Methods for determining cetuximab sensitivity in cancer patients

Abstract

The invention described in this disclosure relates to a panel of biomarkers for determining the response of cancer patients who have been or will be treated with cetuximab.The disclosure provides methods and compositions, such as kits, for evaluating biomarkers, and methods for using such biomarkers to predict the response of cancer patients to cetuximab.Such information can be used to determine the prognosis and treatment options of gastric cancer patients.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to International Application No. PCT / CN2022 / 078536, filed March 1, 2022, the disclosure of which is incorporated herein by reference.

[0002] The present invention relates generally to the diagnosis and treatment of cancer. [Background technology]

[0003] The epidermal growth factor receptor (EGFR, also called ErbB-1, HER1 in humans) is a transmembrane receptor protein of the ErbB family of receptors, which includes four closely related receptor tyrosine kinases: EGFR (ErbB-1), HER2 / neu (ErbB-2), Her3 (ErbB-3) and Her4 (ErbB-4). EGFR is activated by binding of ligands such as epidermal growth factor or transforming growth factor-alpha, leading to homodimerization or heterodimerization with other members of the EGFR family. This receptor activation is followed by phosphorylation of specific tyrosine residues in the cytoplasmic tail, stimulating downstream signaling pathways that control cell proliferation, migration, adhesion, differentiation and survival.

[0004] Genetic amplification, mutated activation, and / or protein overexpression of EGFR have been observed in various solid tumors, including lung cancer, colorectal cancer, bladder cancer, breast cancer, head and neck cancer, esophageal cancer, and gastric cancer. In some tumors, such as non-small cell lung cancer and colorectal cancer, increased EGFR expression is associated with advanced stage and poor prognosis. Over the past 20 years, many EGFR-targeted drugs have been developed, particularly tyrosine kinase inhibitors (TKIs), anti-EGFR antibodies, and antibody-drug conjugates (ADCs).

[0005] Cetuximab (brand name Erbitux™ in the United States and Canada) is a recombinant human / mouse chimeric monoclonal antibody against EGFR and is the oldest EGFR-targeted monoclonal antibody drug. Cetuximab has been approved to treat EGFR-expressing metastatic colorectal cancer (mCRC) and squamous cell carcinoma of the head and neck (SCCHN) without activating KRAS mutations. However, clinical trials of cetuximab for other types of cancer, including non-small cell lung cancer (NSCLC), gastric cancer, and esophageal cancer, have failed in late stages of clinical trials.

[0006] Clinical responses to anticancer therapy are often limited to a subset of patients. Nonclinical and clinical trial data clearly show that there is a significant responding population among cancer types not approved for cetuximab treatment. To maximize the efficiency of anticancer therapy using cetuximab, biomarker-based patient stratification has been proposed. However, the identification of predictive biomarkers that can predict response to cetuximab remains a challenge. For example, EGFR as a single gene biomarker may not be useful under at least two circumstances, namely, when the expression level of EGFR is in the moderate range or when the EGFR gene has deleterious mutations. Therefore, there is a need to identify new biomarkers to accurately predict response to cetuximab in cancer patients. Summary of the Invention

[0007] In one aspect, the present disclosure provides a method for predicting the sensitivity of cetuximab in patients with cancer.In some embodiments, the method includes measuring a set of biomarkers selected from EGFR expression level, TMEM40 expression level, IL1A expression level, PTPRN2 expression level, LCE2A expression level, TREM2 expression level, LY6D expression level, TMEM63B expression level, EIF4EBP1 expression level, C20orf56 expression level, SHC expression level, DSG3 expression level, HES6 expression level, FAM25B expression level, PNMA2 expression level, GSK3B expression level, PPM1H expression level, TOX3 expression level, TYMP expression level, Anxa8L2 expression level, ACP6 expression level, KRAS mutation, APC mutation, MACF1 mutation, NCOR2 mutation, LPP mutation, and combinations thereof in the tumor sample of the patient, and using a machine learning classifier to determine the likelihood that the patient is responsive to cetuximab based on the set of measured biomarkers.

[0008] In some embodiments, the cancer is selected from colon cancer, gastric cancer, lung cancer, head and neck cancer, and esophageal cancer.

[0009] In one aspect, the present disclosure provides a method for predicting the sensitivity of cetuximab in patients with colon cancer.In some embodiments, the method includes measuring a set of biomarkers in the tumor sample of the patient, comprising EGFR expression level, GSK3B expression level, KRAS mutation, LY6D expression level, PNMA2 expression level, C20orf56 expression level, MACF1 mutation and NCOR2 mutation, and using a machine learning classifier to determine the likelihood that the patient is responsive to cetuximab based on the set of biomarkers measured.

[0010] In one aspect, the present disclosure provides a method for predicting the cetuximab sensitivity in patients with gastric cancer.In some embodiments, the method includes measuring a set of biomarkers in the tumor sample of the patient, comprising LPP mutation, EHBP1L1 expression level, EGFR expression level, LY6D expression level, C20orf56 expression level, PTPRN2 expression level, FMOD expression level and NCOR2 mutation, and using a machine learning classifier to determine the likelihood that the patient is responsive to cetuximab based on the set of biomarkers measured.

[0011] In one aspect, the present disclosure provides a method for predicting the cetuximab sensitivity in patients with lung cancer.In some embodiments, the method includes measuring a set of biomarkers in the tumor sample of the patient, comprising LPP mutation, FMOD expression level, EGFR expression level, GSK3B expression level, FAM25B expression level, SHC3 expression level, IL1A expression level, S100A7A expression level, PTPRN2 expression level and AKT3 expression level, and using a machine learning classifier to determine the likelihood that the patient is responsive to cetuximab based on the set of biomarkers measured.

[0012] In one aspect, the disclosure provides a method for predicting cetuximab sensitivity in a patient having head and neck cancer, comprising measuring a set of biomarkers in a tumor sample from the patient, the set including SHC3 expression level, LPP mutation, HES6 expression level, S100A7A expression level, GSK3B expression level and FAM25B expression level, and using a machine learning classifier to determine the likelihood that the patient will be responsive to cetuximab based on the set of measured biomarkers.

[0013] In one aspect, the present disclosure provides a method for predicting the cetuximab sensitivity in patients with esophageal cancer.In some embodiments, the method includes measuring a set of biomarkers in the tumor sample of the patient, comprising LPP mutation, EGFR expression level, FMOD expression level, LY6D expression level, FAM25B expression level, PNMA2 expression level, TOX3 expression level and PTPRN2 expression level, and using a machine learning classifier to determine the likelihood that the patient is responsive to cetuximab based on the set of biomarkers measured.

[0014] In some embodiments, the LPP mutation is selected from R22Q, S123Y, P136S, A174T, and G379E.

[0015] In some embodiments, the biomarkers are measured by an amplification assay, a hybridization assay, a sequencing assay, or an array.

[0016] In some embodiments, the machine learning classifier is constructed by regularized regression methods.

[0017] In some embodiments, the methods disclosed herein further comprise administering cetuximab to the patient.

[0018] In one aspect, the disclosure provides a non-transitory computer readable medium having instructions stored thereon, which, when executed by a processor, cause the processor to retrieve data of a set of biomarkers obtained from a tumor sample of a patient having cancer, the set of biomarkers being EGFR expression level, TMEM40 expression level, IL1A expression level, PTPRN2 expression level, LCE2A expression level, TREM2 expression level, LY6D expression level, TMEM63B expression level, EIF4EBP1 expression level, C20orf56 expression level, and IL1A expression level. a search for a set of biomarkers selected from the group consisting of a marker for the genomic DNA of a patient with a genomic DNA sequence selected from the group consisting of a marker for the genomic DNA sequence ...

[0019] In one aspect, the disclosure provides a non-transitory computer readable medium having instructions stored thereon, which, when executed by a processor, cause the processor to: search data of a set of biomarkers obtained from a tumor sample of a patient having colon cancer, the set of biomarkers comprising KRAS mutations, LY6D expression levels, PNMA2 expression levels, C20orf56 expression levels, MACF1 mutations, EGFR expression levels, GSK3B expression levels, and NCOR2 mutations; and determine, using a machine learning classifier, a likelihood that the patient will be responsive to cetuximab based on the data of the set of biomarkers.

[0020] In one aspect, the disclosure provides a non-transitory computer readable medium having instructions stored thereon, which, when executed by a processor, cause the processor to: search data of a set of biomarkers obtained from a tumor sample of a patient having gastric cancer, the set of biomarkers including LPP mutations, EHBP1L1 expression level, EGFR expression level, LY6D expression level, C20orf56 expression level, PTPRN2 expression level, FMOD expression level, and NCOR2 mutations; and determine, using a machine learning classifier, a likelihood that the patient will be responsive to cetuximab based on the data of the set of biomarkers.

[0021] In one aspect, the disclosure provides a non-transitory computer readable medium having instructions stored thereon, which, when executed by a processor, cause the processor to: search data of a set of biomarkers obtained from a tumor sample of a patient having lung cancer, the set of biomarkers including LPP mutations, FMOD expression level, EGFR expression level, GSK3B expression level, FAM25B expression level, SHC3 expression level, IL1A expression level, S100A7A expression level, PTPRN2 expression level, and AKT3 expression level; and determine, using a machine learning classifier, a likelihood that the patient will be responsive to cetuximab based on the data of the set of biomarkers.

