CTLA-4 is disclosed in U.S. Patent Nos. 9,084,776, 8,728,474, 9,067,999, and 9,073,994. Treatment of tumors with antibodies against PD-1 and CTLA-4, including subtherapeutic doses and PD-L1-negative tumors, is disclosed in U.S. Patent No. 9,358,289. Treatment of tumors with antibodies against PD-L1 and CTLA-4 is disclosed in U.S. Patent Nos. 9,393,301 and 9,402,899, all of which patents and publications are incorporated herein by reference in their entireties.

[0298] Specific therapeutic agents and whether they are approved for the treatment of solid tumors are identified in the table below. [Table 6]

[0299] RORy Agonist Therapeutic Agents: In some embodiments, RORy agonist therapeutic agents are small molecule agonists of RORy (retinoid-related orphan receptor gamma), which belongs to the nuclear hormone receptor family. RORy plays an important role in regulating apoptosis during thymocyte formation and T cell homeostasis. Small molecule agonists in clinical development include LYC-55716 (cintirorgon).

[0300] Tislelizumab Tislelizumab (BGB-A317) is a humanized monoclonal antibody against PD-1. It prevents PD-1 from binding to its ligands PD-L1 and PD-L2 (thus, it is a checkpoint inhibitor). Tislelizumab can be used to treat solid tumors, such as Hodgkin's lymphoma (alone or in combination with adjuvant therapy, such as platinum-containing chemotherapy), urothelial carcinoma, NSCLC, or hepatocellular carcinoma. Sequences related to tislelizumab are provided in the table below. In some embodiments of the present disclosure, tislelizumab, or an antigen-binding portion thereof, can be administered in combination with bavituximab. [Table 7]

[0301] sintilimab Sintilimab (TYVYT®) is a fully human IgG4 monoclonal antibody against PD-1. It prevents PD-1 from binding to its ligands PD-L1 and PD-L2 (thus, it is a checkpoint inhibitor). Sintilimab can be used to treat solid cancers, such as Hodgkin's lymphoma, alone or in combination with adjuvant therapy. Related sequences for sintilimab are provided in the table below. In some embodiments of the present disclosure, sintilimab, or an antigen-binding portion thereof, can be administered in combination with bavituximab. [Table 8]

[0302] III.B:IS TME Phenotype Class-Specific Therapy Gastric cancer (e.g., locally advanced, metastatic gastric cancer, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colon cancer metastatic to the liver), which are classified into the IS (immunosuppressed) TME phenotype class by a classifier of the present disclosure, e.g., a TME Panel 1 classifier (i.e., IS biomarker-positive patients), A predominantly immunosuppressed TME, such as that of tumors from colorectal cancer, ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer), glioma (e.g., metastatic glioma), glioblastoma, or lung cancer (e.g., NSCLC), can be resistant to checkpoint inhibitors unless drugs that reverse immunosuppression, such as phosphatidylserine (anti-PS) and anti-phosphatidylserine targeted therapeutics, PI3Kγ inhibitors, adenosine pathway inhibitors, IDO, TIM, LAG3, TGFβ, and CD47 inhibitors, are also administered.

[0303] Bavituximab is a preferred anti-PS targeted therapeutic. Patients with this biology also have underlying angiogenesis and can similarly benefit from antiangiogenic agents, such as those used for the A TME phenotype.

[0304] Specific therapeutic agents for patients with cancer tumors assigned to the IS TME phenotype class by a classifier disclosed herein, e.g., the TME Panel 1 classifier, are discussed herein. Anti-PS and PS-targeting antibodies include, but are not limited to, bavituximab; PI3Kγ inhibitors such as LY3023414 (samotricisib) and IPI-549; adenosine pathway inhibitors (oral adenosine 2a and 2b receptor antagonists) such as AB-928; IDO inhibitors; anti-TIM, both TIM and TIM-3; LAG3; TGFβ inhibitors such as LY2157299 (galunisertib); and CD47 inhibitors such as Forty Seven's magrolimab (5F9).

