Method and system for screening neoantigens and use thereof

Abstract

The present invention relates to a method, system and uses thereof for screening neoantigens, and more specifically, to a method, system and uses thereof for screening neoantigens derived from genes whose expression is essential for the survival of cancer cells and / or is uniform in all cells within cancer tissue as diagnostic and / or therapeutic targets.

Description

[Technical field]

[0001] The present invention relates to a method, system and use thereof for screening neoantigens, and more specifically, to a method, system and use thereof for screening neoantigens whose expression is essential for the survival of cancer cells and / or which are derived from homogeneous genes in all cells within cancer tissue, as diagnostic and / or therapeutic targets. [Background technology]

[0002] Immune anti-cancer therapy is currently the most popular anti-cancer therapy, and in patients who respond to it, a high cure rate can be expected, presenting a new paradigm in anti-cancer treatment. It is being actively researched, with over 3,400 global immune anti-cancer drug clinical trials being conducted worldwide.

[0003] Anti-cancer vaccines are a type of immune anti-cancer therapeutic agent that activates the immune system using cancer-specific antigens, and are attracting attention as a core technology of immune anti-cancer therapy through combination therapy with immune checkpoint inhibitors. However, the actual success rate of immune checkpoint inhibitors, such as anti-PD-1 / anti-PD-L1, is only 15-20%, and they involve a process of losing the immune regulatory function of T cells in general, and there is a high concern that side effects such as autoimmune diseases may be observed. There is still a demand for therapeutic agents that can improve the response rate of cancer patients and minimize side effects.

[0004] Patient-specific anti-cancer vaccines are being developed as a treatment method that can improve the therapeutic effect of cancer patients and minimize side effects by enabling optimal vaccine design targeting patient-specific neoantigens. It is possible to improve the effect of immune anti-cancer treatment by inducing the immune action of the patient to concentrate on cancer cell-specific neoantigens, and the key to this is the technology to select the optimal neoantigen.

[0005] Cancer cells evolve into a form that can evade anti-cancer immune responses through a process known as immunoediting. Such evolved cancer cell populations are believed to be the cause of resistance to immune anti-cancer treatment and cancer recurrence. There is a need to develop neoantigen targets that can overcome immune evasion mechanisms due to the heterogeneity and plasticity of such tumors. Korean Patent Application Publication No. 2018-0107102 discloses a method for identifying and selecting neoantigens for personalized cancer vaccines, but does not disclose or suggest at all the immune evasion of cancer cells, let alone the need to overcome it.

[0006] The inventors have conducted research into strategies that can circumvent the immune editing mechanisms of cancer cells, which makes it difficult to discover neoantigens that are effective as anti-cancer vaccines, and have completed an invention relating to a method for screening neoantigens that are effective as diagnostic and / or therapeutic targets. Summary of the Invention [Problem to be solved by the invention]

[0007] An object of the present invention is to provide a method for screening neoantigens that overcome immune evasion of cancer cells.

[0008] Another object of the present invention is to provide a system for screening neoantigens that overcome immune evasion of cancer cells.

[0009] Another object of the present invention is to provide an anti-cancer vaccine comprising a neoantigen that overcomes immune evasion of cancer cells.

[0010] Another object of the present invention is to provide a composition for predicting the prognosis of treatment for cancer patients, which comprises a neoantigen that overcomes immune evasion by cancer cells. [Means for solving the problem]

[0011] One aspect of the present invention is obtaining exome, transcriptome, single cell transcriptome, peptidome, or whole genome sequence analysis data from a cancer patient; selecting a cancer cell survival-dependent gene; obtaining a neoantigen derived from said cancer cell survival-dependent gene.

[0012] In one embodiment of the present invention, the method for screening a neoantigen further comprises determining the binding affinity between the neoantigen and HLA (human leukocyte antigen) of an antigen-presenting cell.

[0013] In one embodiment of the present invention, a cancer cell survival-dependent gene is one which, when its expression is reduced or ablated, results in the death of various types of cancer cells or cancer cells derived from the patient.

[0014] In one embodiment of the present invention, the cancer cell survival dependency gene is also a universal dependency gene that is universally essential for the survival of various types of cancers or cancer cells.

[0015] In one embodiment of the present invention, the cancer cell survival-dependence gene is also a cancer patient-specific dependence gene essential for the survival of cancer cells derived from said cancer patient.

[0016] In one embodiment of the present invention, the cancer cell survival dependency genes include universal dependency genes or cancer patient-specific dependency genes, or universal dependency genes and cancer patient-specific dependency genes.

[0017] In one embodiment of the present invention, the step of selecting cancer cell survival-dependent genes includes using a cell survival-dependent prediction model to select cancer cell survival-dependent genes that are essential for cancer cell survival, the cell survival-dependent prediction model being generated by learning the relationship between cellular gene expression patterns and cell death, and genes that cause cancer cell death by reducing or eliminating the expression level of the gene can be selected as cancer cell survival-dependent genes.

[0018] In one embodiment of the present invention, the cell survival dependency prediction model is based on deep learning and is learned from experimental data relating to the relationship between cellular gene expression and cell death.

[0019] In one embodiment of the present invention, the relationship between the expression of the gene and cell death is based on in vitro screening experimental data or in silico data on the death of cancer cell lines by targeted reduction or ablation of gene expression.

[0020] In one embodiment of the present invention, the step of selecting the cancer cell survival-dependent gene further includes a step of determining whether the cancer cell survival-dependent gene selected from the cell survival-dependent prediction model is uniformly expressed in all cancer cells obtained from a cancer patient.

[0021] In one specific example of the present invention, the step of obtaining neoantigens derived from the cancer cell survival-dependent genes also includes a step of comparing sequences from cancer cells and sequences from normal cells from sequence analysis data obtained from the cancer patient to obtain neoantigens from the cancer patient, and a step of collecting neoantigens derived from the cancer cell survival-dependent genes from the obtained neoantigens.

[0022] In one specific example of the present invention, the step of obtaining neoantigens derived from the cancer cell survival-dependent genes also includes a step of comparing sequences from sequence analysis data obtained from a cancer patient with sequences from sequence analysis data obtained from a normal control group to obtain neoantigens from the cancer patient, and a step of collecting neoantigens derived from the cancer cell survival-dependent genes from among the obtained neoantigens.

[0023] In one embodiment of the present invention, the step of collecting the neoantigens further comprises screening for nonsynonymous mutations in the cancer cell survival-dependent genes.

[0024] In one embodiment of the present invention, the step of determining the binding affinity between the neoantigen and the HLA of an antigen-presenting cell includes a step of inputting the sequence of the neoantigen into a neoantigen binding affinity prediction model that predicts the binding affinity between an antigen peptide and the HLA of an antigen-presenting cell to obtain a binding affinity prediction, and the neoantigen binding affinity prediction model is generated by learning with interaction data between amino acids of the peptide and amino acids of the HLA.