[0022] In one aspect, the disclosure provides a non-transitory computer readable medium having instructions stored thereon, which, when executed by a processor, cause the processor to: search data of a set of biomarkers obtained from a tumor sample of a patient having head and neck cancer, the set of biomarkers including SHC3 expression level, LPP mutation, HES6 expression level, S100A7A expression level, GSK3B expression level, and FAM25B expression level; and determine, using a machine learning classifier, a likelihood that the patient will be responsive to cetuximab based on the data of the set of biomarkers.

[0023] In one aspect, the disclosure provides a non-transitory computer readable medium having instructions stored thereon, which, when executed by a processor, cause the processor to: search data of a set of biomarkers obtained from a tumor sample of a patient having esophageal cancer, the set of biomarkers including LPP mutations, EGFR expression level, FMOD expression level, LY6D expression level, FAM25B expression level, PNMA2 expression level, TOX3 expression level, and PTPRN2 expression level; and determine, using a machine learning classifier, a likelihood that the patient will be responsive to cetuximab based on the data of the set of biomarkers.

[0024] In one aspect, the present disclosure provides a method for generating a machine learning model for predicting sensitivity to a drug in a patient with cancer, the method includes obtaining whole genome expression levels from each of a group of tumor models tested for responsiveness to the drug, selecting a first group of genes whose expression levels are increased in the tumor models that respond to the drug compared to the tumor models that do not respond to the drug, selecting a second group of genes whose expression levels are decreased in the tumor models that respond to the drug compared to the tumor models that do not respond to the drug, using a regularized regression method to select a set of biomarkers from the first and second groups of genes, and using a logistic regression model to build a machine learning classifier. In some embodiments, the drug is cetuximab. In some embodiments, the tumor model is a xenograft model.

[0025] In some embodiments, the first and second groups of genes are selected by correlation between gene expression levels and AUCr or by model performance of the ROC index.

[0026] The following drawings form part of the present specification and are included to further demonstrate certain aspects of the present disclosure. The present disclosure may be better understood by reference to one or more of these drawings in combination with the detailed description of specific embodiments presented herein. [Brief description of the drawings]

[0027] [Figure 1] Figure 1 shows that adding interaction terms between EGFR and other genes greatly improves model fit. [Diagram 2] FIG. 2 shows that EGFR pathway genes are important for building a model to predict cetuximab responsiveness in colon cancer patients. [Diagram 3] FIG. 3 shows that EGFR pathway genes are important for building a model to predict cetuximab responsiveness in head and neck cancer patients. [Figure 4] FIG. 4 shows a machine learning model for predicting cetuximab responsiveness in colon cancer. [Diagram 5] Figure 5 shows a machine learning model for predicting cetuximab responsiveness in gastric cancer. [Figure 6] FIG. 6 shows a machine learning model for predicting cetuximab responsiveness in lung cancer. [Figure 7] FIG. 7 shows a machine learning model for predicting cetuximab responsiveness in head and neck cancer. [Figure 8] FIG. 8 shows a machine learning model for predicting cetuximab responsiveness in esophageal cancer. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0028] Before describing the present disclosure in more detail, it is to be understood that the present disclosure is not limited to particular embodiments described, as such may, of course, vary. It is also to be understood that the terminology used herein is for the purpose of describing particular embodiments only, and is not intended to be limiting, since the scope of the present disclosure will be limited only by the appended claims.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. Although any methods and materials similar or equivalent to those described herein can be used in the practice or testing of this disclosure, the preferred methods and materials are now described.

[0030] All publications and patents cited herein are incorporated by reference as if each individual publication or patent was specifically and individually indicated to be incorporated by reference, and are incorporated by reference herein to disclose and describe the methods and / or materials in connection with which the publications are cited. The citation of any publication is for its disclosure prior to the date of the present application and should not be construed as an admission that the disclosure is not entitled to antedate such publication by virtue of prior disclosure. Further, the dates of publication provided may be different from the actual publication dates which may need to be independently confirmed.

[0031] As will be apparent to one of skill in the art upon reading this disclosure, each of the individual embodiments described and illustrated herein has distinct components and characteristics which may be readily separated from or combined with the characteristics of any of the other several embodiments without departing from the scope or spirit of the disclosure. Any recited method can be carried out in the order of events recited or in any other order which is logically possible.

[0032] definition

[0033] The following definitions are provided to assist the reader. Unless otherwise specified, all technical terms, notations and other scientific or medical terms or terminology used herein shall have the meanings that are commonly understood by those skilled in the chemical and medical arts. In some cases, terms that have commonly understood meanings are defined herein for clarity and / or ease of reference, and the inclusion of such definitions herein should not necessarily be interpreted as representing a substantial difference in the definition of the terms as they are commonly understood in the art.

[0034] As used herein, the singular forms "a," "an," and "the" include plural references unless the context clearly dictates otherwise.

[0035] The term "amount" or "level" refers to the amount of a polynucleotide or polypeptide of interest present in a sample. Such amount may be expressed in absolute terms, i.e., the total amount of polynucleotide or polypeptide in the sample, or in relative terms, i.e., the concentration of the polynucleotide or polypeptide in the sample.

[0036] As used herein, the term "cancer" refers to any disease involving abnormal cell proliferation, and includes all stages and forms of disease that affect any tissue, organ or cell in the body.This term includes all known cancers and neoplastic conditions characterized as malignant, benign, soft tissue or solid, as well as all stages and grades of cancer, including pre-metastatic and post-metastatic cancer.Generally, cancer can be classified according to the tissue or organ in which it is located or originates, and the morphology of cancerous tissue and cells. As used herein, types of cancer include, but are not limited to, acute lymphocytic leukemia (ALL), acute myeloid leukemia, adrenocortical carcinoma, anal cancer, astrocytoma, childhood cerebellar or cerebral basal cell carcinoma, bile duct cancer, bladder cancer, bone tumors, brain cancer, cerebellar astrocytoma, cerebral astrocytoma / malignant glioma, ependymoma, medulloblastoma, supratentorial primitive neuroectodermal tumor, visual pathway and hypothalamic glioma, breast cancer, Burkitt's lymphoma, cervical cancer, chronic lymphocytic leukemia, chronic myeloid leukemia, colon cancer, emphysema, endometrial cancer, ependymoma, esophageal cancer, and pulmonary tumors. Cancer, Ewing's sarcoma, retinoblastoma, gastric (stomach) cancer, glioma, head and neck cancer, heart cancer, Hodgkin's lymphoma, islet cell carcinoma (endocrine pancreas), Kaposi's sarcoma, kidney cancer (renal cell carcinoma), pharyngeal cancer, leukemia, liver cancer, lung cancer, neuroblastoma, non-Hodgkin's lymphoma, ovarian cancer, pancreatic cancer, pharyngeal cancer, prostate cancer, rectal cancer, renal cell carcinoma (kidney cancer), retinoblastoma, Ewing's family of tumors, skin cancer, stomach cancer, testicular cancer, pharyngeal cancer, thyroid cancer, and vaginal cancer.

[0037] It should be noted that in this disclosure, terms such as "comprises," "comprised," "comprising," "contains," "containing," and the like have the meanings ascribed to them in U.S. patent law, which are inclusive or open-ended and do not exclude additional, unrecited elements or method steps. Terms such as "consisting essentially of" and "consists essentially of" have the meanings ascribed to them in U.S. patent law, which permit the inclusion of additional components or steps that do not materially affect the basic and novel characteristics of the claimed invention. The terms "consist of" and "consisting of" have the meanings ascribed to them in U.S. patent law, i.e., these terms are closed-ended.