[0305] Specific therapeutic agents for patients with cancer tumors assigned to the IS TME phenotype class by a classifier disclosed herein, e.g., the TME Panel 1 classifier, further include anti-TIGIT drugs, which are immunosuppressive by (among other activities) inducing CD155 (cluster of differentiation 155) on dendritic cells and the expression of Treg subsets in tumors. A preferred anti-TIGIT antibody is AB-154. Another preferred anti-TIGIT antibody is BGB-A1217 (osipellimab). Anti-activin A therapeutic agents are useful because activin A promotes the differentiation of M2-like tumor macrophages and suppresses the generation of NK cells. Anti-BMP therapeutic agents, like bone morphogenetic protein (BMP), are useful because they promote the differentiation of M2-like tumor macrophages and suppress CTLs and DCs.

[0306] Additional specific therapeutic agents for patients with cancer tumors assigned to the IS TME phenotype class by a classifier disclosed herein, e.g., the TME Panel 1 classifier, include TAM (Tyro3, ​​Axl, and Mer receptor) inhibitors or TAM product inhibitors; anti-IL-10 (interleukin) or anti-IL-10R (interleukin-10 receptor) (because IL-10 is immunosuppressive); anti-M-CSF (because macrophage colony-stimulating factor (M-CSF) antagonism has been shown to deplete TAMs); anti-CCL2 (CC motif chemokine ligand 2) or anti-CCL2R (CC motif chemokine ligand 2 receptor) (because targeting of specific pathways by these drugs recruits myeloid cells to tumors); MERTK (tyrosine protein kinase protein kinase Mer) antagonists (inhibition of this receptor tyrosine kinase induces a pro-inflammatory TAM phenotype and tumor CD8 + Further examples include (as it increases cell proliferation).

[0307] Other therapeutic agents for patients with cancer tumors assigned to the IS TME phenotype class by a classifier disclosed herein, e.g., the TME Panel 1 classifier, include STING agonists (stimulators of interferon genes that stimulate cytosolic DNA sensing) and anti-tumor CD8 +enhances DC stimulation of T cells, and agonists are part of STINGVAX®; CCL3 (CC motif chemokine 3), CCL4 (CC motif chemokine 4), CCL5 (CC motif chemokine 5), or their common receptor CCR5 (CC motif chemokine receptor type 5) (because these chemokines are products of myeloid-derived suppressor cells (MDSCs) and activate CCR5 on regulatory T cells (Tregs)); inhibitors of arginase-1 (because arginase-1 is produced by M2-like TAMs and reduces the production of tumor-infiltrating lymphocytes (TILs) and increases the production of Tregs); CCR Antibodies against CCR4 (CC motif chemokine receptor type 4) can be used to deplete Tregs; antibodies against CCL17 (CC motif chemokine 17) or CCL22 (CC motif chemokine 22) can inhibit CCR4 (CC motif chemokine receptor type 4) activation on Tregs; antibodies against GITR (glucocorticoid-induced TNFR-related protein) can be used to deplete Tregs; and inhibitors of DNA methyltransferases (DNMTs) or histone deacetylases (HDACs), such as entinostat, which reverse epigenetic silencing of immune genes.

[0308] In preclinical models, inhibitors of phosphodiesterase-5, sildenafil, and tadalafil significantly inhibited MDSC function, which may benefit patients with colorectal cancer tumors with an IS TME phenotype. All-trans retinoic acid (ATRA), used to differentiate MDSCs into mature dendritic cells (DCs) and macrophages, may benefit patients with an IS phenotype. VEGF and c-kit signaling have been reported to be involved in the generation of MDSCs. Sunitinib treatment of patients with metastatic renal cell carcinoma has been reported to reduce the number of circulating MDSCs, which may benefit patients with colorectal cancer tumors with an IS TME phenotype.

[0309] Cancer tumors classified into the IS TME phenotype class are a target population for bavituximab treatment in combination with checkpoint inhibitors such as anti-PD-1 (e.g., sintilimab, tislelizumab, pembrolizumab, or antigen-binding portions thereof), anti-PD-L1, or anti-CTLA-4, as the present disclosure notes that immune responses occurring in the presence of angiogenesis are indicative of immunosuppression, and bavituximab can restore immune activity to immunosuppressed cells.

