Scoring model, method and system for screening tumor neoantigen peptides

Abstract

The invention relates to a scoring model, method and system for screening tumor neoantigen peptides, and the scoring model is used for screening tumor neoantigen peptides. The construction method comprises the following steps: acquiring whole exome sequencing data of tumor tissues and normal tissues of a plurality of subjects, whole transcriptome sequencing data of the tumor tissues and immunogenicity results of variant polypeptide fragments; processing the sequencing data to obtain a score feature corresponding to each variant polypeptide fragment; and screening the scoring features by using a Lasso regression model to obtain scoring key features and regression coefficients thereof, and establishing a tumor neoantigen peptide scoring model. The scoring model constructed by the invention has better prediction efficiency, and is suitable for screening personalized tumor neoantigen peptides.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of tumor immunity, in particular to a scoring model, method and system for screening tumor neoantigen peptides. BACKGROUND

[0002] Tumor therapeutic vaccine as a frontier field of tumor immunotherapy has attracted extensive attention. The core of tumor therapeutic vaccine is to introduce tumor neoantigen into the body of patients to overcome the immunosuppression caused by tumors, so as to rebuild or enhance the immune response, thereby improving the recognition and killing ability of the immune system to tumor cells expressing such neoantigens, and achieving the purpose of controlling or eliminating tumors. Compared with traditional tumor treatment methods, tumor therapeutic vaccine can effectively activate the immune system without damaging normal cells, showing great application potential and development prospects.

[0003] Tumor neoantigens are derived from gene mutations produced by tumor cells during development, and the DNA or RNA fragments carrying mutations are translated into polypeptides which are recognized and presented to immune cells by human leukocyte antigen (HLA) due to their immunogenicity, thereby producing specific anti-tumor immune response. Therefore, accurately screening specific tumor neoantigens is crucial for the successful development of tumor therapeutic vaccine, and if the neoantigens cannot be accurately predicted, the subsequent treatment effect based on the prediction results will be affected. However, this process faces many challenges: there are many somatic mutations of tumor cells, and not all mutations can produce immune response; in addition, the process of experimental screening of neoantigens is not only time-consuming and labor-intensive, but also costly.

[0004] With the rapid development of second-generation sequencing technology, combined with advanced bioinformatics tools, the somatic mutation information of tumor patients and HLA typing can be efficiently obtained at present. However, it is still a complex problem to accurately predict the immunogenicity of neoantigens from massive mutation information. The immune recognition of neoantigens involves multiple links, including the expression of mutations, the processing of polypeptides, the combination with HLA molecules, and the final presentation. At present, most neoantigen prediction models only consider part of these biological characteristics, resulting in limited accuracy and efficiency of prediction. For example, the international tumor neoantigen screening alliance (Tumor Neoantigen Selection Alliance, TESLA) found that (Wells DK, et al. Cell. 2020;183(3):818-834.e13.doi:10.1016 / j.cell.2020.09.015) among the 608 variant polypeptide fragments with high scores predicted by 25 teams, only 37 variant polypeptide fragments (6.1%) were immunogenic; among the top 100 variant polypeptide fragments predicted by each team (combined with the prediction scores of all specific subjects), none of the teams predicted more than 20 of the 37 variant polypeptide fragments with immunogenicity; further analysis of the proportion of each team in the specific subject who included the variant polypeptide fragments with immunogenicity in the top 100 scores showed that it was concentrated in less than 50%, with a median of 33.3%.

[0005] Therefore, researchers urgently need to develop more efficient computer prediction algorithms to assist in screening tumor neoantigens. These algorithms need to consider the biological characteristics and immunological features of polypeptides, improve the accuracy of prediction, and reduce the burden of experimental screening, thereby providing a solid foundation for the development of tumor therapeutic vaccines. The present application is proposed in such a background, which is a new tumor neoantigen prediction method, aiming to improve the efficiency and quality of tumor neoantigen screening through comprehensive analysis and computational model. SUMMARY

[0006] The purpose of the present application is to provide a scoring model, method and system for screening tumor neoantigen peptides, which can effectively improve the accuracy and clinical applicability of screening tumor neoantigen peptides.

[0007] To achieve the above object, the present application provides a method for constructing a tumor neoantigen peptide scoring model, comprising: obtaining whole exome sequencing data of tumor tissue and normal tissue, whole transcriptome sequencing data of tumor tissue and immunogenicity results of variant polypeptide fragments of a plurality of subjects; processing the sequencing data to obtain scoring features corresponding to each variant polypeptide fragment; screening the scoring features using a Lasso regression model to obtain scoring key features and their regression coefficients, and establishing a tumor neoantigen peptide scoring model.