[0038] As used herein, a "cell" may be a prokaryotic or eukaryotic cell. Prokaryotic cells include, for example, bacteria. Eukaryotic cells include, for example, fungi, plant cells, and animal cells.Types of animal cells (e.g., mammalian cells or human cells) include, for example, cells of the circulatory / immune system or organs (e.g., B cells, T cells (cytotoxic T cells, natural killer T cells, regulatory T cells, helper T cells), natural killer cells, granulocytes (e.g., basophilic granulocytes, eosinophilic granulocytes, neutrophilic granulocytes, and hypersegmented neutrophils), monocytes or macrophages, red blood cells (e.g., reticulocytes), mast cells, platelets or megakaryocytes, and dendritic cells), cells of the endocrine system or organs (e.g., thyroid cells (e.g., thyroid epithelial cells, parafollicular cells, and thyroid cells), and cells of the endocrine system or organs (e.g., thyroid cells (e.g., thyroid epithelial cells, parafollicular cells, and thyroid cells). cells), parathyroid cells (e.g. parathyroid chief cells, eosinophil cells), adrenal cells (e.g. chromaffin cells) and cells of the pineal gland (e.g. pinealocytes)), cells of the nervous system or organs (e.g. glioblasts (e.g. astrocytes and oligodendrocytes), microglia, giant cell neurosecretory cells, astrocytes, Bechtel cells and pituitary cells (e.g. gonadotropes, corticotropes, thyrotropes, somatotropes and lactotropes)), cells of the respiratory system or organs (e.g. lung alveolar cells (type I pneumocytes and type II pneumocytes), Clara cells, goblet cells, alveolar macrophages), cells of the circulatory system or organs (e.g., cardiac myocytes and peripheral cells), cells of the digestive system or organs (e.g., gastric chief cells, parietal cells, goblet cells, Paneth cells, G cells, D cells, ECL cells, I cells, K cells, S cells, enteroendocrine cells, enterochromaffin cells, APUD cells, cells of the liver (e.g., hepatocytes and Kupffer cells)), cells of the envelope system or organs (e.g., bone cells (e.g., osteoblasts, osteocytes, and osteoclasts), dental cells (e.g., cerebrospinal fluid cells, osteoblasts, osteoblasts, osteoclasts, osteoblast ... cells of the urinary system or organs (e.g., podocytes, juxtaglomerular cells, intraglomerular mesangial cells, extraglomerular mesangial cells, renal proximal tubule brush border cells and macula densa cells) and cells of the reproductive system or organs (e.g., sperm, Sertoli cells, Leydig cells, eggs, oocytes).The cells may be normal, healthy cells or diseased, unhealthy cells (e.g., cancer cells). The cells further include stem cells, including mammalian zygotic or embryonic stem cells, fetal stem cells, induced pluripotent stem cells, and adult stem cells. Stem cells are cells that can undergo cycles of cell division while maintaining an undifferentiated state and differentiate into specialized cell types. Stem cells may be totipotent, pluripotent, multipotent, oligopotent, and unipotent stem cells, all of which can be derived from somatic cells. Stem cells may also include cancer stem cells. The mammalian cells may be rodent cells, e.g., mouse, rat, hamster cells. The mammalian cells may be lagomorph cells, e.g., rabbit cells. The mammalian cells may also be primate cells, e.g., human cells. In certain examples, the cells are cells used in mass bioproduction, e.g., CHO cells.

[0039] The terms "determine," "assess," "assay," "measure," and "detect" can be used interchangeably and refer to both quantitative and semi-quantitative determinations. When referring to either quantitative and semi-quantitative determinations, the phrase "determine the level" of a polynucleotide or polypeptide of interest or "detect" a polynucleotide or polypeptide of interest can be used.

[0040] The term "gene product" or "gene expression product" means an RNA or protein encoded by a gene.

[0041] The term "hybridization" refers to the binding, duplex formation or hybridization of nucleic acid molecules preferentially to a specific nucleotide sequence under stringent conditions. The term "stringent conditions" refers to conditions under which a probe hybridizes preferentially to its target subsequence in a mixed population (e.g., cell lysate or DNA preparation from tissue biopsy) and hybridizes to a lesser extent or not at all to other sequences. In the case of nucleic acid hybridization (e.g., as in array, microarray, Southern or Northern hybridization), "stringent hybridization" and "stringent hybridization washing conditions" are sequence-dependent and are different under different environmental parameters. Extensive guidelines for nucleic acid hybridization are found, for example, in Tijssen Laboratory Techniques in Biochemistry and Molecular Biology-Hybridization with Nucleic Acid Probes part I, Ch. 2, "Overview of principles of Hybridization and the strategy of Nucleic acid probe assays," (1993) Elsevier, NY. In general, highly stringent hybridization and washing conditions are selected to be about 5°C lower than the melting temperature (Tm) of the specific sequence at a defined ionic strength and pH. The Tm is the temperature (defined ionic strength and pH) at which 50% of the target sequence hybridizes to a perfectly matched probe. Very stringent conditions are selected to be equal to the Tm of the specific probe.An example of stringent hybridization conditions for hybridization of complementary nucleic acids having more than 100 complementary residues on an array or filter in a Southern or Northern blot is 42°C using a standard hybridization solution (see, for example, Sambrook and Russell Molecular Cloning: A Laboratory Manual (3rd ed.) Vol. 1-3 (2001) Cold Spring Harbor Laboratory, Cold Spring Harbor Press, NY). An example of high stringency washing conditions is NaCl 0.15M at 72°C for about 15 minutes. An example of stringent washing conditions is a 0.2xSSC wash at 65°C for about 15 minutes. Often, a low stringency wash precedes a high stringency wash to remove background probe signals. For example, an example of a moderate stringency wash for duplexes of more than 100 nucleotides is 1xSSC at 45°C for 15 minutes. For example, an example of a low stringency wash for a duplex of greater than 100 nucleotides is 4xSSC to 6xSSC at 40°C for 15 minutes.

[0042] The terms "nucleic acid" and "polynucleotide" are used interchangeably and refer to a polymeric form of nucleotides of any length, either deoxyribonucleotides or ribonucleotides, or analogs thereof. Polynucleotides can have any three-dimensional structure and can perform any function, known or unknown. Non-limiting examples of polynucleotides include genes, gene fragments, exons, introns, messenger RNA (mRNA), transfer RNA, ribosomal RNA, ribozymes, cDNA, shRNA, single-stranded short or long RNA, recombinant polynucleotides, branched polynucleotides, plasmids, vectors, isolated DNA of any sequence, regulatory regions, isolated RNA of any sequence, nucleic acid probes and primers. Nucleic acid molecules can be linear or circular.

[0043] The term "overall survival" means the period from either the time of diagnosis or the start of treatment that the patient is still alive.

[0044] As used herein, the term "prognose" or "prognosing" refers to a prediction or forecast of the course or outcome of a disease or condition.

[0045] The term "progression-free survival" means the time from treatment of a patient to progression of cancer or death of the patient, whichever occurs first.

[0046] Generally, a "protein" is a polypeptide (i.e., a string of at least two amino acids linked together by peptide bonds). A protein may contain moieties other than amino acids (e.g., may be a glycoprotein) and / or may be otherwise processed or modified. Those of skill in the art will understand that a "protein" may be an entire polypeptide chain (with or without a signal sequence) as produced by a cell, or a functional portion thereof. Those of skill in the art will further understand that a protein may sometimes include more than one polypeptide chain, for example, linked by one or more disulfide bonds or associated by other means.

[0047] The term "recommending" or "suggesting" in the context of disease treatment means making a suggestion or recommendation for a therapeutic action (e.g., drug therapy, adjunctive therapy, etc.) and / or disease management that is particularly applicable to a patient.

[0048] The terms "responsive", "clinical response", "positive clinical response", etc., as used in the case of patient response to cancer therapy, are used interchangeably and refer to the desired patient response to treatment as opposed to an undesired response, i.e., an adverse event. In patients, a beneficial response can be expressed using several clinical parameters, including disappearance of detectable tumor (complete response, CR), reduction in tumor size and / or cancer cell count (partial response, PR), tumor growth arrest (stable disease, SD), enhancement of antitumor immune response, potential to cause tumor regression or elimination; alleviation to some extent of one or more symptoms associated with the tumor; extension of the length of survival after treatment; and / or reduction in mortality at a given time point after treatment. A continued increase in tumor size and / or cancer cell count and / or tumor metastasis is indicative of a lack of beneficial response to treatment. In a population, the clinical benefit of a drug, i.e., its efficacy, can be evaluated based on one or more endpoints. For example, analysis of objective response rate (ORR) classifies patients who experience a CR or PR after drug treatment as responders. Analysis of disease control (DC) classifies patients who experience CR, PR or SD after drug treatment as responders. A positive clinical response can be evaluated using any endpoint and shows benefit to the patient including, but not limited to, (1) inhibition to some extent, including slowing and complete cessation of tumor growth, (2) reduction in tumor cell count, (3) reduction in tumor size, (4) inhibition (i.e., reduction, slowing or complete cessation) of tumor cell invasion into adjacent surrounding organs and / or tissues, (5) inhibition of metastasis, (6) enhancement of anti-tumor immune response that may cause tumor regression or elimination, (7) relief to some extent of one or more symptoms associated with the tumor, (8) increased length of survival after treatment, and / or (9) decreased mortality at a given time point after treatment. A positive clinical response can also be expressed using various measures of clinical outcome. Positive clinical outcomes can also be considered for individual outcomes versus outcomes for a patient population with a similar clinical diagnosis and can be assessed using various endpoints such as increased recurrence-free interval (RFI), increased survival time compared to overall survival (OS) for the population, increased disease-free survival (DFS), and increased distant recurrence-free interval (DRFI).Additional endpoints include chance of event-free (AE) survival, chance of metastatic recurrence-free (MR) survival (MRFS), chance of disease-free survival (DFS), chance of recurrence-free survival (RFS), chance of first progression (FP) and chance of distant metastasis-free survival (DMFS). An increased chance of a positive clinical response corresponds to a decreased chance of cancer recurrence or relapse.

[0049] As used herein, the term "subject" refers to a human or any non-human animal (e.g., mouse, rat, rabbit, dog, cat, cow, pig, sheep, horse, or primate). Human includes prenatal and postnatal forms. In many embodiments, the subject is a human being. The subject may be a patient, which refers to a person who visits a medical institution for diagnosis or treatment of a disease. The term "subject" is used interchangeably herein with "individual" or "patient." A subject may suffer from or be susceptible to a disease or disorder, but may or may not exhibit symptoms of the disease or disorder.