[0310] For single-agent bavituximab to work, the ongoing immune response would have to be highly active to the extent that blocking immunosuppression would be sufficient to unleash the full potential of the patient's immune response. However, most advanced cancer patients need to maintain their immune response and may need to combine bavituximab with checkpoint inhibitors. Thus, the IS TME phenotype classes disclosed herein can be used to determine which colorectal cancer patients are likely to respond to bavituximab and checkpoint inhibitors. Bavituximab

[0311] Bavituximab is a PS-targeting antibody. Its binding to phosphatidylserine (PS) is mediated by the serum protein β2 glycoprotein 1 (β2GPI). β2GPI is also known as apolipoprotein H (Apo-H). Bavituximab has been used in clinical trials for breast cancer, liver cancer (hepatocellular carcinoma), malignant melanoma, colorectal cancer, and prostate cancer.

[0312] In some embodiments, bavituximab can be administered to a subject, for example, a patient with gastric cancer (e.g., locally advanced, metastatic, or previously untreated gastric cancer), breast cancer (e.g., locally advanced or metastatic HER2-negative breast cancer), prostate cancer (e.g., castration-resistant metastatic prostate cancer), liver cancer (e.g., hepatocellular carcinoma, such as advanced metastatic hepatocellular carcinoma), head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma), melanoma, colorectal cancer (e.g., advanced colorectal cancer metastatic to the liver), or ovarian cancer (e.g., platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer) according to the methods described herein. Sequences related to bavituximab are provided in the table below. [Table 9]

[0313] In some aspects, bavituximab is administered to cancer patients in combination with an anti-PD-1 antibody (e.g., sintilimab, tislelizumab, pembrolizumab, or an antigen-binding portion thereof). In some embodiments, bavituximab is administered in combination with pembrolizumab. In some embodiments, bavituximab is administered in combination with sintilimab. In some aspects, bavituximab is administered in combination with tislelizumab.

[0314] In some embodiments, bavituximab is administered to patients with gastric cancer (e.g., locally advanced or metastatic gastric cancer) as a combination therapy that includes a chemotherapeutic agent and an immune checkpoint inhibitor, e.g., an anti-PD-1 antibody (e.g., sintilimab, tislelizumab, pembrolizumab, or an antigen-binding portion thereof).

[0315] In some embodiments, bavituximab is administered to patients with liver cancer (e.g., advanced metastatic hepatocellular carcinoma) or head and neck cancer (e.g., recurrent or metastatic head and neck squamous cell carcinoma) as a combination therapy that includes an immune checkpoint inhibitor, e.g., an anti-PD-1 antibody (e.g., sintilimab, tislelizumab, pembrolizumab, or an antigen-binding portion thereof).

[0316] In some embodiments, bavituximab is administered to patients with melanoma as a combination therapy that includes radiation. III.C:A TME Phenotype Class-Specific Therapy

[0317] In the A TME phenotype class, where angiogenic activity is predominant, patients may respond to VEGF-targeted therapy, DLL4-targeted therapy, angiopoietin / TIE2-targeted therapy, anti-VEGF / anti-DLL4 bispecific antibodies such as nabicikizumab, and anti-VEGF or anti-VEGF receptor antibodies such as valisacumab, ramucirumab, and bevacizumab.

[0318] In some embodiments, dual variable domain immunoglobulin molecules, drugs, or therapies with anti-angiogenic effects, such as those with anti-DLL4 and / or anti-VEGF activity, can be selected for treating patients with cancers classified as having an A TME phenotype class, e.g., by applying the TME panel 1 classifier. In some embodiments, the dual variable domain immunoglobulin molecule, drug, or therapy is zilpakimab (ABT165). In some embodiments, dual targeting proteins, drugs, or therapies with anti-angiogenic effects, such as those with anti-DLL4 and / or anti-VEGF activity, can be selected for treating patients with cancers identified as having an A TME phenotype, e.g., by applying the TME panel 1 classifier. In some embodiments, the dual targeting protein, drug, or therapy is ABL001 (NOV1501, TR009), as taught by U.S. Patent Application Publication No. 2016 / 0159929, which is incorporated herein by reference in its entirety.