[0008] In some embodiments, wherein the immunogenicity results of the variant polypeptide fragments are obtained by immunological experiments.

[0009] In some embodiments, wherein the processing of the sequencing data to obtain scoring features corresponding to each variant polypeptide fragment comprises: aligning the whole exome sequencing data of the tumor tissue and the normal tissue with a reference genome to obtain DNA alignment results; aligning the whole transcriptome sequencing data of the tumor tissue with a reference genome to obtain RNA alignment results; analyzing somatic mutation sites in the tumor tissue relative to the normal tissue according to the DNA alignment results, and predicting the HLA typing of the subject; calculating variant polypeptide fragment features according to variant polypeptide fragments with known immunogenicity results; obtaining expression features of somatic mutations corresponding to the variant polypeptide fragments according to the somatic mutation sites, the DNA alignment results and the RNA alignment results; predicting the binding features of the variant polypeptide fragments and HLA molecules based on the variant polypeptide fragments and the HLA typing results.

[0010] In some embodiments, wherein the variant polypeptide fragment features at least include the length of the variant polypeptide fragment and the hydrophobic value of the variant polypeptide fragment.

[0011] In some embodiments, wherein the expression features of somatic mutations corresponding to the variant polypeptide fragments at least include the variant allele frequency at the DNA level, the corrected variant allele frequency at the DNA level, the total number of sequencing fragments covering the mutation site in the transcriptome sequencing, the number of sequencing fragments of the variant allele in the transcriptome sequencing, the variant allele frequency at the RNA level, the expression amount of the gene where the mutation is located, and the tumor purity.

[0012] In some embodiments, wherein the binding features of the variant polypeptide fragments and HLA molecules at least include the pMHC affinity value and the pMHC ligand elution value.

[0013] In some embodiments, wherein for MHC-I type variant polypeptide fragments, the binding features of the variant polypeptide fragments and HLA molecules at least further include antigen processing scores and antigen presentation scores.

[0014] In some embodiments, wherein before the score features are screened using the Lasso regression model, the method further comprises preliminarily filtering the variant polypeptide fragments according to the binding features of the variant polypeptide fragments to HLA molecules, and the preliminary filtering criteria are: for MHC-I type variant polypeptide fragments, the pMHC affinity value predicted by MHC-I type affinity prediction software is not higher than 500 nM; for MHC-II type variant polypeptide fragments, the pMHC affinity value predicted by MHC-II type affinity prediction software is not higher than 200 nM.

[0015] In some embodiments, wherein the score features are screened using the Lasso regression model, the score key features and their regression coefficients are obtained, and the tumor neoantigen peptide scoring model is established, and the tumor neoantigen peptide scoring model is established for MHC-I type variant polypeptide fragments and MHC-II type variant polypeptide fragments respectively.

[0016] In some embodiments, wherein before the score features are screened using the Lasso regression model, the method further comprises normalizing the values of the score features.

[0017] In some embodiments, wherein the Lasso regression model selects the optimal penalty parameter based on the cross-validation method, so as to determine the score key features and their regression coefficients for establishing the tumor neoantigen peptide scoring model.

[0018] In some embodiments, wherein the Lasso regression model selects the penalty parameter corresponding to the slowest increase in the average value of the area under the curve as the optimal penalty parameter.

[0019] In some embodiments, wherein the tumor neoantigen peptide scoring model is expressed in the form of a predicted probability:

[0020]

[0021] wherein,

[0022] MScore =∑(βi×Xi) + β0

[0023] Xi is the value of the normalized score key feature, βi is its regression coefficient, and β0 is the intercept term.

[0024] In some embodiments, wherein the score key features of the MHC-I type tumor neoantigen peptide scoring model include the variant allele frequency at the RNA level, the pMHC affinity value, the pMHC ligand elution value, the antigen processing score, and the antigen presentation score; and the score key features of the MHC-II type tumor neoantigen peptide scoring model include the length of the variant polypeptide fragment, the variant allele frequency at the RNA level, the pMHC affinity value, and the pMHC ligand elution value.

[0025] The second aspect of the present application provides a tumor neoantigen peptide scoring model, which is constructed by the method of the first aspect of the present application.