[0025] In one embodiment of the invention, the antigen presenting cells are dendritic cells, macrophages, B cells, or a combination thereof.

[0026] In one embodiment of the invention, the HLA is MHC (major histocompatibility class) class I or MHC class II.

[0027] In one embodiment of the present invention, the neoantigen is also considered to have binding affinity when the CNN-MHC number between the neoantigen and the HLA of the antigen-presenting cell is >0.5.

[0028] In one embodiment of the present invention, the cancer cell survival-dependent genes are selected from a plurality of genes based on their essentiality for cancer cell survival.

[0029] In one embodiment of the present invention, the cancer cell survival-dependent genes are selected from the top and bottom 5, 10, or more genes according to their essentiality for cancer cell survival.

[0030] In one embodiment of the present invention, the neoantigen is also a cancer patient-specific neoantigen.

[0031] Another aspect of the present invention is at least one processor for executing the at least one instruction stored in a memory; The processor executes the at least one instruction to: A cell survival dependency prediction model is generated by learning the relationship between the amount of gene expression in a cell and cell death, and predicting the dependency of cell survival on gene expression; A gene expression profile of a cancer patient is input into the cell survival dependence prediction model to select a cancer cell survival dependence gene, and a neoantigen derived from the cancer cell survival dependence gene is obtained from the gene expression profile of the cancer patient; A neoantigen binding affinity prediction model is generated that predicts binding affinity based on amino acid interactions between the peptide and the antigen-presenting cell; The present invention provides a neoantigen screening system that utilizes the neoantigen binding affinity prediction model to select neoantigens that have binding affinity with the HLA of antigen-presenting cells.

[0032] In one embodiment of the invention, said system is also used to carry out a method for screening neoantigens according to one embodiment of the invention.

[0033] In one embodiment of the present invention, the cell survival dependence prediction model is capable of selecting cell survival dependence genes or predicting the cell survival dependence of genes from which neoantigens are derived.

[0034] In one specific example of the present invention, the cell survival dependence prediction model is generated by learning the relationship between cellular gene expression levels and cell death, and also selects genes that cause cancer cell death by reducing or eliminating their expression levels as cancer cell survival dependence genes.

[0035] In one embodiment of the present invention, the cancer cell survival dependency gene is also a universal dependency gene that is universally essential for the survival of various types of cancers or cancer cells.

[0036] In one embodiment of the present invention, a cancer cell survival-dependence gene is also a cancer patient-specific dependence gene that is essential for the survival of cancer cells derived from a specific cancer patient.

[0037] In one embodiment of the present invention, the cancer cell survival dependency genes include universal dependency genes or cancer patient-specific dependency genes, or universal dependency genes and cancer patient-specific dependency genes.

[0038] In one embodiment of the present invention, the processor, by executing the at least one instruction, also selects a neoantigen as having binding affinity if the CNN-MHC number between the neoantigen and the HLA of the antigen-presenting cell is >0.5.

[0039] In one embodiment of the present invention, the processor, by executing the at least one instruction, also learns the relationship between gene expression and cell death, and the relationship between the binding affinity of neoantigens to HLA of antigen-presenting cells, respectively.

[0040] In one embodiment of the present invention, the learning is also performed based on deep learning.

[0041] In one embodiment of the invention, the relationship between cellular gene expression and cell death is also based on in vitro or in silico data relating to the death of cancer cell lines by targeted reduction or ablation of gene expression.

[0042] In one embodiment of the present invention, the neoantigen binding affinity prediction model is also generated by learning with interaction data between amino acids of peptides and amino acids of HLA.

[0043] In one embodiment of the present invention, the gene expression profile of the cancer patient is also sequence analysis data of the exome, transcriptome, single cell transcriptome, peptidome or whole genome.

[0044] In one embodiment of the invention, said HLA is also MHC class I or MHC class II.

[0045] In one embodiment of the present invention, the neoantigen is also considered to have binding affinity when the CNN-MHC number between the neoantigen and the HLA of the antigen-presenting cell is >0.5.

[0046] In one embodiment of the present invention, the cancer cell survival-dependent genes are selected from a plurality of genes based on their essentiality for cancer cell survival.

[0047] Yet another aspect of the present invention is Obtaining a neoantigen by the screening method according to one aspect of the present invention described above; and producing an anti-cancer vaccine comprising the neoantigen.

[0048] In one embodiment of the present invention, the method for producing the anti-cancer vaccine includes the steps of obtaining a peptide sequence consisting of 9 to 30 amino acids including the neoantigen site; and selecting a peptide sequence having hydrophilicity and stability from the peptide sequences.

[0049] In one embodiment of the invention, the selected peptide sequences have a Kyte-Doolittle GRAVY<0 and an InstaIndex<40.

[0050] In one embodiment of the invention, the peptide sequence is composed of 12 to 30, 15 to 30, or even 15 to 25 amino acids.

[0051] Yet another aspect of the present invention provides an anti-cancer vaccine comprising a neoantigen obtained by the screening method according to the above-mentioned aspect of the present invention.

[0052] The anti-cancer vaccine may induce a specific cytotoxic T cell response and / or a specific helper T cell response.

[0053] In one embodiment of the invention, the anti-cancer vaccine also comprises a peptide comprising a neoantigen.

[0054] In one embodiment of the present invention, the peptide has a length of 15 to 30 amino acids and is capable of binding to the HLA of antigen-presenting cells and activating T cells specific to neoantigens.

[0055] In one embodiment of the invention, the anti-cancer vaccine also comprises peptides comprising two or more neoantigens.

[0056] In one embodiment of the present invention, the anti-cancer vaccine further comprises additional active ingredients, such as anti-cancer drugs, additives, excipients, and the like.

[0057] The anti-cancer vaccine is administered in an amount sufficient to induce an antigen-specific immune response. The amount of peptide contained in the anti-cancer vaccine and the dosage of the anti-cancer vaccine can be determined by those skilled in the art without undue experimentation. The anti-cancer vaccine can also be administered by intravenous injection, subcutaneous injection, intradermal injection, intraperitoneal injection, or intramuscular injection. In the anti-cancer vaccine, the concentration of the peptide can vary widely, such as less than about 0.1% by weight, about 2% by weight to about 20% by weight, about 50% by weight or more, and is determined in consideration of the administration method, etc. The dosage of the anti-cancer vaccine is determined by a clinician in consideration of the composition of the peptide containing the neoantigen, the administration method, the stage and severity of the target disease, the weight and health condition of the patient, etc., and generally, for a 70 kg patient, the amount of peptide can be administered in the range of about 1.0 μg to about 50,000 μg.