[0050] The term "tumor sample" includes biological samples or samples from biological sources that contain one or more tumor cells. Biological samples include samples of body fluids, such as blood, plasma, serum or urine, or samples obtained, for example, by biopsy, from cells, tissues or organs, preferably tumor tissues suspected of containing or essentially consisting of cancer cells.

[0051] The term "treatment", "treat" or "treating" refers to a method of reducing the effects or symptoms of cancer (e.g., breast cancer, lung cancer, ovarian cancer, etc.). Thus, in the disclosed methods, treatment can refer to a 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90% or 100% reduction in the severity of cancer or symptoms of cancer. For example, a method of treating a disease is considered to be a treatment if there is a 10% reduction in one or more symptoms of the disease in a subject compared to a control. Thus, the reduction may be a 10%, 20%, 30%, 40%, 50%, 60%, 70%, 80%, 90%, 100% or any percentage reduction between 10 and 100% compared to the natural or control level. It is to be understood that treatment does not necessarily refer to a cure or complete disappearance of a disease, illness, or symptoms of a disease or illness.

[0052] Biomarkers for predicting response to cetuximab

[0053] The methods and compositions described herein are based in part on the discovery of a panel of biomarkers that are correlated with cetuximab sensitivity in cancer patients.In one aspect, the present disclosure provides a method for predicting response to cetuximab in patients with gastric cancer.In certain embodiments, the biomarkers include the expression levels of certain genes and certain gene mutations. In certain embodiments, the biomarker is selected from EGFR expression level, TMEM40 expression level, IL1A expression level, PTPRN2 expression level, LCE2A expression level, TREM2 expression level, LY6D expression level, TMEM63B expression level, EIF4EBP1 expression level, KRAS mutation, APC mutation, MACF1 mutation, NCOR2 mutation, LPP mutation, C20orf56 expression level, SHC expression level, DSG3 expression level, HES6 expression level, FAM25B expression level, PNMA2 expression level, GSK3B expression level, PPM1H expression level, TOX3 expression level, TYMP expression level, Anxa8L2 expression level and ACP6 expression level. Certain biomarker information can be found in Table 1. mRNA and protein sequences are referred to by NCBI sequence number.

[0054] [Table 1-1] [Table 1-2] [Table 1-3] [Table 1-4]

[0055] Biomarker measurements The methods of the present disclosure include detecting or measuring at least some of the predictive biomarkers disclosed herein in a tumor sample obtained from a patient suspected of having or at risk of having cancer, in some embodiments, the patient has been diagnosed with cancer.

[0056] Sample preparation

[0057] The tumor sample may be a biological sample containing cancer cells. In some embodiments, the tumor sample is a fresh or archived sample obtained from a tumor, for example, by tumor biopsy or fine needle aspiration. The sample may also be any biological fluid that contains cancer cells. The tumor sample may be isolated or obtained from any number of primary tumors, including, but not limited to, breast, lung, prostate, brain, liver, kidney, intestine, colon, spleen, pancreas, thymus, testis, ovary, uterus, etc. In some embodiments, the tumor sample is from a tumor cell line. The collection of the tumor sample from the subject is performed according to standard protocols that hospitals or clinics generally follow, such as during a biopsy.

[0058] In certain embodiments, the method further includes isolating or extracting cancer cells (such as circulating tumor cells) from a biological fluid sample (such as a peripheral blood sample) or a tissue sample obtained from the subject. Cancer cells can be separated by immunomagnetic separation techniques, such as those available from Immunicon (Huntingdon Valley, Pa.).

[0059] In certain embodiments, the tissue sample may be processed to perform in situ hybridization, for example, the tissue sample may be embedded in paraffin and then fixed to a glass microscope slide, and then deparaffinized with a solvent, typically xylene.

[0060] In some embodiments, the method further comprises isolating nucleic acid, for example DNA or RNA, from the sample.Various extraction methods, such as phenol and chloroform extraction, and various other methods described, for example, in Ausubel et al., Current Protocols of Molecular Biology (1997) John Wiley & Sons, and Sambrook, and Russell, Molecular Cloning: A Laboratory Manual 3rd ed. (2001), are suitable for isolating DNA or RNA from cells or tissues.

[0061] Commercially available kits such as, for example, NucliSens extraction kit (Biomerieux, Marcy l'Etoile, France), QIAamp™ mini blood kit, Agencourt Genfind™, Rneasy® mini columns (Qiagen), PureLink® RNA mini kit (Thermo Fisher Scientific), and Eppendorf Phase Lock Gels™ can also be used to isolate DNA and / or RNA. A person skilled in the art can easily extract or isolate RNA or DNA following the manufacturer's procedures.

[0062] How to measure biomarkers

[0063] The biomarkers disclosed herein can be detected at the DNA (e.g., genomic DNA) or RNA (e.g., mRNA) level using suitable methods known in the art, such as, but not limited to, amplification assays, hybridization assays and sequencing assays. Gene expression levels can be detected at the RNA (e.g., mRNA) level or protein level using suitable methods known in the art. Gene mutations can be detected at the DNA level or RNA level using suitable methods known in the art.

[0064] Sequencing methods

[0065] Sequencing methods useful for measuring biomarkers include sequencing of target nucleic acids. Any sequencing method known in the art can be used to detect the biomarker of interest. Generally, sequencing methods can be categorized into traditional or classical methods and high-throughput sequencing methods (next-generation sequencing methods). Traditional sequencing methods include Maxam-Gilbert sequencing (also known as chemical sequencing) and Sanger sequencing (also known as chain termination).

[0066] By using methods that distinguish them from traditional methods such as Sanger sequencing, high-throughput sequencing, i.e., next-generation sequencing, is highly scalable and can sequence an entire genome or transcriptome at once. High-throughput sequencing methods include sequencing by synthesis, sequencing by ligation, and ultra-deep sequencing (such as those described by Marguiles et al., Nature 437(7057):376-80(2005)). Sequencing by synthesis involves synthesizing a complementary strand of a target nucleic acid by incorporating a labeled nucleotide or nucleotide analogue into a polymerase amplification. Immediately after or upon successful incorporation of the labeled nucleotide, the signal of the label is measured and the identity of the nucleotide is recorded. The detectable label on the incorporated nucleotide is removed prior to incorporation, and the detection and identification steps are repeated. Examples of sequencing by synthesis are known in the art and are described in, for example, U.S. Patent No. 7,056,676, U.S. Patent No. 8,802,368, and U.S. Patent No. 7,169,560, the contents of which are incorporated herein by reference.Sequencing by synthesis can be carried out on a solid surface (or microarray or chip) using foldback PCR and anchor primers.Target nucleic acid fragments can be bound to a solid surface by hybridizing to anchor primers and bridge amplified.This technology is used, for example, in the Illumina® sequencing platform.

[0067] Pyrosequencing involves hybridization of the target nucleic acid region to a primer and, in the presence of a polymerase, extension of the new strand by sequential incorporation of deoxynucleotide triphosphates corresponding to the bases A, C, G, and T (U). With each incorporation of a base, pyrophosphate is released, which is converted to ATP by sulfurylase, synthesizing oxyluciferin and emitting visible light. Since pyrophosphate release is equimolar to the number of bases incorporated, the light emitted is proportional to the number of nucleotides added in one step. This method is repeated until the entire sequence is determined.

[0068] In certain embodiments, the biomarkers described herein are detected by whole transcriptome shotgun sequencing (RNA sequencing). Methods for RNA sequencing have been described (see Wang Z, Gerstein M and Snyder M, Nature Review Genetics (2009) 10:57-63; Maher CA et al., Nature (2009) 458:97-101; Kukurba K & Montgomery SB, Cold Spring Harbor Protocols (2015) 2015(11):951-969).

[0069] Amplification assays

[0070] Nucleic acid amplification assays involve copying a target nucleic acid (e.g., DNA or RNA), thereby increasing the number of copies of the amplified nucleic acid sequence. Amplification can be exponential or linear. Examples of nucleic acid amplification methods include, but are not limited to, amplification using the polymerase chain reaction ("PCR", see U.S. Pat. Nos. 4,683,195 and 4,683,202; PCR Protocols: A Guide To Methods And Applications (Innis et al., eds, 1990)), reverse transcriptase polymerase chain reaction (RT-PCR), quantitative real-time PCR (qRT-PCR); quantitative PCR such as TaqMan®, nested PCR, ligase chain reaction (see Abravaya, K., et al., Nucleic Acids Research, 23:675-682, (1995)), branched DNA signal amplification (see Urdea, MS, et al., AIDS, 7(suppl 2):S11-S14, (1993)), amplifiable RNA reporters, Q-beta replication (see Lizardi et al., J. Am. Soc. 1999, 11:111-112, (1999)). al., Biotechnology (1988) 6:1197), transcription-based amplification (see Kwoh et al., Proc. Natl. Acad. Sci. USA (1989) 86:1173-1177), boomerang DNA amplification, strand displacement activation, cycling probe technology, self-sustained sequence replication (Guatelli et al., Proc. Natl. Acad. Sci. USA (1990) 87:1874-1878), rolling circle replication (U.S. Patent No. 5,854,033), isothermal nucleic acid sequence based amplification (NASBA), and serial analysis of gene expression (SAGE).