[0319] Bevacizumab Bevacizumab, sold under the trade name AVASTIN®, is a drug used to treat several types of cancer, such as colorectal, breast, or ovarian cancer. It is administered by slow infusion into a vein (intravenously). Bevacizumab is a monoclonal antibody that functions as an angiogenesis inhibitor. It slows the growth of new blood vessels by inhibiting vascular endothelial growth factor A (VEGF-A), i.e., as an anti-VEGF therapy. Bevacizumab was approved in the United States in 2004 for use in colorectal cancer with standard chemotherapy (as first-line treatment). In 2006, it was approved with 5-fluorouracil-based therapy for second-line treatment of metastatic colorectal cancer. It has also been approved by the European Medicines Agency (EMA) for use in colorectal cancer. Bevacizumab has also been tested as an add-on to other chemotherapy agents in people with nonmetastatic colon cancer. In the European Union, bevacizumab in combination with a fluoropyrimidine chemotherapy agent is indicated for the treatment of adults with metastatic carcinoma of the colon or rectum. In the EU, bevacizumab in combination with paclitaxel is indicated for the first-line treatment of adults with metastatic breast cancer. Bevacizumab in combination with capecitabine is indicated for the first-line treatment of adults with metastatic breast cancer when other chemotherapy options, including taxanes or anthracyclines, are not considered appropriate.

[0320] In 2018, the U.S. Food and Drug Administration (FDA) approved bevacizumab in combination with chemotherapy followed by single-agent bevacizumab for stage III or IV ovarian cancer after initial surgery. Approval was based on a study of the addition of bevacizumab to carboplatin and paclitaxel. Progression-free survival increased from 13 to 18 months. In the EU, bevacizumab in combination with carboplatin and paclitaxel is indicated for the frontline treatment of adults with advanced (World Federation of Gynecology and Obstetrics (FIGO) stages IIIB, IIIC, and IV) epithelial ovarian, fallopian tube, or primary peritoneal cancer. Bevacizumab in combination with carboplatin and gemcitabine, or carboplatin and paclitaxel, is indicated for the treatment of adults with first recurrence of platinum-sensitive epithelial ovarian, fallopian tube, or primary peritoneal cancer who have not been previously treated with bevacizumab or other VEGF inhibitors or VEGF receptor-targeted agents. In May 2020, the Food and Drug Administration expanded the labeling of olaparib to include in combination with bevacizumab for the first-line maintenance treatment of adults with advanced epithelial ovarian, fallopian tube, or primary peritoneal cancer who have had a complete or partial response to first-line platinum-based chemotherapy and whose cancer is associated with a deleterious or suspected deleterious BRCA mutation and / or homologous recombination deficiency-positive status defined by genomic instability.

[0321] Bevacizumab Biosimilars: The FDA and European Union have approved Amgen's biosimilar (generic name bevacizumab-awwb, trade name Mvasi), and Zirabev (Pfizer), Aybintio (Samsung Bioepis), and Equidacent (Centus Biotherapeutics) have been approved for use in the European Union. On January 28, 2021, the European Medicines Agency's (EMA) Committee for Medicinal Products for Human Use (CHMP) adopted a recommendation for the grant of a new drug recommendation and marketing authorization for the drug Alymsys (Mabxience Research) for the treatment of colorectal cancer. [Table 10]

[0322] Nabikizumab The anti-VEGF / anti-DLL4 bispecific antibody nabicikizumab is described in detail, for example, in U.S. Patent Nos. 9,376,488, 9,574,009, and 9,879,084, each of which is incorporated herein by reference in its entirety. [Table 11]

[0323] In some embodiments, nabiquizumab is administered to patients with gastric cancer (e.g., locally advanced or metastatic gastric cancer) as a combination therapy further including a chemotherapeutic agent (e.g., docetaxel, cabazitaxel, etc.) and an immune checkpoint inhibitor, e.g., an anti-PD-1 antibody (e.g., sintilimab, tislelizumab, pembrolizumab, or an antigen-binding portion thereof).