[0058] A further aspect of the present invention is a method for screening a neoantigen comprising the steps of: obtaining a neoantigen by a screening method according to one aspect of the present invention; measuring the amount of said neoantigen from a sample from a cancer patient.

[0059] In one embodiment of the present invention, the method for providing information predicting the treatment prognosis of a cancer patient further comprises a step of comparing the amount of the obtained neoantigen with the amount of the neoantigen obtained from a control group of cancer patients whose treatment prognosis has been confirmed.

[0060] In one embodiment of the present invention, the control group consists of cancer patients confirmed to have a good prognosis after treatment, or alternatively, consists of cancer patients confirmed to have a poor prognosis after treatment.

[0061] In one embodiment of the invention, the amount of neoantigen is also the number of neoantigens.

[0062] In one embodiment of the present invention, if the amount or number of neoantigens in a sample from a cancer patient is greater than that of a control group of cancer patients with poor prognosis, the prognosis of the cancer patient is also predicted to be good.

[0063] In one embodiment of the present invention, if the amount or number of neoantigens in a sample from a cancer patient is smaller than that of a control group of cancer patients with a good prognosis, it is also predicted that the cancer patient will have a poor prognosis.

[0064] Yet another aspect of the present invention provides a composition for predicting the prognosis of a cancer patient, comprising a neoantigen obtained by the screening method according to the above-mentioned aspect of the present invention.

[0065] The composition for predicting the treatment prognosis of a cancer patient according to one embodiment of the present invention further comprises additional components necessary for predicting the treatment prognosis of a cancer patient. Effect of the Invention

[0066] A method according to one embodiment of the present invention enables screening of neoantigens derived from universal addiction genes or cancer patient-specific addiction genes that are essential for the survival of cancer cells as diagnostic and / or therapeutic targets, and can be usefully used in developing neoantigen-based anti-cancer vaccines that can maximize the immunotherapeutic effects against cancer cells without worrying about immune evasion by cancer cells, and in predicting the prognosis of cancer patient treatment. [Brief description of the drawings]

[0067] [Figure 1] FIG. 1 is a schematic diagram of a method for in silico cancer cell survival-dependent gene prediction according to one embodiment of the present invention. [Diagram 2] 1 is a flow chart of a method for screening a neoantigen according to an embodiment of the present invention. The order and number of times of determining the dependency of the gene from which the neoantigen is derived on cancer cell survival and determining the binding strength between the neoantigen and an antigen-presenting cell may be adjusted as necessary. [Diagram 3] FIG. 1 is a block diagram of a system for neoantigen screening according to one embodiment of the present invention. [Figure 4] FIG. 2 is a block diagram of a processor according to one embodiment of the present invention. [Diagram 5] The significance of neoantigens obtained by a method for screening neoantigens using in vitro-dependent data according to one embodiment of the present invention as diagnostic and / or therapeutic targets is shown. The usefulness of a method for screening neoantigens based on cancer cell survival dependence according to one embodiment of the present invention was verified using in vitro dependence data and expression homogeneity data of published cohorts of lung cancer and melanoma. The gray dotted line indicates the difference in the number of neoantigens calculated for all genes by a standard method currently used. [Figure 6] This is a schematic diagram of the immune evasion mechanism of cancer cells. When the protein from which the neoantigen is derived is essential for the survival of the cancer cells and is a constitutive neoantigen that is uniformly expressed in all cells, the cancer cells respond effectively to immunotherapy. However, when the neoantigen is not essential for the survival of the cancer cells or is a facultative neoantigen that is not uniformly expressed in all cells (lower end), an immune evasion mechanism against immunotherapy occurs. [Figure 7]1 shows the response profile of neoantigens to immunotherapy obtained by a method for screening neoantigens based on in vitro-dependence data and single-cell gene expression data according to one embodiment of the present invention. The results of comparing clonal changes and gene expression changes before and after treatment in a melanoma immune checkpoint inhibitor (checkpoint blockade) treatment cohort (Riaz) are shown, divided into the essentiality (high dependency vs. low dependency) of the gene function from which the neoantigen is derived for cancer survival, and the homogenousness (homogenous vs. heterogenous expression). CR / PR and SD / PD mean positive prognosis and negative prognosis, respectively. In patients with a positive prognosis, it can be seen that neoantigens derived from genes with high essentiality were mainly attacked by the immune system, resulting in clonal contraction, and as a result of immune editing, gene expression (RNA expression) was reduced, while in patients with a negative prognosis, neoantigens derived from genes with low essentiality were mainly attacked by the immune system, resulting in clonal expansion and reduced expression. [Figure 8]The significance of neoantigens obtained by the method for screening neoantigens based on in vitro-dependent data according to one embodiment of the present invention as diagnostic / therapeutic targets is shown. Similar analyses to those presented in FIG. 5 were performed on the MSKCC panel (468 genes) results for all types of cancer patient samples, and survival analyses for immunotherapy cohorts of lung cancer and melanoma. The top of FIG. 8 (MSK pan-cancer) shows the results of analyzing mutation-derived genes by the entire MSKCC panel (all), or by the top 50% genes (high fitness) or bottom 50% genes (low fitness) essential for cancer cell survival, while the middle (lung cancer) and bottom (melanoma) show the results of analyzing published immunotherapy cohorts of lung cancer and melanoma by all genes, or by the top 500 genes (high fitness) and bottom 500 genes (low fitness) essential for cancer cell survival. Patients were divided into those with high mutation burden or neoantigen load (high) and those with low mutation burden (low) for each gene group, and their survival rates after treatment were compared, with the hazard ratio (HR) and p-value shown. The lower the HR and the lower the p-value, the higher the explanatory power of the gene group in question regarding treatment prognosis. As a result, it can be seen that using a small number of gene groups that are highly dependent on cancer cell survival provides a better explanation of treatment prognosis than using the entire gene group using the standard method. [Figure 9]The significance of neoantigens as diagnostic / therapeutic targets obtained based on cancer patient-specific in silico dependency data according to one embodiment of the present invention is shown. Based on the transcriptome data (https: / / portal.gdc.cancer.gov / ) for lung cancer and breast cancer patients published in TCGA (The Cancer Genome Atlas), the in silico dependency predictions shown in FIG. 1 were performed. Since these patients were not actually treated with immune anticancer therapy, they were divided into samples with high immune cell penetrance (high leukocyte fraction) and samples with low immune cell penetrance (low leukocyte fraction), and it was assumed that a similar effect to immune therapy would be shown in samples with high immune cell penetrance. As shown in FIG. 5, the explanatory power of genes from which neoantigens are derived for patient survival was measured, and similar results were obtained only in samples with high immune cell penetrance (R: responder; NR: nonresponder). The gray dotted line indicates the difference in the number of neoantigens calculated for all genes by the standard method used in the past. We were able to confirm that cancer patient-specific dependency data generally showed higher explanatory power for differences compared to universal dependency data. [Figure 10] 1 is a schematic diagram of a model for predicting the binding affinity between a neoantigen and MHC used in a screening method for neoantigens according to one embodiment of the present invention, which is constructed based on amino acid similarity matrix information of a two-dimensional matrix of HLA and neoantigen peptides, and applies a convolutional neural network (CNN) to predict the binding affinity between HLA and antigen peptides. [Figure 11] Using the test dataset officially provided by the IEDB (Immune Epitope Database), we present a comparison of the performance (AUC, F1 score) of CNN according to one embodiment of the present invention with other methods (NetMHCpan, NetMHCcon, ANN, SMMPMBEC). [Figure 12]The results of confirming the HLA binding ability of neoantigens predicted by one embodiment of the present invention are shown below. It was confirmed that 80% of peptides containing neoantigens predicted to bind to HLA-A02 actually bind to HLA-A02. [Figure 13] The results of confirming immune responses induced by neoantigens selected according to one embodiment of the present invention are shown below. Regarding confirmation of CD8+ T cell responses to neoantigens, ELISpot analysis based on INFγ secretion confirmed that 10 (66.7%) of 15 candidate neoantigen peptides predicted to be reactive induced INFγ secretion by T cells. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0068] The present invention will become clearer with reference to the following detailed description of the embodiments in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments, and may be embodied in various forms. The embodiments are provided solely for the purpose of making the disclosure of the present invention complete and fully conveying the scope of the invention to those skilled in the art to which the present invention pertains. The present invention is defined by the claims.