[0071] In certain embodiments, nucleic acid amplification assay is a PCR-based method. PCR is started with a primer pair that hybridizes to the target nucleic acid sequence to be amplified, and the primer is extended by polymerase that uses the target nucleic acid sequence as a template and dNTPs as building blocks to synthesize a new strand.The new strand and the target strand are then denatured to allow the primer to bind for the next extension and synthesis cycle.After multiple amplification cycles, the total number of copies of the target nucleic acid sequence can increase exponentially.

[0072] In some embodiments, an intercalating agent can be used that generates a signal when it is intercalated into double-stranded DNA.Examples of agents include SYBR GREEN™ and SYBR GOLD™.These agents are not template-specific, so it is considered that the signal is generated based on template-specific amplification.This can be confirmed by monitoring the signal as a function of temperature, since the melting temperature of template sequence is generally much higher than that of, for example, primer-dimer.

[0073] In certain embodiments, a detectably labeled primer or detectably labeled probe may be used to allow for detection of the biomarker corresponding to that primer or probe, hi certain embodiments, multiple labeled primers or probes with different detectable labels may be used to allow for simultaneous detection of multiple biomarkers.

[0074] Hybridization assay

[0075] Nucleic acid hybridization assays use probes to hybridize to target nucleic acids, thereby allowing detection of the target nucleic acids. Non-limiting examples of hybridization assays include Northern blotting, Southern blotting, in situ hybridization, microarray analysis, and multiplex hybridization-based assays.

[0076] In certain embodiments, the probe for hybridization is detectably labeled. In certain embodiments, the nucleic acid-based probe for hybridization is unlabeled. Such unlabeled probes can be immobilized on a solid support such as a microarray and can hybridize to a target nucleic acid molecule that is detectably labeled.

[0077] In certain embodiments, hybridization assays can be performed by isolating nucleic acid (e.g., RNA or DNA), separating the nucleic acid (e.g., by gel electrophoresis), and then transferring the separated nucleic acid to a suitable membrane filter (e.g., nitrocellulose filter), where the probe hybridizes to the target nucleic acid and allows detection.See, for example, Molecular Cloning: A Laboratory Manual, J. Sambrook et al., eds., 2nd edition, Cold Spring Harbor Laboratory Press, 1989, Chapter 7. The hybridization of the probe and the target nucleic acid can be detected or measured by methods known in the art. For example, autoradiographic detection of hybridization can be performed by exposing the hybridized filter to photographic film.

[0078] In some embodiments, the hybridization assay can be performed on a microarray. Microarrays provide a method for the simultaneous measurement of the levels of multiple target nucleic acid molecules. The target nucleic acid can be RNA, DNA, cDNA reverse transcribed from mRNA, or chromosomal DNA. The target nucleic acid can be hybridized to a microarray that includes a substrate with multiple immobilized nucleic acid probes, arrayed at a density of up to several million probes per square centimeter of substrate surface. The RNA or DNA in the sample hybridizes to the complementary probe on the array and is then detected by laser scanning. Hybridization intensity is determined for each probe on the array and converted to a quantitative value that represents the relative level of RNA or DNA. See U.S. Patent Nos. 6,040,138, 5,800,992 and 6,020,135, 6,033,860 and 6,344,316.

[0079] Techniques for synthesizing these arrays using mechanical synthesis methods are described, for example, in U.S. Patent No. 5,384,261. Although planar array surfaces are often used, arrays can be fabricated on practically any shape of surface, and can be fabricated on multiple surfaces. Arrays can be peptides or nucleic acids on beads, gels, polymer surfaces, fibers such as fiber optics, glass, or any other suitable substrate, see U.S. Patent Nos. 5,770,358, 5,789,162, 5,708,153, 6,040,193, and 5,800,992. Arrays can be packaged in a manner that allows for diagnostics or other operation of the comprehensive device. Useful microarrays are also commercially available, such as microarrays from Affymetrix, Nano String Technologies, and QuantiGene 2.0 multiplex assays from Panomics.

[0080] In certain embodiments, the hybridization assay is an in situ hybridization assay. In situ hybridization assays are useful for detecting the presence of genetic mutations. Probes useful for in situ hybridization assays can be mutation-specific probes that hybridize to specific genetic mutations to detect the presence or absence of a specific mutation of interest. Methods for using probes of unique sequences for in situ hybridization are described in U.S. Patent No. 5,447,841, incorporated herein by reference. Probes can be examined using a fluorescent microscope and appropriate filters for each fluorophore, or by using dual or triple bandpass filter sets to observe multiple fluorophores. See, for example, U.S. Patent No. 5,776,688 to Bittner et al., incorporated herein by reference. Hybridized probes can be visualized using any suitable microscopic imaging method, such as an automated digital imaging system. Alternatively, techniques such as flow cytometry can be used to examine the hybridization pattern of the probes.

[0081] Immunoassays

[0082] Immunoassays as used herein typically include the use of antibodies that specifically bind to biomarker proteins. Such antibodies can be obtained using methods known in the art (e.g., Huse et al., Science (1989) 246:1275-1281; Ward et al, Nature (1989) 341:544-546) or from commercial sources. Examples of immunoassays include, but are not limited to, Western blotting, enzyme-linked immunosorbent assay (ELISA), enzyme immunoassay (EIA), radioimmunoassay (RIA), immunoprecipitation, sandwich assay, competitive assay, immunofluorescence staining and imaging, immunohistochemistry (IHC), and fluorescence-activated cell sorting (FACS). For a general review of immunological and immunoassay techniques, see Basic and Clinical Immunology (Stites & Terr eds., 7th ed. 1991). Furthermore, immunoassays can be performed in any of a number of formats, which are reviewed in detail in Enzyme Immunoassay (Maggio, ed., 1980) and Harlow & Lane, supra. For general reviews of immunoassays, see also Methods in Cell Biology: Antibodies in Cell Biology, volume 37 (Asai, ed. 1993); Basic and Clinical Immunology (Stites & Terr, eds., 7th ed. 1991).

[0083] Any of the assays and methods provided herein for measuring gene expression levels can be adapted or optimized for use in automated and semi-automated systems, or point-of-care assay systems.

[0084] The gene expression levels described herein can be normalized using suitable methods known in the art.For example, gene expression levels can be normalized to the standard level of a standard marker, which can be pre-determined, simultaneously determined, or determined after obtaining a sample from a subject.The standard marker can be operated in the same assay, or can be a known standard marker from a previous assay.In another example, gene expression levels can be normalized to an internal control, which can be an internal marker, or an average level, or the total level of multiple internal markers.

[0085] The level of mRNA expression of each of the biomarkers described herein can be normalized to the reference level of a control gene. The control value can be determined in advance, at the same time, or after obtaining a sample from a subject. The standard can be operated on the same array, or can be a known standard from a previous assay. If the level of RNA expression is determined by RNA sequencing, the level of RNA expression of each of the biomarkers can be normalized to the total reads of sequencing. The normalized level of mRNA expression of the biomarker gene can be converted into a score, for example, using the methods and models described herein.

[0086] Methods for predicting response to cetuximab

[0087] After measuring the panel of biomarkers, the methods disclosed herein include determining the likelihood that the patient will be responsive to cetuximab. In certain embodiments, this probability is determined using techniques such as partial least squares (Wold S et al., PLS for Multivariate Linear Modeling, Chemometric Methods in Molecular Design (1995) Han van de Waterbeemd (ed.), pp. 195-218. VCH, Weinheim), elastic nets (Zou H et al., Regularization and Variable Selection via the Elastic Net, Journal of the Royal Statistical Society, Series B (2005) 67(2):301-320), support vector machines (Vapnik V, The Nature of Statistical Learning Theory (2010) Springer), random forests (Breiman L, Random Forests, Machine Learning (2001) 45:5-32), neural nets (Bishop C, Neural Networks for Pattern Recognition (1995) Oxford University Press, Oxford) and gradient boosting machines (Friedman J, Greedy Function The likelihood can be determined based on a model using machine learning techniques such as the approximation: a gradient boosting machine, Annals of Statistics (2001) 29(5), 1189-1232. In some cases, the likelihood is determined based on a model constructed by regularized regression.

[0088] As used herein, "machine learning" refers to a computer-implemented technology that gives a computer system the ability to gradually improve its performance on a specific processing task with data, i.e., to learn from data, without being explicitly programmed. Machine learning technology employs algorithms that can learn from data and make predictions about data by building models from sample inputs, i.e., describing the system using mathematical concepts. The main goal of machine learning is to generalize from experience, i.e., to perform accurately on new data after experiencing a learning data set. For biomedical diagnosis or prediction, machine learning techniques generally include supervised learning methods, where a computer is presented with example inputs (e.g., gene expression signatures) and desired outputs (e.g., responsiveness) to learn general principles that map the inputs to the outputs. Different models, i.e., hypotheses, can be used in generalization. For best generalization performance, the complexity of the hypotheses should match the complexity of the features underlying the data.

[0089] Machine learning models can be classified as supervised or unsupervised. Supervised learning involves learning a function that maps an input to an output based on examples of input-output pairs. Supervised models can be subdivided into either regression or classification models. In regression models, the output is continuous. Common types of regression models include linear regression, decision trees, random forests, and neural networks. On the other hand, in classification models, the output is discrete. Common types of classification models include logistic regression, support vector machines, naive Bayes, decision trees, random forests, and neural networks. For certain methods disclosed in this application, the machine learning model is a classification model because the output is whether a cancer subject is likely to respond to cetuximab or not.