[0324] In some embodiments, nabiquizumab is administered to patients with gastric cancer (e.g., locally advanced or metastatic gastric cancer) or prostate cancer (e.g., castration-resistant metastatic prostate cancer) ...

Claims

1. 1. A method for identifying a human subject suffering from cancer suitable for treatment with a TME phenotype class-specific therapy, comprising applying an artificial neural network (ANN) classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject, wherein the cancer tumor is assigned to a TME phenotype class selected from the group consisting of IS (immunosuppressed), A (angiogenic), IA (immunoactive), ID (immune desert), and combinations thereof, wherein the assigned TME phenotype class indicates that a TME phenotype class-specific therapy can be administered to treat the cancer.

2. The ANN classifier is (a) an input layer comprising 2-100 nodes, each node in the input layer corresponding to a gene in a gene panel selected from the genes shown in Table 1 and Table 2, the gene panel comprising: (i) 1-63 genes selected from Table 1 and 1-61 genes selected from Table 2, (ii) a gene panel comprising genes selected from Table 3 and Table 4, (iii) a gene panel of Table 5, or (iv) any of the gene panels (gene sets) disclosed in Figures 9A-G; (b) a hidden layer containing two nodes; and (c) an output layer including four output nodes, each of the four output nodes in the output layer corresponding to a TME phenotype class, the four TME phenotype classes being IA, IS, ID, and A; and 2. The method of claim 1, further comprising optionally applying a logistic regression classifier including a softmax function to the output of the ANN, the softmax function assigning a probability to each TME phenotype class.

3. The TME phenotype class specific therapy comprises: (i) activators of stimulatory immune checkpoint molecules, such as antibody molecules against GITR, OX-40, ICOS, 4-1BB, or combinations thereof; (ii) a RORγ agonist; (iii) Inhibitory antibodies against PD-1 (e.g., nivolumab, pembrolizumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, TSR-042, or antigen-binding portions thereof), antibodies against PD-L1 (e.g., avelumab, atezolizumab, durvalumab, CX-072, LY3300054, or antigen-binding portions thereof), antibodies against PD-L2, or antibodies against CTLA-4. Inhibitors of immune checkpoint molecules, alone or in combination, or in combination with inhibitors of TIM-3, LAG-3, BTLA, TIGIT, VISTA, TGF-β, LAIR1, CD160, 2B4, GITR, OX40, 4-1BB, CD2, CD27, CDS, ICAM-1, LFA-1, ICOS, CD30, CD40, BAFFR, HVEM, CD7, LIGHT, NKG2C, SLAMF7, NKp80, or CD86; or (iv) combinations thereof, 3. The method of claim 1 or 2, wherein the IA TME phenotype class-specific therapy comprises a checkpoint modulator therapy, comprising administering