[0026] The third aspect of the present application provides the use of the tumor neoantigen peptide scoring model of the second aspect of the present application in screening tumor neoantigen peptides.

[0027] The fourth aspect of the present application provides a method for screening tumor neoantigen peptides, which comprises: obtaining whole exome sequencing data of tumor tissue and normal tissue of a subject to be tested, whole transcriptome sequencing data of the tumor tissue; processing the sequencing data to obtain variant polypeptide fragments, and extracting scoring features corresponding to each variant polypeptide fragment; inputting the scoring features corresponding to each variant polypeptide fragment as input data into the scoring model constructed according to the method of the first aspect of the present application or the scoring model of the second aspect of the present application, and calculating the prediction probability of each variant polypeptide fragment constituting a tumor neoantigen peptide; according to the prediction probability from high to low, selecting a variant polypeptide fragment with high prediction probability of constituting a tumor neoantigen peptide as a predicted tumor neoantigen peptide.

[0028] In some embodiments, wherein the processing of the sequencing data, obtaining variant polypeptide fragments, and extracting scoring features corresponding to each variant polypeptide fragment comprise: aligning the whole exome sequencing data of the tumor tissue and the normal tissue with a reference genome to obtain DNA alignment results; aligning the whole transcriptome sequencing data of the tumor tissue with a reference genome to obtain RNA alignment results; analyzing the somatic mutation sites in the tumor tissue relative to the normal tissue according to the DNA alignment results, and predicting the HLA typing of the subject; obtaining variant polypeptide fragments according to the somatic mutation sites in the tumor tissue relative to the normal tissue; predicting the binding of the variant polypeptide fragments to MHC class I molecules or to MHC class II molecules based on the variant polypeptide fragments and the HLA typing results; and extracting scoring features in the MHC-I type tumor neoantigen peptide scoring model or the MHC-II type tumor neoantigen peptide scoring model according to the results of the binding of the variant polypeptide fragments to MHC class I molecules or to MHC class II molecules.

[0029] In some embodiments, wherein the score features in the MHC-I type tumor neoantigen peptide scoring model or the MHC-II type tumor neoantigen peptide scoring model are extracted according to the results of the variant polypeptide fragment binding to MHC-I class molecule or binding to MHC-II class molecule, the score features extracted at least include, for the MHC-I type variant polypeptide fragment, the variant allele frequency at RNA level obtained according to the somatic mutation site and the RNA alignment result, and the pMHC affinity value, the pMHC ligand elution value, the antigen processing score and the antigen presentation score obtained according to the variant polypeptide fragment and the HLA typing result; for the MHC-II type variant polypeptide fragment, the score features extracted at least include the variant polypeptide fragment length obtained according to the variant polypeptide fragment, the variant allele frequency at RNA level obtained according to the somatic mutation site and the RNA alignment result, and the pMHC affinity value and the pMHC ligand elution value obtained according to the variant polypeptide fragment and the HLA typing result.

[0030] The fifth aspect of the present application provides a tumor neoantigen peptide scoring model construction system, which comprises the following modules: a data acquisition module for acquiring whole exome sequencing data of tumor tissue and normal tissue, whole transcriptome sequencing data of tumor tissue and immunogenicity results of variant polypeptide fragments of a plurality of subjects; a data processing module for processing the sequencing data to obtain score features corresponding to each variant polypeptide fragment; a scoring model training module for screening the score features using a Lasso regression model to obtain score key features and their regression coefficients, and establishing a tumor neoantigen peptide scoring model.

[0031] The sixth aspect of the present application provides a tumor neoantigen peptide screening system, which comprises the following modules: a data acquisition module for obtaining whole exome sequencing data of tumor tissue and normal tissue, whole transcriptome sequencing data of tumor tissue of a subject to be tested; a data processing module for processing the sequencing data to obtain variant polypeptide fragments, and extracting score features corresponding to each variant polypeptide fragment; a scoring module for inputting the score features corresponding to each variant polypeptide fragment as input data into the scoring model constructed according to the system of the fifth aspect of the present application or the scoring model of the second aspect of the present application, calculating the prediction probability of each variant polypeptide fragment constituting a tumor neoantigen peptide; an output module for sorting according to the prediction probability from high to low, and selecting the variant polypeptide fragment with high prediction probability of constituting a tumor neoantigen peptide as a predicted tumor neoantigen peptide.

[0032] The seventh aspect of the present application provides a computing device comprising: a memory configured to store one or more programs; and a processing unit coupled to the memory and configured to execute the one or more programs to cause the device to perform the steps of the method of the first aspect or the third aspect of the present application.