[0090] Unsupervised learning, on the other hand, involves drawing inferences and finding patterns in input data without reference to labeled grades. The two main methods used in unsupervised learning include clustering and dimensionality reduction.

[0091] Clustering is an unsupervised technique that involves grouping or clustering of data points. Common clustering algorithms include k-means clustering, hierarchical clustering, mean-shift clustering, and density-based clustering.

[0092] Dimensionality reduction is a method for reducing the number of random variables to obtain a set of principle variables. Common dimensionality reduction algorithms include Principal Component Analysis (PCA), Regularized Regression, and Boruta.

[0093] Classification Models

[0094] As mentioned above, in a classification model, the output is discrete. In certain embodiments, the methods disclosed herein include a classification model, i.e., a machine learning classifier. In certain embodiments, the machine learning classifier is constructed by a logistic regression model.

[0095] Logistic regression

[0096] The simplest idea of ​​linear regression is to find the line that best fits the data. Extensions of linear regression include multiple linear regression (e.g., finding the best-fit surface) and polynomial regression (e.g., finding the best-fit curve). Logistic regression is similar to linear regression, but is used to model the probability of a finite number of outcomes, typically two.

[0097] Support Vector Machine

[0098] A Support Vector Machine (SVM), at its most basic level, is a supervised classification technique that finds a hyperplane or boundary between two classes of data that maximizes the margin between the two classes. There may be many planes that can separate the two classes, but only one plane can maximize the margin or distance between the classes.

[0099] Decision Tree

[0100] A decision tree is a decision support tool that uses a tree-like model of decisions and their possible outcomes, such as chance outcomes, resource costs and utilities. Typically, a decision tree is structured like a flow chart, with each internal node representing a "test" of attributes, each branch representing the result of the test, and each leaf node representing a class label (the decision made after computationally processing all the attributes). The path from the root to the leaves represents the classification rule. Decision trees are intuitive and easy to build, but suffer from poor accuracy.

[0101] Random Forest

[0102] Random forest is an ensemble learning technique that builds decision trees. It involves generating multiple decision trees using a bootstrap dataset of the original data and randomly selecting a subset of variables at each step of the decision tree. The model then selects the most frequent value of all predictions for each decision tree. By relying on a "majority wins" model, the risk of error in individual trees is reduced.

[0103] Artificial Neural Networks

[0104] Artificial neural networks (ANNs), usually simply called neural networks (NNs), are inspired by the biological neural networks that make up animal brains. ANNs are based on a collection of connected units or nodes called artificial neurons, loosely modeled on neurons in biological brains. Each connection can transmit a signal to other neurons, like synapses in biological brains. An artificial neuron can receive a signal, then process it and send a signal to the neurons connected to it. The "signals" at the connections are real numbers, and the output of each neuron is calculated by some nonlinear function of the sum of the inputs. The connections are called edges. Neurons and edges usually have weights that are adjusted as learning progresses. Weights increase or decrease the strength of the signal at the connection. Neurons may have thresholds so that a signal is sent only if the aggregate signal exceeds that threshold. Typically, neurons are aggregated into layers. Different layers may perform different transformations on their inputs. Signals travel from the first layer (input layer) to the last layer (output layer), possibly passing through the layers multiple times.

[0105] Dimensionality reduction

[0106] Dimensionality reduction is a method to reduce the number of random variables under consideration by obtaining a set of key variables. In the field of diagnostic data mining, dimension reduction methods reduce the number of biomarkers that can be used for diagnosis. Dimensionality reduction methods can be divided into feature selection and feature extraction. In certain embodiments, machine learning classifiers are constructed by dimension reduction using regularized linear regression methods.

[0107] Feature Selection

[0108] Feature selection approaches, also known as variable selection, attribute selection or variable subset selection, attempt to find a subset of input variables to use in building a model. The simplest algorithm for feature selection is to test each possible subset of features and find the one that minimizes the error rate. Feature selection algorithms can be classified into three categories: wrappers, filters and embedding methods.

[0109] Wrapper methods use a predictive model to score a feature subset. Each new subset is used to train a model and then tested on a holdout set. Counting the number of mistakes made on that holdout set (the model's error rate) gives the score for that subset. Because wrapper methods train a new model for each subset, they are very computationally intensive, but they usually provide the best-performing feature set for a particular type of model or typical problem.

[0110] Filter methods use an alternative criterion instead of error rate to score feature subsets. This metric is chosen to be computationally fast while still capturing the usefulness of the feature set. Common criteria include mutual information, pointwise mutual information, Pearson product moment correlation coefficient, Relief-based algorithms, and inter-class / intra-class distances, or significance test scores for each class / feature combination. Filters are typically less computationally intensive than wrappers, but produce feature sets that are not tuned to a specific type of predictive model. This lack of tuning means that the feature set from the filter is more general than the set from the wrapper, and will usually have lower predictive performance than the wrapper. However, the feature set is more useful for revealing relationships between features, since it does not include the assumptions of the predictive model. Many filters provide feature rankings rather than an explicit best feature subset, and the cutoff point for the ranking is chosen by cross-validation. Filter methods are also used as a preprocessing step for wrapper methods, allowing the wrapper to be used for larger problems. Another common approach is the recursive feature reduction algorithm, commonly used in support vector machines, which iteratively builds a model to remove low-weight features.

[0111] Embedding methods are a comprehensive group of techniques that perform feature selection as part of the model building method. A typical example of this approach is regularized regression methods, e.g., the least absolute shrinkage and selection operator (LASSO) method for building linear models, which imposes an L1 penalty on the regression coefficients to compress many of the penalties to zero. Any features with non-zero regression coefficients are "selected" by the LASSO algorithm. As a result, LASSO methods perform both feature selection and regularization to increase the predictive accuracy and interpretability of the resulting machine learning model. Improvements on LASSO include Bolasso, which bootstraps samples; elastic net regularization, which combines the L1 penalty of LASSO with the L2 penalty of ridge regression; and FeaLect, which scores all features based on a combinatorial analysis of the regression coefficients. AEFS further extends LASSO to nonlinear scenarios with autoencoders. These approaches tend to fall between filters and wrappers in terms of computational complexity.

[0112] Feature Extraction

[0113] Feature extraction is a method to construct a set of features from the initial features that are intended to be informative and non-redundant to facilitate subsequent learning and generalization steps. Examples of feature extraction algorithms include principle component analysis (PCA), isomap, partial least squares, nonlinear dimensionality reduction, multiple linear subspace learning, positive semidefinite embedding, and autoencoders.

[0114] Principal component analysis (PCA)

[0115] PCA involves projecting high-dimensional data into a smaller space, reducing the dimensionality of the data while retaining all the original variables in the model. PCA can be thought of as fitting a p-dimensional ellipsoid to the data, where each axis of the ellipsoid represents a principal component. If an axis of the ellipsoid is smaller, the variance along that axis will also be smaller.

[0116] To find the axes of the ellipsoid, the values ​​of each variable in the data set are first centered at zero by subtracting the mean of the observed values ​​of the variable from each value. These transformed values ​​are used in place of the original observed values ​​of each of the variables. Next, a covariance matrix is ​​calculated for the data, and the eigenvalues ​​and corresponding eigenvectors of this covariance matrix are calculated. Each of the orthogonal eigenvectors are then normalized and converted into a unit vector. Once this is done, each of the mutually orthogonal unit eigenvectors can be interpreted as an axis of an ellipsoid that has been fitted to the data. This choice of basis converts the covariance matrix into a diagonal form where the diagonal elements represent the variance of each axis. The percentage of variance represented by each eigenvector can be calculated by dividing the eigenvalue corresponding to that eigenvector by the sum of all eigenvalues. Simply put, PCA is a method to summarize (1) a measure of how each variable (e.g., biomarkers) is related to each other using the covariance matrix, (2) the directions in which the data is distributed using the eigenvectors, and (3) the relative importance of these different directions using the eigenvalues.

[0117] COMPUTER-IMPLEMENTED METHODS, SYSTEMS AND APPARATUS - Patent application

[0118] Any of the methods described herein may be implemented, in whole or in part, in a computer system including one or more processors that may be configured to perform the steps. Thus, the embodiments are directed to a computer system configured to perform the steps of any of the methods described herein, which may have different components performing each step or each group of steps. Although presented as numbered steps, the steps of the methods herein may be performed simultaneously or in different orders. Furthermore, some of these steps may be used with some of the other steps of other methods. Also, all or some of the steps may be optional. Any steps of any method may be implemented with modules, circuits or other means for performing these steps.