4. The TME phenotype class specific therapy comprises: (1) checkpoint modulator therapy and anti-immunosuppressant therapy, and / or (2) antiangiogenic therapy, and administering to said patient a dose of at least one of said TME classes. The checkpoint modulator therapy includes: (a) an antibody against PD-1 selected from the group consisting of pembrolizumab, nivolumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, TSR-042, antigen-binding portions thereof, and combinations thereof; (b) an antibody against PD-L1 selected from the group consisting of avelumab, atezolizumab, CX-072, LY3300054, durvalumab, antigen-binding portions thereof, and combinations thereof; (c) an antibody against PD-L2, or an antigen-binding portion thereof; (d) an antibody against CTLA-4 selected from ipilimumab and the bispecific antibody XmAb20717 (anti-PD-1 / anti-CTLA-4); or (e) administering an inhibitor of an inhibitory immune checkpoint molecule, including any of these combinations; The antiangiogenic therapy comprises: (a) an anti-VEGF antibody selected from the group consisting of valisacumab, bevacizumab, nabicikizumab (anti-DLL4 / anti-VEGF bispecific), ABL101 (NOV1501) (anti-DLL4 / anti-VEGF), ABT165 (anti-DLL4 / anti-VEGF), and combinations thereof; (b) an anti-VEGFR2 antibody, including ramucirumab; or (c) administering the combination thereof; and The anti-immunosuppressant therapy comprises: (a) an anti-PS antibody, an anti-PS targeting antibody, an antibody that binds β2 glycoprotein 1, an inhibitor of PI3Kγ, an adenosine pathway inhibitor, an inhibitor of IDO, an inhibitor of TIM, an inhibitor of LAG3, an inhibitor of TGFβ, a CD47 inhibitor, or a combination thereof; the anti-PS targeting antibody is bavituximab or an antibody that binds to β2 glycoprotein 1; the PI3Kγ inhibitor is LY3023414 (samotricisib) or IPI-549; the adenosine pathway inhibitor is AB-928; the TGFβ inhibitor is LY2157299 (galunisertib) or the TGFβR1 inhibitor is LY3200882; the CD47 inhibitor is magrolimab (5F9); and the CD47 inhibitor targets SIRPα; (b) TIM-3, LAG-3, BTLA, TIGIT, VISTA, an inhibitor of TGF-β or its receptor, LAIR1, CD160, 2B4, GITR, OX40, 4-1BB, CD2, CD27, CDS, ICAM-1, LFA-1, ICOS, CD30, CD40, BAFFR, HVEM, CD7, LIGHT, NKG2C, SLAMF7, an inhibitor of NKp80, an agonist of CD86, or a combination thereof; or 3. The method of claim 1 or 2, comprising administering: (c) a combination thereof.

5. The TME phenotype class specific therapy comprises: (i) VEGF targeted therapy, inhibitors of angiopoietin 1 (Ang1), inhibitors of angiopoietin 2 (Ang2), inhibitors of DLL4, bispecific antibodies of anti-VEGF and anti-DLL4, TKI inhibitors, anti-FGF antibodies, anti-FGFR1 antibodies, anti-FGFR2 antibodies, small molecules that inhibit FGFR1, small molecules that inhibit FGFR2, anti-PLGF antibodies, small molecules against the PLGF receptor, antibodies against the PLGF receptor, anti-VEGFB antibodies, anti-VEGFC antibodies, anti-VEGFD antibodies, antibodies against VEGF / PLGF trapping molecules such as aflibercept or ziv-aflibercept, anti-DLL4 antibodies, inhibitors of gamma secretase, etc. Which anti-Notch therapy, or any combination thereof, wherein the TKI inhibitor is selected from the group consisting of cabozantinib, vandetanib, tivozanib, axitinib, lenvatinib, sorafenib, regorafenib, sunitinib, fruquitinib, pazopanib, and any combination thereof, and said VEGF targeted therapy comprises administering (a) an anti-VEGF antibody, including valisacumab, bevacizumab, an antigen binding portion thereof, or a combination thereof; (b) an anti-VEGFR2 antibody, including ramucirumab, or an antigen binding portion thereof; or (c) a combination thereof; (ii) angiopoietin / TIE2 targeted therapy, including endoglin and / or angiopoietin; or 3. The method of claim 1 or 2, wherein the method is an ATME phenotype class-specific therapy comprising administering (iii) a DLL4 targeted therapy comprising nabicikizumab, ABL101 (NOV1501), ABT165, or a combination thereof.

6. The TME phenotype class specific therapy comprises: an ID TME phenotype class specific therapy comprising administering a checkpoint modulator therapy concomitantly with or subsequent to administration of a therapy that initiates an immune response, said checkpoint modulator therapy comprising administration of an inhibitor of an inhibitory immune checkpoint molecule, such as an antibody against PD-1, PD-L1, PD-L2, CTLA-4, or a combination thereof, and said therapy that initiates an immune response is a vaccine, CAR-T, or a neo-epitope vaccine; (i) the anti-PD-1 antibody comprises nivolumab, pembrolizumab, cemiplimab, PDR001, CBT-501, CX-188, sintilimab, tislelizumab, or TSR-042, or an antigen-binding portion thereof; (ii) the anti-PD-L1 antibody comprises avelumab, atezolizumab, CX-072, LY3300054, durvalumab, or an antigen-binding portion thereof; or (iii) The method of claim 1 or 2, wherein the anti-CTLA-4 antibody comprises ipilimumab or the bispecific antibody XmAb20717 (anti-PD-1 / anti-CTLA-4), or an antigen-binding portion thereof.