[0069] As used herein, the term "exome" refers to the collection of exons present in a cell, a group of cells, or an individual.

[0070] As used herein, the term "transcriptome" refers to the collection of expressed RNA present in a cell, a group of cells, or an individual.

[0071] As used herein, the term "gene expression profile" refers to an analysis of the genes expressed or transcribed from the genome of a cell and a set of values ​​indicative of the expression levels of genes, including the mRNA levels of one or more genes.

[0072] As used herein, the term "dependence" means essentiality for cell growth or survival, and is used interchangeably with "essentiality."

[0073] The term "dependency gene" as used herein means a gene essential for cell proliferation or survival. More specifically, a dependency gene is a gene whose expression, when reduced or eliminated, leads to reduced proliferation and / or death of a cell, and means a gene on which a cell depends for its survival, and includes universal dependency genes that are understood to be universally essential for the survival of cancers or cancer cells of various types and / or origins, and / or cancer patient-specific dependency genes that are understood to be specifically essential for the survival of cancer cells derived from individual cancer patients. The dependency gene may mean a gene that is essential and constitutively expressed in a cell and is uniformly expressed in all individual cells.

[0074] As used herein, the term "universal dependency gene" refers to a gene that has been identified through in vitro data as being essential for the survival of a large number of diverse cancer cells, such as known cancer cell lines.

[0075] As used herein, the term "cancer patient-specific dependency gene" refers to a gene that is understood to be essential for the survival of cancer cells derived from an individual cancer patient.

[0076] The term "neoantigen" as used herein refers to a peptide that elicits an immune response. That is, the neoantigen is also an immunogenic peptide. The neoantigen may be derived by cancer cell-specific mutation and may be presented as an epitope of the cancer cell. It is an antigen that has at least one alteration that allows it to be differentiated from the corresponding wild-type, parental antigen through a mutation in the cancer cell or a post-translational modification specific to the cancer cell. The neoantigen may also include an amino acid sequence or a nucleotide sequence. The mutation may also include a frameshift or non-frameshift mutation, an indel, a missense or nonsense substitution, a splice site alteration, a genomic rearrangement or gene fusion, or any genomic or expression alteration that causes a neo-ORF.

[0077] Neoantigens derived from universally-addicted or cancer patient-specific-addicted genes do not disappear due to immune evasion caused by immunoediting of cancer cells, and therefore serve as effective therapeutic targets for cancer vaccines tailored to cancer patients that can bring about high immunotherapeutic effects in cancer patients, as well as effective diagnostic targets as markers related to the prognosis of immunotherapy.

[0078] The term "immunotherapy" as used herein refers to a therapeutic method using immune response. The immunotherapy is also used for cancer treatment. For example, the immunotherapy is also a treatment using immune checkpoint inhibitors, and the immune checkpoint inhibitors are anti-CTLA4 blockers or anti-PD-1 / PD-L1 blockers, but are not limited thereto, and are various immunotherapies.

[0079] The term "binding affinity" as used herein means the binding strength between a neoantigen peptide and the MHC of an antigen-presenting cell, and is also expressed by a CNN-MHC value. The "CNN-MHC value" is a value obtained by constructing a deep learning model based on experimental values ​​obtained by converting the binding strength between each amino acid between a neoantigen and an MHC into a matrix form, and means a probability value between 0 and 1 converted into a sigmoid activity function. Specifically, an immunogenic peptide capable of binding to an MHC class I protein or an MHC class II protein may have an MHC CNN-MHC value of 0.5 or more. In addition, the closer the CNN-MHC value is to 1, the stronger the binding strength between the MHC class I protein or the MHC class II protein and the immunogenic peptide.

[0080] The term "antigen-presenting cells" as used herein refers to cells that receive and process protein antigens, and then present antigen-derived peptide fragments, together with MHC class II, to T cells to activate them, and includes macrophages, B cells, dendritic cells, etc.

[0081] The term "cell survival dependency prediction model" as used herein means a model that predicts the probability of cell survival or death due to the reduction or elimination of the expression of an individual gene. Specifically, the model learns the effect of gene knockout / knockdown on cell survival by machine learning and predicts the effect of a specific gene on cell survival, and the machine learning is also performed using in vitro data of gene knockout / knockdown using RNAi or CRISPR / Cas9. By inputting a gene expression profile or sequence into the cell survival dependency prediction model, the cell survival dependency of the gene from which the sequence is derived can be predicted.

[0082] The term "neoantigen binding affinity prediction model" as used herein means a model for predicting the binding affinity between a neoantigen and an antigen-presenting cell, specifically, the binding affinity between a neoantigen and the MHC of an antigen-presenting cell. The binding affinity based on the amino acid interaction between a peptide sequence and an HLA sequence is learned by machine learning, and by inputting the peptide sequence of a neoantigen, the binding affinity between the peptide sequence and the HLA of an antigen-presenting cell can be predicted, and the neoantigen can be classified according to the set scale of binding affinity.