[0119] Any of the computer systems mentioned herein may utilize any suitable number of subsystems. In some embodiments, the computer system includes a single computer device, where the subsystems may be components of the computer device. In other embodiments, the computer system may include multiple computer devices, each of which is a subsystem, with internal components. The subsystems may be interconnected via a system bus. Additional subsystems include, for example, printers, keyboards, storage devices, monitors coupled to display adapters, and others. Peripheral devices and input / output (I / O) devices coupled to an I / O controller may be coupled to the computer system by any number of means known in the art, such as a serial port. For example, a serial port or an external interface (e.g., Ethernet, Wi-Fi, etc.) may be used to couple the computer system to a wide area network, such as the Internet, a mouse input device, or a scanner. The interconnection via a system bus allows the central processing unit to communicate with each subsystem and control the execution of instructions from the system memory or storage devices (e.g., fixed disks such as hard drives or optical disks), and allows for the exchange of information between the subsystems. The system memory and / or storage devices may incorporate computer-readable media. Any of the data referred to herein may be output from one component to another or may be output to a user.

[0120] A computer system may include multiple identical components or subsystems, for example, linked together by external interfaces or by internal interfaces. In some embodiments, a computer system, subsystem, or device may be in communication with a network. In such a case, one computer may be considered a client and another computer a server, each of which may be part of the same computer system. Each of the client and server may include multiple systems, subsystems, or components.

[0121] It should be understood that any of the embodiments of the present disclosure can be implemented using hardware (e.g., application specific integrated circuits or field programmable gate arrays) and / or using computer software with a modular or integrated general programmable processor in the form of control logic. As used herein, a processor includes a multi-core processor on the same integrated chip, or multiple processing units on a single circuit board, or multiple processing units networked together. Based on the disclosure and teachings provided herein, a person skilled in the art will recognize and understand other ways and / or methods for implementing the embodiments of the present disclosure using hardware and combinations of hardware and software.

[0122] Any of the software components or functions described in this application can be implemented as software code executed by a processor using any suitable computer language, such as, for example, Java, C++, or Perl, using, for example, conventional or object-oriented techniques. The software code can be stored as a series of instructions or commands on a computer-readable medium for storage and / or transmission, suitable media including random access memory (RAM), read-only memory (ROM), magnetic media such as hard drives or floppy disks, or optical media such as compact disks (CDs) or DVDs (digital versatile disks), flash memory, and the like. The computer-readable medium may be any combination of such storage or transmission devices.

[0123] Such programs may also be encoded and transmitted using carrier signals adapted for transmission over wired, optical and / or wireless networks conforming to various protocols, including the Internet. Thus, computer-readable media according to embodiments of the present invention may be produced using such program-encoded data signals. Computer-readable media encoded with program code may be packaged in a compatible device or provided separately from other devices (e.g., via Internet download). Any such computer-readable media may reside on or within a single computer product (e.g., a hard drive, CD or entire computer system) or may reside on or within different computer products within a system or network. The computer system may include a monitor, printer or other suitable display to provide a user with any of the results mentioned herein. EXAMPLES

[0124] The following examples are provided to better illustrate the claimed invention and are not intended to limit the scope of the invention. The specific compositions, materials and methods described below are all within the scope of the invention, in whole or in part. These specific compositions, materials and methods are not intended to limit the invention, but merely exemplify specific embodiments within the scope of the invention. Those skilled in the art can develop equivalent compositions, materials and methods without exerting inventive capacity and without departing from the scope of the invention. It should be understood that many modifications can be made in the procedures described herein while remaining within the scope of the invention. It is the intention of the inventors that such modifications are included within the scope of the invention.

[0125] [Example 1] This example illustrates the identification of biomarkers to predict response to cetuximab in patient-derived xenograft (PDX) models.

[0126] material and method

[0127] A cohort of 207 PDX models of five types of cancer, including colon cancer, esophageal cancer, gastric cancer, head and neck cancer, and lung cancer, was used in this study. These models had received cetuximab treatment for at least two weeks. Both a newly developed scale, TGI, and the median of the AUC (area under the curve) ratio were calculated to evaluate drug efficacy. Next, these models were divided into three categories based on the median of the AUC ratio: 53 models that responded to cetuximab treatment (AUCr < 0.3 and TGI > 0.7), 82 models that responded partially to cetuximab treatment (0.3 < AUCr < 0.7 and 0.3 < TGI < 0.7), and 72 models that did not respond to cetuximab treatment (AUC > 0.7 and TGI < 0.3). The response categories for each type of cancer are shown in Table 2.

[0128] Genome-wide gene expression levels in the grafts were measured using RNA-seq. Genes with low expression levels or small fluctuations in expression levels were removed. Next, the normalized expression levels were used as input for feature selection and modeling operations.

[0129] Next, the inventors identified differentially expressed genes (DEGs) between responders and non-responders for various cancers and defined genes that are normally upregulated, such as genes that are upregulated in at least three types of cancer. The inventors found that 26 out of 52 genes function in epidermal development. The inventors found that EGFR is the most relevant gene in all samples combined. When examining various types of cancer, the higher the EGFR expression, the higher the efficacy of ES, GA, and LU, showing a correlation, but there is no significant correlation with CR and HN (presumably because the expression range is limited). This result indicates that EGFR expression is a good biomarker but is not sufficient to determine cetuximab responsiveness in some types of cancer.

[0130] According to receiver operating characteristics (ROC), expression of the following 10 genes was most associated with cetuximab response: FMOD, REPIN1, PTPRN2, FOXA2, C20orf56, EGFR, LY6D, SFN, MICALL1, and IL1A. Other highly associated genes included TREM2, KLK5, SPRR2D, LCE2B, FAM25B, NLRP10, LCE2A, SFN, DSG3, DEFB103B, ANXA8, LCE3E, TMEM40, ANXA8L2, and S100A7A.

[0131] We also performed mutation data analysis using the WES data of 205 models according to the method of the CancerGenomeInterpreter database. We found that the following gene mutations were significantly correlated with cetuximab response: APC, LPP, KRAS, GNAS, TRRAP, HERC2, MACF1, CDKN2A, ABCB1, NCOR2.

[0132] Yang M et al. reported that the combination of APC and TP53 mutations can be used to predict cetuximab response (Yang M, Schell MJ, Loboda A, et al. Repurposing EGFR inhibitor utility in colorectal cancer in mutant APC and TP53 subpopulations [J]. Cancer Epidemiology and Prevention Biomarkers, 2019, 28 (7): 1141-1152). However, the present inventors found that among 25 TP53 and APC double mutation models, only 2 models were responsive and 14 models were non-responsive, indicating that the combination of APC and TP53 mutations is not a good predictor of cetuximab response.

[0133] We also found that more than 25% of non-responders had KRAS mutations (see Table 3).

[0134] Next, we tested gene interactions in the EGFR signaling pathway. There are 75 genes in the EGFR signaling pathway that pass the low expression filter. We used a linear mixed model to model the second order (two genes interaction) for cetuximab treatment in gastric, colon, and neck and head cancers. We calculated the AIC, the coefficient and p-value of the two genes interaction on treatment. The results showed that EGFR interacted with SRC, GSK3B, EIP4EBP1, AKT3 and SHC3 with high significance and good model performance. We constructed three linear models of genes including EGFR, SRC, GSK3B, EIF4EBP1, AKT3 and SHC3: (1) model EGFR: AUC~EGFR; (2) additional model: AUC~EGFR+other genes; (3) interaction model: AUC~EGFR*(other genes). The results showed that adding interaction terms between EGFR and other genes significantly improved the model fit (see Figures 1-3).

[0135] Next, we combined all cancer types for biomarker analysis, based on the hypothesis that more samples would provide greater power.

[0136] For the selected biomarker candidates, we first selected four sets of genes / pathways: (1) Highly correlated genes: ranking genes by correlation (SCC between gene expression and AUCr); (2) High ROC genes: ranking genes by model performance (ROC index based on categorical endpoints); (3) Highly correlated pathways: ranking pathways by correlation (SCC between GSVA score and AUCr); (4) High ROC pathways: ranking pathways by model performance (ROC index based on categorical endpoints). We selected the top 10 up- and down-regulated genes / pathways in the above four sets and combined them together (total of 64 biomarker candidates).

[0137] We used regularized regression (LASSO) for feature selection and constructed a LASSO model with 15 features.

[0138] Based on the features selected by LASSO, we constructed a logistic regression model. Then, we further simplified the model using a stepwise model selection method. Only six features were left (see Table 4). The accuracy of this simple model was 0.912 (see Table 5).

[0139] Next, we added the EGFR pathway interaction terms to the model, and the overall accuracy was slightly higher than the model without the EGFR pathway interaction terms (see Table 6).

[0140] We further added mutational biomarkers to the model, and the accuracy of the final model was 100% based on logistic regression.

[0141] We used partial response data, which is more difficult to predict, as test data (82 models) (not used for modeling). Response (R) was defined as predicted probability >0.5. Results showed that predicted response (R) had lower AURr and higher TGI than NR.

[0142] We also tested the model performance based on the resampling method. We resampled 100 times, randomly selecting 80% samples as training data set and 20% samples as test data set each time, and calculate the overall accuracy for each resampling. The results were compared with the EGFR+KRAS mutation model. The average accuracy of the EGFR+KRAS mutation model is about 0.8, while the average accuracy of the full model is about 0.9.

[0143] To better fit the models to each type of cancer, the inventors further used samples for each type of cancer to generate cancer type-specific cutoffs and generate separate models for each type of cancer (see Figures 4-8).