7. The method of any one of claims 1 to 6, wherein the cancer is recurrent, refractory, metastatic, dMMR, or a combination thereof.

8. The cancer is (i) gastric cancer, such as locally advanced, metastatic, or previously untreated gastric cancer; (ii) breast cancer, such as locally advanced, triple-negative breast cancer, or metastatic HER2-negative breast cancer; (iii) prostate cancer, such as castration-resistant metastatic prostate cancer; (iv) liver cancer, such as advanced metastatic hepatocellular carcinoma; (v) Head and neck cancer, such as recurrent or metastatic head and neck squamous cell carcinoma; (vi) melanoma, including metastatic melanoma; (vii) colorectal cancer, including advanced colorectal cancer metastatic to the liver; (viii) ovarian cancer, such as platinum-resistant ovarian cancer or platinum-sensitive recurrent ovarian cancer; (ix) glioma, such as metastatic glioma; (x) lung cancer, such as non-small cell lung cancer (NSCLC); and (xi) glioblastoma.

9. Administering a TME phenotype class-specific therapy (i) reducing the cancer burden by at least about 10%, 20%, 30%, 40%, or 50% compared to the cancer burden before said administration; (ii) progression-free survival of at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 months, or at least about 1, 2, 3, 4, or 5 years, following a first administration of said TME phenotype class-specific therapy; (iii) stable disease for about 1 month, about 2 months, about 3 months, about 4 months, about 5 months, about 6 months, about 7 months, about 8 months, about 9 months, about 10 months, about 11 months, about 1 year, about 18 months, about 2 years, about 3 years, about 4 years, or about 5 years after the first administration of the TME phenotype class-specific therapy; (iv) a partial response at about 1 month, about 2 months, about 3 months, about 4 months, about 5 months, about 6 months, about 7 months, about 8 months, about 9 months, about 10 months, about 11 months, about 1 year, about 18 months, about 2 years, about 3 years, about 4 years, or about 5 years after the first administration of the TME phenotype class-specific therapy; (v) a complete response at about 1 month, about 2 months, about 3 months, about 4 months, about 5 months, about 6 months, about 7 months, about 8 months, about 9 months, about 10 months, about 11 months, about 1 year, about 18 months, about 2 years, about 3 years, about 4 years, or about 5 years after the first administration of the TME-phenotype class-specific therapy; (vi) an improvement in progression free survival probability of at least about 10%, at least about 20%, at least about 30%, at least about 40%, at least about 50%, at least about 60%, at least about 70%, at least about 80%, at least about 90%, at least about 100%, at least about 110%, at least about 120%, at least about 130%, at least about 140%, or at least about 150%, compared to the progression free survival probability of a subject who did not receive a TME phenotype class-specific therapy assigned using an ANN classifier, such as TME Panel-1; (vii) an improvement in overall survival probability of at least about 25%, at least about 50%, at least about 75%, at least about 100%, at least about 125%, at least about 150%, at least about 175%, at least about 200%, at least about 225%, at least about 250%, at least about 275%, at least about 300%, at least about 325%, at least about 350%, or at least about 375% compared to the overall survival probability of a subject who did not receive a TME phenotype class-specific therapy assigned using an ANN classifier, such as TME Panel-1; or (viii) a combination thereof.

10. 1. A method of assigning a TME phenotype class to cancer in a subject in need of treatment, comprising: (i) generating an ANN classifier by training the ANN on a training set comprising RNA expression levels for each gene in a gene panel of a plurality of samples from a plurality of subjects, where each sample is assigned a TME phenotype classification; and assigning a TME phenotype class to cancer in the subject using the ANN classifier, wherein input to the ANN classifier includes RNA expression levels for each gene in the gene panel of a test sample from the subject; (ii) generating an ANN classifier by training an ANN with a training set comprising RNA expression levels for each gene in a gene panel of a plurality of samples from a plurality of subjects, each sample being assigned a TME phenotype classification; the ANN classifier using as input the RNA expression levels for each gene in the gene panel of test samples from the subjects to assign a TME phenotype class to the cancer of the subject; or (iii) predicting the TME phenotype class of the subject's cancer using an ANN classifier, wherein the ANN classifier is generated by training an ANN with a training set comprising RNA expression levels for each gene in a gene panel of multiple samples from multiple subjects, and each sample is assigned to a TME phenotype class or a combination thereof.