[0083] Example 1. Screening of cancer cell survival-dependent genes To determine whether a gene function is essential for cancer or cancer cell survival, high-throughput screening (HTS) can be performed using RNAi or CRISPR libraries that can perform knockout / knockdown of all genes. Specifically, cancer cells are transfected with shRNA libraries or CRISPR sgRNA, and after a certain period of time, deep sequencing is performed. By comparing the results with the initial sequencing results, it is possible to quantitatively profile which genes have been inactivated and which cells have died, and to select genes essential for cancer cell survival in vitro. Figure 1 shows a schematic diagram of a method for predicting genes essential for cancer cell survival.

[0084] Such methods continue to produce in vitro data on genes essential for cancer cell survival for an increasing number of cell lines, and dependency data on a significant number of cancer cell lines have been established for major solid cancers, such as lung, ovarian, colon, gastric, and breast cancers (https: / / depmap.org / portal / and https: / / depmap.sanger.ac.uk / ). Such data are used, according to one embodiment of the present invention, as data on universal dependency genes or for the purpose of in silico prediction of cancer patient-specific dependency genes based on deep learning.

[0085] In vitro data on cancer cell lines can be used to obtain universal dependency genes, while data derived from cancer patients, such as transcriptome data from cancer patients, can be applied to cell survival dependency prediction models learned from in vitro data on cancer cell lines to predict dependency for each patient sample and select dependency genes specific to cancer patients.

[0086] In addition, by applying a cell survival dependency prediction model to single-cell transcriptome data, it is possible to understand the dependency patterns of different cells due to tumor heterogeneity, and genes that show the same dependency in various cells can be selected as universal dependency genes. Genes that are essential for cancer cell survival and show uniform expression at the individual cell level are selected as dependency genes that can be used as effective diagnostics and / or targets.

[0087] The cell survival dependency prediction model is a model based on a deep learning neural network composed of an input layer, multiple hidden layers, and an output layer. When the in vitro data is input to the input layer, the neural network is configured such that the relationship between the gene expression pattern of the cell and cell death is learned in the multiple hidden layers, and a predetermined probability value is output in the output layer. At this time, the predetermined probability value includes a value indicating the probability that the cell will die and a value indicating the probability that the cell will grow.

[0088] If single-cell transcriptome data on cancer patients is available, it can be used to select cancer patient-specific dependency genes via cell survival dependency models; if not, previously published data on other cancer patient samples can be used to select universal dependency genes.

[0089] Example 2. Selection of neoantigens derived from universal addiction genes and their significance as diagnostic / therapeutic targets Neoantigens are peptides that are protein fragments that are not present in normal cells but are generated by cancer-specific mutations, which bind to MHC proteins, are delivered to the surface of cancer cells, and are recognized as antigens by immune cells. Neoantigens are a core element of immune anticancer therapy, and it is known that the more neoantigens there are, the better the response to immune anticancer therapy. The amount of neoantigens (neoantigen load) is used as a diagnostic marker. However, since the neoantigens are derived from various genes and have different properties, their usefulness as diagnostic markers and therapeutic targets will differ. In this example, as described in Example 1, neoantigens derived from cancer cell survival-dependent genes were selected as useful neoantigens that can maximize the effect of immune anticancer therapy, and their immunotherapy reactivity was confirmed.

[0090] Specifically, immune checkpoint inhibitor regimen cohorts listed in Table 1 were utilized to compare immunotherapy responsiveness.

[0091] [Table 1]

[0092] 2-1. Pre-therapy cohort The usefulness of a method for screening neoantigens based on the dependency on cancer cell survival according to one embodiment of the present invention was verified using in vitro dependency data and expression uniformity data of published cohorts of lung cancer and melanoma with only pretherapy results. The analysis results are shown in FIG. 5. Specifically, after arranging the genes from which the neoantigens are derived according to their essentiality for cancer cell survival, 500 to 2,000 genes with high and low dependency were selected, and the differential neoantigen load per gene between patients with good prognosis and those with poor prognosis for immune anti-cancer therapy was calculated using the formula:

[0093]

number

[0094] The difference in the number of neoantigens was calculated using the expression homogeneity and gene expression level. It can be seen that the greater the difference, the better the explanatory power of the treatment prognosis. The gray dotted line shows the difference in the number of neoantigens calculated for all genes using the existing standard method. As a result, it can be seen that the use of a small number of highly essential genes or uniformly expressed genes shows better explanatory power than the use of all genes, and the explanatory power is not good when low essential genes or non-uniformly expressed genes are used. Similarly, the difference in the number of neoantigens was calculated using the expression homogeneity and gene expression level. This shows that gene essentiality and expression homogeneity have better explanatory power of the treatment prognosis than the gene expression level itself. Specifically, after arranging the genes from which the neoantigens are derived according to their essentiality for cancer cell survival, 500 to 2,000 genes with high and low dependency were selected, and the differential neoantigen load between patients with good prognosis and those with poor prognosis for gene-specific immune anticancer therapy was calculated. The larger this difference is, the better the explanatory power for treatment prognosis is. The gray dotted line shows the difference in the neoantigen load calculated for all genes by the existing standard method. As a result, it can be seen that the use of a small number of highly essential genes or uniformly expressed genes shows better explanatory power than the use of all genes, and the explanatory power is not good when low essential genes or non-uniformly expressed genes are used. Similarly, the difference in the neoantigen load was calculated according to expression homogeneity and gene expression level. This indicates that gene essentiality and expression uniformity have better explanatory power for treatment prognosis than the gene expression value itself.

[0095] Such results imply that immune evasion mechanisms may occur more actively when the function of the gene from which the immune response-focused neoantigen is derived is not essential for the proliferation and survival of cancer. Figure 6 shows a schematic diagram of the immune evasion mechanism of cancer cells.

[0096] 2-2. Pre-therapy and on-therapy cohorts Using data from the Riaz cohort, which had data before and during immune checkpoint inhibitor treatment, we analyzed the dependency of neoantigen-derived genes and immunotherapy responsiveness, and the results are shown in Figure 7.

[0097] Specifically, in the melanoma immune checkpoint inhibitor cohort (Riaz), we compare clonal changes and gene expression changes before and after treatment, dividing them into the essentiality of the gene function from which the neoantigen originates for cancer survival (high dependency vs. low dependency) and the homogeneity of gene expression (homogenous vs. heterogenous expression). CR / PR and SD / PD mean positive and negative prognosis, respectively.

[0098] In the patient group with a good response to treatment (CR (complete remission) / PR (partial remission)), neoantigens derived from genes with uniform expression and dependent genes essential for cancer cell survival were the main targets of the anti-cancer immune response, resulting in clonal contraction and a decrease in expression at the RNA level. On the other hand, in the patient group with a poor response to treatment (SD (stable disease) and PD (progressive disease)), neoantigens derived from genes with heterogeneous expression and genes that are not essential for cancer cell survival were the main targets of immune attack, and gene expression was reduced through immune editing, resulting in successful immune evasion, which in turn led to clonal expansion.