[0144] [Table 2]

[0145] [Table 3]

[0146] [Table 4]

[0147] [Table 5]

[0148] [Table 6]

[0149] While the present disclosure has been particularly shown and described with reference to specific embodiments, some of which are preferred embodiments, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the disclosure disclosed herein.

Claims

1. 1. A method for predicting cetuximab sensitivity in a patient having cancer, comprising: measuring a set of biomarkers in a tumor sample from said patient, comprising EGFR expression level, TMEM40 expression level, IL1A expression level, PTPRN2 expression level, LCE2A expression level, TREM2 expression level, LY6D expression level, TMEM63B expression level, EIF4EBP1 expression level, KRAS mutation, APC mutation, MACF1 mutation, NCOR2 mutation and LPP mutation; and Using a machine learning classifier to determine the likelihood that the patient will be responsive to cetuximab based on the set of measured biomarkers. A method comprising:

2. 2. The method of claim 1, wherein the cancer is selected from colon cancer, gastric cancer, lung cancer, head and neck cancer and esophageal cancer.

3. 2. The method of claim 1, wherein the set of biomarkers further comprises a C20orf56 expression level, a SHC expression level, a DSG3 expression level, a HES6 expression level, a FAM25B expression level, a PNMA2 expression level, a GSK3B expression level, a PPM1H expression level, a TOX3 expression level, a TYMP expression level, an Anxa8L2 expression level, or an ACP6 expression level.

4. 1. A method for predicting cetuximab sensitivity in a patient with colon cancer, comprising: measuring a set of biomarkers in the patient's tumor sample, the set including KRAS mutations, LY6D expression levels, PNMA2 expression levels, EGFR expression levels, GSK3B expression levels, C20orf56 expression levels, MACF1 mutations and NCOR2 mutations; Using a machine learning classifier to determine the likelihood that the patient will be responsive to cetuximab based on the set of measured biomarkers. A method comprising:

5. 1. A method for predicting cetuximab sensitivity in a patient with gastric cancer, comprising: measuring a set of biomarkers in the patient's tumor sample, the set including LPP mutations, EHBP1L1 expression levels, EGFR expression levels, LY6D expression levels, C20orf56 expression levels, PTPRN2 expression levels, FMOD expression levels and NCOR2 mutations; and Using a machine learning classifier to determine the likelihood that the patient will be responsive to cetuximab based on the set of measured biomarkers. A method comprising:

6. 1. A method for predicting cetuximab sensitivity in a patient with lung cancer, comprising: measuring a set of biomarkers in the patient's tumor sample, the set including LPP mutations, FMOD expression levels, EGFR expression levels, GSK3B expression levels, FAM25B expression levels, SHC3 expression levels, IL1A expression levels, S100A7A expression levels, PTPRN2 expression levels, and AKT3 expression levels; Using a machine learning classifier to determine the likelihood that the patient will be responsive to cetuximab based on the set of measured biomarkers. A method comprising:

7. 1. A method for predicting cetuximab sensitivity in a patient with head and neck cancer, comprising: measuring a set of biomarkers in the patient's tumor sample, the set including SHC3 expression level, LPP mutation, HES6 expression level, S100A7A expression level, GSK3B expression level and FAM25B expression level; Using a machine learning classifier to determine the likelihood that the patient will be responsive to cetuximab based on the set of measured biomarkers. A method comprising:

8. 1. A method for predicting cetuximab sensitivity in a patient with esophageal cancer, comprising: measuring a set of biomarkers in the patient's tumor sample, the set including LPP mutations, EGFR expression levels, FMOD expression levels, LY6D expression levels, FAM25B expression levels, PNMA2 expression levels, TOX3 expression levels, and PTPRN2 expression levels; Using a machine learning classifier to determine the likelihood that the patient will be responsive to cetuximab based on the set of measured biomarkers. A method comprising:

9. The method of any one of claims 1 to 8, wherein the biomarkers are measured by an amplification assay, a hybridization assay, a sequencing assay or an array.

10. The method of any one of claims 1 to 8, wherein the machine learning classifier is constructed by a regularized regression method and a logistic regression model.

11. The method of any one of claims 1 to 8, further comprising administering to the patient cetuximab.

12. A non-transitory computer readable medium having instructions stored thereon, the instructions, when executed by a processor, causing the processor to: a search for data of a set of biomarkers obtained from tumor samples of patients with cancer, said set of biomarkers comprising EGFR expression level, TMEM40 expression level, IL1A expression level, PTPRN2 expression level, LCE2A expression level, TREM2 expression level, LY6D expression level, TMEM63B expression level, EIF4EBP1 expression level, KRAS mutation, APC mutation, MACF1 mutation, NCOR2 mutation, and LPP mutation; and Determining the likelihood that the patient will be responsive to cetuximab based on the data for a set of biomarkers using a machine learning classifier. A computer-readable medium for causing

13. 13. The non-transitory computer readable medium of claim 12, wherein the cancer is selected from colon cancer, gastric cancer, lung cancer, head and neck cancer, and esophageal cancer.

14. The non-transient computer readable medium of claim 12, wherein the set of biomarkers further comprises a C20orf56 expression level, a SHC expression level, a DSG3 expression level, a HES6 expression level, a FAM25B expression level, a PNMA2 expression level, a GSK3B expression level, a PPM1H expression level, a TOX3 expression level, a TYMP expression level, an Anxa8L2 expression level, or an ACP6 expression level.

15. A non-transitory computer readable medium having instructions stored thereon, the instructions, when executed by a processor, causing the processor to: A search for a set of biomarkers obtained from tumor samples of patients with colon cancer, the set of biomarkers comprising KRAS mutations, LY6D expression levels, PNMA2 expression levels, EGFR expression levels, GSK3B expression levels, C20orf56 expression levels, MACF1 mutations, and NCOR2 mutations; and Determining the likelihood that the patient will be responsive to cetuximab based on the data for a set of biomarkers using a machine learning classifier. A computer-readable medium for causing

16. A non-transitory computer readable medium having instructions stored thereon, the instructions, when executed by a processor, causing the processor to: A search for data of a set of biomarkers obtained from tumor samples of patients with gastric cancer, the set of biomarkers comprising LPP mutations, EHBP1L1 expression levels, EGFR expression levels, LY6D expression levels, C20orf56 expression levels, PTPRN2 expression levels, FMOD expression levels and NCOR2 mutations; and Determining the likelihood that the patient will be responsive to cetuximab based on the data for a set of biomarkers using a machine learning classifier. A computer-readable medium for causing

17. A non-transitory computer readable medium having instructions stored thereon, the instructions, when executed by a processor, causing the processor to: A search for data of a set of biomarkers obtained from tumor samples of patients with lung cancer, the set of biomarkers comprising LPP mutations, FMOD expression levels, EGFR expression levels, GSK3B expression levels, FAM25B expression levels, SHC3 expression levels, IL1A expression levels, S100A7A expression levels, PTPRN2 expression levels, and AKT3 expression levels; and Determining the likelihood that the patient will be responsive to cetuximab based on the data for a set of biomarkers using a machine learning classifier. A computer-readable medium for causing

18. A non-transitory computer readable medium having instructions stored thereon, the instructions, when executed by a processor, causing the processor to: A search of data for a set of biomarkers obtained from tumor samples of patients with head and neck cancer, the set of biomarkers comprising SHC3 expression levels, LPP mutations, HES6 expression levels, S100A7A expression levels, GSK3B expression levels, and FAM25B expression levels; and Determining the likelihood that said patient will be responsive to cetuximab based on data from a set of biomarkers using a machine learning classifier. A computer-readable medium for causing

19. A non-transitory computer readable medium having instructions stored thereon, the instructions, when executed by a processor, causing the processor to: A search for a set of biomarkers obtained from tumor samples of patients with esophageal cancer, the set of biomarkers including LPP mutations, EGFR expression levels, FMOD expression levels, LY6D expression levels, FAM25B expression levels, PNMA2 expression levels, TOX3 expression levels, and PTPRN2 expression levels; and Determining the likelihood that said patient will be responsive to cetuximab based on data from a set of biomarkers using a machine learning classifier. A computer-readable medium for causing

20. The non-transitory computer readable medium of any one of claims 12 to 19, wherein the biomarkers are measured by an amplification assay, a hybridization assay, a sequencing assay, or an array.

21. 20. The non-transitory computer-readable medium of any one of claims 12 to 19, wherein the machine learning classifier is constructed by a regularized regression method and a logistic regression model.

22. 1. A method for generating a machine learning model for predicting sensitivity to a drug in a patient having cancer, comprising: obtaining whole genome expression levels from each of a group of tumor models tested for responsiveness to said agent; selecting a first group of genes that have increased expression levels in a tumor model that is responsive to the agent compared to a tumor model that is not responsive to the agent; selecting a second group of genes that have decreased expression levels in the tumor model that is responsive to the agent compared to the tumor model that is not responsive to the agent; selecting a set of biomarkers from said first and second groups of genes using a regularized regression method; and Steps to build a machine learning classifier using a logistic regression model A method comprising:

23. 23. The method of claim 22, wherein the agent is cetuximab.

24. 23. The method of claim 22, wherein the tumor model is a xenograft model.

25. 23. The method of claim 22, wherein the first and second groups of genes are selected by correlation of gene expression levels with AUCr or by model performance of a ROC index.

Images