11. 11. The method of claim 10, wherein the method is implemented in a computer system including at least one processing device and at least one storage device, the at least one storage device including instructions that, when executed by the at least one processing device, cause the at least one processing device to implement a machine learning model.

12. The method according to claim 1, 12. The method of claim 11, further comprising: (i) inputting the ANN classifier code into a storage device of the computer system; (ii) inputting the gene panel input data corresponding to the subject into a storage device of the computer system, the input data comprising RNA expression levels; (iii) executing the ANN classifier code; or (v) any combination thereof. (i) the cancer is locally advanced, metastatic gastric cancer and the TME phenotype is IA, A, or IS; (ii) the cancer is untreated gastric cancer and the TME phenotype is IS or A; (iii) the cancer is advanced / metastatic HER2-negative breast cancer and the TME phenotype is A or IS; (iv) the cancer is castration-resistant metastatic prostate cancer and the TME phenotype is A or IS; (v) the cancer is advanced metastatic hepatocellular carcinoma and the TME phenotype is IA or IS; (vi) the cancer is recurrent / metastatic head and neck squamous cell carcinoma and the TME phenotype is IA or IS; (vii) the cancer is melanoma and the TME phenotype is IA or IS; (viii) the cancer is advanced colorectal cancer metastatic to the liver and the TME phenotype is ID; (ix) the cancer is platinum-resistant or platinum-sensitive recurrent ovarian cancer and the TME phenotype is IA, IS, or A; (x) the cancer is platinum-resistant or platinum-sensitive recurrent triple-negative breast cancer and the TME phenotype is IA, IS, or A; (xi) the cancer is metastatic colorectal cancer and the TME phenotype is A or IS; (xii) the cancer is glioma or glioblastoma and the TME phenotype is IS or IA; (xiii) the cancer is non-small cell lung cancer and the TME phenotype is IS or IA; The TME phenotype class is assigned by applying an ANN classifier to a plurality of RNA expression levels obtained from a gene panel from a cancer tumor sample from the subject, the ANN classifier comprising: (a) an input layer comprising 2-100 nodes, each node in the input layer corresponding to a gene in a gene panel selected from the genes shown in Table 1 and Table 2, the gene panel comprising: (i) 1-63 genes selected from Table 1 and 1-61 genes selected from Table 2, (ii) a gene panel comprising genes selected from Table 3 and Table 4, (iii) a gene panel of Table 5, or (iv) any of the gene panels (gene sets) disclosed in Figures 9A-G; (b) a hidden layer containing two nodes; and (c) an output layer including four output nodes, each of the four output nodes in the output layer corresponding to a TME phenotype class, the four TME phenotype classes being IA, IS, ID, and A; and 13. The method of any one of claims 1 to 12, optionally further comprising a logistic regression classifier comprising a softmax function on the output of the ANN, said softmax function assigning a probability to each TME phenotype class.

14. 9A-9G), and (ii) a plurality of oligonucleotide probes capable of specifically detecting RNA encoding the genetic biomarkers of Table 2 (or FIGS. 9A-9G), wherein the kit or article of manufacture comprises a microarray.

15. A computer system comprising at least one processing device and at least one storage device, wherein the at least one storage device contains instructions that, when executed by the at least one processing device, cause the at least one processing device to implement a machine learning model, and the at least one processing device executes the instructions to implement a method according to any one of claims 1 to 13.

16. A computer-readable medium having computer-readable program code embedded therein for executing an application program on a computer having a database to implement a method according to any one of claims 1 to 13.

17. A computer program product comprising a computer readable medium having computer readable program code embedded in the medium for executing an application program on a computer having a database to implement a method according to any one of claims 1 to 13.

Images