[0099] 2-3.MSK-IMPACT (integrated mutation profiling of actionable cancer targets) We used MSK-IMPACT to examine neoantigen-derived gene dependency and immunotherapy responsiveness in various cancer types.

[0100] Previous research measured the mutation load (proportional to the amount of neoantigens) of 1,662 patients with a variety of cancers who received immune checkpoint inhibitor treatment, focusing only on 468 genes belonging to the so-called MSKCC (Memorial Sloan Kettering Cancer Center) panel, and revealed that the load is an important determinant of immunotherapy responsiveness (Nat. Genet. 51: 202-206, 2019).

[0101] In this embodiment, the top 50% of genes with high dependency on cancer cell survival among 468 genes in the MSKCC panel were selected, and it was confirmed that the mutation load was more highly correlated with immune checkpoint inhibitor treatment responsiveness than when all 468 genes were used. The results are shown in the upper part of FIG. 8. Survival rates after treatment were compared using the results of survival analysis performed using the entire MSKCC panel (all), the top 50% genes with high dependency on cancer cell survival (high fitness), and the bottom 50% genes (low fitness), and HR (hazard ratio) values ​​and p-values ​​are shown. The lower the HR and the lower the p-value, the higher the explanatory power of the gene group in relation to treatment prognosis. The mutation load of 226 genes belonging to the top 50% with high dependency on cancer cell survival showed a higher correlation with immune treatment responsiveness than the mutation load of all 486 genes.

[0102] In addition, the middle and bottom of Figure 8 show the results of analyzing published immunotherapy cohorts for lung cancer and melanoma by all genes, or the top 500 and bottom 500 genes essential for cancer cell survival, respectively. For each gene group, patients were divided into those with high and low mutation burden or neoantigen load, and survival rates after treatment were compared, and HR and p-values ​​are shown. As a result, it can be seen that using a small number of gene groups highly dependent on cancer cell survival provides a better explanation of treatment prognosis than using the entire gene group by standard methods.

[0103] Taking these results together, we confirmed that the essentiality of genes from which antigens (including neoantigens and surface antigens) are derived in cancer cells and the uniformity of their expression patterns among individual cells in tissues play a very important role in the immune response to cancer. This indicates that in terms of treatment, inducing the concentration of essential genes involved in cancer survival and neoantigens derived from genes that show uniform expression is essential to minimize immune evasion of cancer, and that the amount of such neoantigens can also be used as a marker to predict the prognosis of immunotherapy.

[0104] Example 3. Selection of neoantigens derived from cancer patient-specific addiction genes and their significance as diagnostic / therapeutic targets In this embodiment, data obtained from samples from specific cancer patients was applied to a cell survival dependency prediction model to obtain neoantigens derived from selected cancer patient-specific dependency genes, and the association between such neoantigens and cancer patient survival was confirmed.

[0105] In Example 2, the relevance of neoantigens derived from universally dependent genes to the survival of immunotherapy patients was examined. The dependency data used in Example 2 was derived from the in vitro cancer cell line experiment described in Example 1, and was obtained by digging up genes that are universally dependent in various cancer cell lines, rather than dependency on a given patient. As described above, a cell survival dependency prediction model that predicts the dependency of cell survival on gene expression is generated by learning the relationship between gene expression patterns and cell death from the in vitro dependency data, and gene expression profiles of cancer patients are input into the prediction model to select cancer patient-specific dependency genes. In order to confirm the significance of neoantigens obtained based on such cancer patient-specific in silico dependency data as diagnostic / therapeutic targets, transcriptome data (https: / / portal.gdc.cancer.gov / ) related to lung cancer patients and breast cancer patients published in TCGA (The Cancer Genome Atlas) was used. Because these were not patients who had actually received immune anti-cancer therapy, we divided the samples into those with high immune cell penetrance (high leukocyte fraction) and those with low immune cell penetrance (low leukocyte fraction), and hypothesized that the samples with high immune cell penetrance would show effects similar to those of immune therapy.

[0106] The results are shown in Figure 9. The number of neoantigens derived from cancer patient-specific dependency genes (high dependency), which are essential for the proliferation and survival of cancer cells (black bars in Figure 9), had higher explanatory power than the total number of neoantigens (horizontal line in Figure 9) and the number of neoantigens derived from non-dependency genes (low dependency), which are not useful for survival (gray bars in Figure 9). In particular, such results were observed only in samples with high immune cell penetration, and it was confirmed that the cancer patient-specific dependency data generally showed a greater difference in explanatory power compared to the universal dependency data.

[0107] Example 4. Model for predicting binding strength between neoantigens and antigen-presenting cells For neoantigens to be immunotherapeutic reactive, they must be processed by antigen-presenting cells and bind to HLA on the cell surface.

[0108] To construct and verify a model for predicting the binding strength between neoantigens and antigen-presenting cells, we used benchmark data from the Immune Epitope Database (IEDB) used in previous studies and compared the predictive power results of existing machine learning algorithms published in the IEDB.

[0109] The binding strength prediction model between neoantigens and antigen-presenting cells is a CNN model that includes multiple convolution layers, a fully connected layer, and an output layer, and does not include a pooling layer because all output values ​​of the convolution layer are used for prediction.

[0110] The multiple convolution layers extract interaction features from input data, and the input data refers to map data including parameters indicating the binding affinity between amino acids of a peptide and amino acids of HLA. The multiple convolution layers perform convolution using a specific number of kernels or weight matrices. The full connection layer receives and integrates the output values ​​of the convolution layers as input, and the output layer uses a sigmoid function to output information related to the binding possibility. Figure 10 is a schematic diagram of a model (CNN-MHC) for predicting the binding affinity between a neoantigen and MHC used in a method for screening a neoantigen according to one embodiment of the present invention.

[0111] The constructed model was trained using over 50,000 peptide-MHC in vitro binding experiment results from the IEDB, and its performance was evaluated using test data sets that are newly released by the IEDB every week. As a result, it was confirmed that the model outperformed not only existing algorithms such as SMMPMBEC, ANN, and NetMHCcon, but also the most widely used NetMHCpan, for over 70% of the test data sets. The results are shown in Figure 11.

[0112] 4-1. Confirmation of binding between neoantigens and HLA In order to confirm the actual binding of neoantigens to HLA molecules predicted using the model (CNN-MHC) of this embodiment, the binding ability of neoantigens to HLA was tested. Specifically, the binding ability of peptides predicted to bind to HLA-A02 was analyzed using a T2 cell line (ATCC CRL-1992) expressing HLA-A02. Peptides predicted to bind to HLA-A02 are shown in Table 2 (SEQ ID NO: 1 to SEQ ID NO: 50). In Table 2, the CNN-MHC value is a value obtained by constructing a deep learning model based on experimental values ​​in which the binding strength between each amino acid of a neoantigen and MHC is converted into a matrix form, and means a value converted into a probability of 0 and 1, and the NetMHC value is the latest version of the NetMHC tool commonly used to predict the binding of neoantigens, and means a binding prediction value between MHC proteins and peptides calculated by NetMHCPan-4.1, a model trained based on mass spectrometer data. The higher the two values, the higher the binding probability. Specifically, cultured T2 cell line (1x10 6 After treatment with 50μg / ml of peptide for 24 hours, APCs were stained with labeled HLA-A2 monoclonal antibody. The binding ability was confirmed by analysis using a flow cytometer. DMSO was used as a negative control, and Mart-1 and NY-ESO were used as positive controls.

[0113] [Table 2]

[0114] Figure 12 shows the results of MHC binding analysis. 80% of the peptides predicted by the model of this example (CNN-MHC) were confirmed to actually bind to HLA-A02. When compared with the widely used prediction value of NetMHCpan-4.1 (http: / / www.cbs.dtu.dk / services / NetMHCpan-4.1 / ), the model of this example has better predictive power.

[0115] Meanwhile, the above-mentioned embodiment may also be performed by a processor 120 executing at least one instruction stored in the memory 110 of the system as shown in Fig. 3. Referring to Fig. 4, the processor 120 may also include a data acquisition unit 121 for acquiring sequence analysis data of an exome, a transcriptome, or a whole genome, a prediction model generation unit 122 for generating a cell survival-dependent prediction model and an avidity prediction model, a gene selection unit 123 for selecting cancer cell survival-dependent genes using the cell survival-dependent prediction model, and a neoantigen selection unit 124 for collecting neoantigens from gene expression profiles of cancer patients and selecting neoantigens having binding affinity with HLA of antigen-presenting cells using the avidity prediction model.

[0116] The cell survival dependence prediction model and the avidity prediction model can be updated based on new input data by repeating the instructions of the prediction model generation unit 122. The prediction models generated and updated via the prediction model generation unit are also stored in the memory.

[0117] Example 5. Anti-cancer vaccine containing neoantigens As described in the previous examples, anti-cancer vaccines can be produced that contain neoantigens derived from cancer cell survival-dependent genes.

[0118] Exome sequencing of mouse cell lines transplanted with cancer cells or tissues can be performed to find mutations not found in normal cells, and these mutations can be applied to a binding strength prediction model with antigen-presenting cells to select neoantigen candidates. Using a cell survival dependency model, the essentiality of the gene from which the neoantigen is derived for cancer cell survival can be determined from dependency data related to the carcinoma, and the neoantigens can be arranged according to essentiality, and the top 5 and bottom 5 can be selected. All possible peptide sequences consisting of 9 to 30 amino acids including the selected neoantigen site can be generated, and sequences with chemical properties suitable for peptide synthesis, for example, high hydrophilicity (Kyte-Doolittle GRAVY<0) and low instability index (InstaIndex<40) can be selected to synthesize peptides, thereby producing an anti-cancer vaccine against the neoantigen.

[0119] 5-1. Selection of neoantigens Using the cell survival dependency prediction model and the binding strength prediction model between neoantigens and antigen-presenting cells described in Examples 1 to 4, the exome and transcriptome information of cancer cell lines was analyzed, and mutations derived from dependency genes that are highly essential for cell survival and have HLA binding ability were selected.

[0120] 5-2. Confirmation of immune response to neoantigens To confirm whether or not neoantigens induce immune responses, we tested whether or not T cells were generated that recognize neoantigen peptides containing the neoantigen peptides selected in 5-1 and that had predicted or experimentally verified HLA binding ability. Specifically, to confirm the T cell pool that exhibits specificity, mice transplanted with mouse cancer cell lines were used. The mouse model consisted of 1x10 mouse lung cancer cell line LLC-1 (ATCC CRL-1642) 6 was produced by subcutaneous injection into the flank of 6-week-old, 20 g-weighing male C57BL / 6 mice.

[0121] Specifically, 15 types of mutations predicted to have high cell survival essentiality and high binding strength to antigen-presenting cells through the analysis of exome and transcriptome of cancer cell lines were selected as a group of neoantigens capable of predicting HLA binding and CD8+ T cell reactivity. The selected peptide information is summarized in Table 3 (SEQ ID NOs: 51 to 65). The CNN-MHC values ​​listed in Table 3 are values ​​obtained by constructing a deep learning model based on experimental values ​​in which the binding strength between each amino acid of neoantigen and MHC is converted into a matrix form, and are values ​​converted into probabilities of 0 and 1, and the higher the value, the higher the probability of binding. Using spleen cells extracted from the spleens of mice vaccinated with 9mer or 15mer peptides synthesized in the mutant sequence, ELISpot analysis based on IFNγ secretion was performed to confirm the CD8+ T cell response to the neoantigen. IFNγ secretion by T cells was confirmed by 10 peptides (66.7%) out of 15 candidate neoantigen peptides predicted to be reactive. The results are shown in Figure 13. It was confirmed that the neoantigens selected using the method according to one embodiment of the present invention can actually bind to HLA and effectively induce immune responses.

[0122] [Table 3]

[0123] Example 6. Prediction of prognosis of cancer patients using neoantigens As described in the above examples, neoantigens derived from cancer cell survival-dependent genes can be used to predict the treatment prognosis of cancer patients.

[0124] Exome sequencing of patient samples is performed to find mutations that are not present in normal cells but are present only in cancer cells, and these mutations are applied to a model predicting binding strength with antigen-presenting cells to select neoantigen candidates. Using a cell survival dependency model, the essentiality of the gene from which the neoantigen is derived from the cancer cell survival is determined from the dependency data related to the carcinoma, the neoantigens are arranged according to their essentiality for survival, and the same number of genes (e.g., 500) are selected from the top and bottom. A group of patients who have received immune anti-cancer therapy is classified into a group of patients who responded to the immune anti-cancer therapy and a group of patients who did not respond, and the numbers of neoantigens derived from the selected top and bottom essential genes are compared, and it can be determined whether the number of neoantigens derived from the top essential genes is highly related to the patient's treatment prognosis. By repeatedly determining the correlation between the number of neoantigens and the patient's prognosis while changing the number of selected genes, the dependent gene-derived neoantigens required for predicting the patient's prognosis can be determined.

[0125] Figure 8 shows that the amount of mutations derived from genes that are essential for cancer cell survival and the number of neoantigens have a high explanatory power for treatment prognosis and can also be used to predict the treatment prognosis of cancer patients.

[0126] The above description of the present invention is for illustrative purposes only, and those skilled in the art to which the present invention pertains can easily understand that the present invention can be easily modified into other specific forms without changing the technical idea or essential features of the present invention. Therefore, the above-described specific examples should be understood to be illustrative in all respects and not limiting. For example, each component described as being single may also be implemented in a distributed form, and similarly, each component described as being distributed may also be implemented in a combined form.

Claims

1. Obtaining sequence analysis data of an exome, a transcriptome, a single cell transcriptome, a peptidome, or a whole genome from a cancer patient sample; selecting a cancer cell survival-dependent gene; obtaining a neoantigen derived from said cancer cell survival-dependent gene; The step of selecting the cancer cell survival-dependent genes includes selecting the cancer cell survival-dependent genes essential for the survival of the cancer cells using a cell survival-dependent prediction model; The cell survival dependence prediction model is generated by learning the relationship between cellular gene expression and cell death, and a gene that causes cancer cell death by reducing or eliminating the expression level of the gene is selected as a cancer cell survival dependence gene; The step of obtaining a neoantigen derived from the cancer cell survival-dependent gene comprises: Obtaining neoantigens from the cancer patient by comparing sequences from cancer cells with sequences from normal cells from the sequence analysis data obtained from the cancer patient, and selecting as neoantigens the sequences from the cancer cells having at least one nonsynonymous mutation compared to the sequences from the normal cells; and collecting, from among the obtained neoantigens, neoantigens derived from the cancer cell survival-dependent genes.

2. The method of claim 1, further comprising determining the binding affinity of the neoantigen to HLA of an antigen-presenting cell.

3. The method of claim 1, wherein the selected cancer cell survival-dependent gene, when its expression level is reduced or when it is eliminated, causes cancer cell death but does not affect the survival of normal cells.

4. The method of claim 1, wherein the relationship between cellular gene expression and cell death is based on in vitro or in silico data regarding whether targeted reduction or ablation of gene expression results in the death of cancer cell lines.

5. The method of claim 1, wherein the step of selecting the cancer cell survival-dependent gene further comprises a step of determining whether the cancer cell survival-dependent gene selected from the cell survival-dependent prediction model is uniformly expressed in all cancer cells obtained from a cancer patient.

6. The method of claim 1 , wherein the neoantigen is cancer patient-specific.

7. The step of determining the binding affinity between the neoantigen and the HLA of the antigen-presenting cell comprises: inputting the sequence of the neoantigen into a neoantigen binding affinity prediction model that predicts the binding affinity between an antigen peptide and HLA of an antigen-presenting cell, and obtaining a binding affinity prediction; The neoantigen binding affinity prediction model is generated by learning with interaction data between amino acids of a peptide and amino acids of HLA, The method of claim 2 , wherein the HLA is MHC class I or MHC class II.

8. The method of claim 7, wherein the antigen presenting cells are dendritic cells, macrophages, B cells, or a combination thereof.

9. The method of claim 7, wherein the binding affinity between the neoantigen and the HLA of the antigen-presenting cell is determined to be present if the CNN-MHC number is >0.

5.

10. a memory for storing at least one instruction; at least one processor for executing the at least one instruction stored in the memory; The processor executes the at least one instruction to: A cell survival dependency prediction model is generated by learning the relationship between the amount of gene expression in a cell and cell death, and predicting the dependency of cell survival on gene expression; The cell survival dependence prediction model is generated by learning the relationship between the gene expression level of a cell and cell death, and selects genes that cause cancer cell death by reducing or removing their expression levels as cancer cell survival dependence genes, compares the gene expression profile of a cancer patient with the gene expression profile of a normal cell or a normal control, and selects, as a neoantigen, a sequence of the cancer patient's cell that has at least one nonsynonymous mutation compared to the sequence of the normal cell or the normal control, thereby obtaining a neoantigen, Here, the gene expression profile may be an exome, a transcriptome, a single cell transcriptome, or a cell transcriptome, peptidome or whole genome sequence analysis data; collecting neoantigens derived from said cancer cell survival-dependent genes; A gene expression profile of a cancer patient is input into the cell survival dependence prediction model, a cancer cell survival dependence gene is selected, and a neoantigen derived from the cancer cell survival dependence gene is obtained; A neoantigen binding affinity prediction model is generated that predicts binding affinity based on amino acid interactions between the peptide and the antigen-presenting cell; A neoantigen screening system that uses the neoantigen binding affinity prediction model to select neoantigens that have binding affinity for the HLA of antigen-presenting cells.

11. The processor executes the at least one instruction to: The system according to claim 10, wherein if the CNN-MHC number between the neoantigen and the HLA of the antigen-presenting cell is >0.5, the neoantigen is selected as having binding affinity.

12. The system of claim 10, wherein the processor executes the at least one instruction to learn the relationship between gene expression and cell death, and the relationship between the binding affinity of neoantigens to HLA of antigen-presenting cells, respectively.

13. The system of claim 10, wherein the relationship between cellular gene expression and cell death is based on in vitro or in silico data regarding whether targeted reduction or removal of gene expression results in the death of a cancer cell line.

14. The system according to claim 10 , wherein the neoantigen binding affinity prediction model is generated by learning with interaction data between amino acids of a peptide and amino acids of HLA.

15. Obtaining a neoantigen by the method according to any one of claims 1 to 9; and producing an anti-cancer vaccine comprising the neoantigen, The step of producing the anti-cancer vaccine comprises: obtaining a peptide sequence consisting of 9 to 30 amino acids comprising the neoantigen; The method for producing an anti-cancer vaccine further comprises the step of selecting a peptide sequence having hydrophilicity and stability from the peptide sequences.

16. The method of claim 15, wherein the selected peptide sequences have a Kyte-Doolittle GRAVY<0 and an InstaIndex<40.

17. Obtaining a neoantigen by the method according to any one of claims 1 to 9; measuring the amount of said neoantigen in a sample from a cancer patient; and comparing the amount of the neoantigen in the cancer patient sample with the amount of the neoantigen obtained from a control group of cancer patients whose prognosis for treatment has been confirmed; When the amount of the neoantigen in the sample from the cancer patient is greater than the amount of the neoantigen obtained from a control group of cancer patients with poor prognosis, the prognosis of the cancer patient is predicted to be good; A method for predicting the treatment prognosis of a cancer patient, in which if the amount of the neoantigen in a sample from the cancer patient is smaller than the amount of the neoantigen obtained from a control group of cancer patients with a good treatment prognosis, the treatment prognosis of the cancer patient is predicted to be poor.

Images