[0033] The eighth aspect of the present application provides a computer-readable storage medium having stored thereon machine executable instructions that, when executed, cause a machine to perform the steps of the method of the first aspect or the third aspect of the present application.

[0034] The present application systematically evaluates a large number of tumor neoantigen peptide responses of subjects, and generates a scoring feature for inputting a tumor neoantigen peptide scoring model based on multi-dimensional data such as clinical experimental data and genomic mutation information of paired tissue samples of the subjects, thereby improving the accuracy of the prediction results, and the accuracy of predicting MHC-I type tumor neoantigen peptides and MHC-II type tumor neoantigen peptides is 30.8% to 88.9%, with an average of 67.1%, which is significantly higher than the average level of existing tumor neoantigen prediction methods; the tumor neoantigen peptide scoring model of the present application is trained based on the Lasso dimension reduction algorithm, which reduces the complexity of the model by introducing L1 regularization, and improves the generalization ability and interpretability of the model, while preventing overfitting, and since the model has low computational complexity, the tumor neoantigen scoring model provided by the present application can be applied to large-scale data sets. Thus, the tumor neoantigen screening method of the present application not only improves the accuracy of predicting clinically effective tumor neoantigens, but also has good clinical applicability and generalization. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 : AUC plot for screening MHC-I type tumor neoantigen peptide scoring model, the best penalty parameter is selected in the green box.

[0036] Figure 2 : AUC plot for screening MHC-II type tumor neoantigen peptide scoring model, the best penalty parameter is selected in the green box.

[0037] Figure 3 : Statistical results of MHC-I type tumor neoantigen peptide scoring model predicting tumor neoantigens. Neoantigen refers to the number of true positive tumor neoantigens identified by ELISPOT experiment, Neoantigen in Top20 refers to the number of true positive tumor neoantigens in the top 20 tumor neoantigen peptides selected according to the score, the X-axis represents the prediction results in each patient, and the Y-axis represents the number of tumor neoantigens.

[0038] Figure 4: Statistical results of MHC-II tumor neoantigen peptide scoring model in predicting tumor neoantigens. Neoantigen refers to the number of true positive tumor neoantigens identified by ELISPOT experiment, Neoantigen in Top20 refers to the number of true positive tumor neoantigens in the top 20 tumor neoantigen peptides selected according to the score ranking, the X axis represents the prediction result in each patient, and the Y axis represents the number of tumor neoantigens.

[0039] Figure 5 : Statistical results when predicting MHC-I / II tumor neoantigen peptides simultaneously. Neoantigen refers to the number of true positive tumor neoantigens identified by ELISPOT experiment, Neoantigen in Top20 refers to the number of true positive tumor neoantigens in the top 20 tumor neoantigen peptides selected according to the score ranking, the X axis represents the prediction result in each patient, and the Y axis represents the number of tumor neoantigens.

[0040] Figure 6 : A block diagram schematically illustrates a computing device suitable for use in implementing embodiments of the present invention DETAILED DESCRIPTION

[0041] Before the present disclosure is described in detail, it is to be understood that certain terminology used herein is for the purpose of describing particular embodiments only. 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. Also, the singular forms of words herein are intended to include the plural forms, unless the context clearly indicates otherwise. Additionally, the use of “including,” “comprising,” “having,” “containing,” and variations thereof herein is meant to encompass the items listed thereafter, but does not exclude other like items not specifically listed. The use of “consisting essentially of’ means excluding other elements of any essential nature to the specified elements, steps or components. The use of “consisting of’ means excluding any element not specified. The use of “comprising” or “including” means that other elements not specifically listed are not excluded. The use of “or” means “and / or” unless otherwise indicated.

[0042] As used herein, the terms “comprises,” “comprising,” “includes,” “including,” “has,” “having,” “contains,” “containing,” “characterized by,” and variations thereof, are intended to mean that the subject matter includes the recited elements, but not to the exclusion of other elements. The term “consisting essentially of’ does not exclude other elements not specified, as long as the essential nature of the subject matter is not changed by the other elements. The term “consisting of’ excludes all but the specified elements. The term “substantially” means that the recited characteristic, parameter, or value need not be achieved exactly, but that deviations, within manufacturing tolerances and / or variations, are acceptable, as long as the essential nature of a structure, function, or result is not changed.

[0043] As used herein, “at least one” or “one or more” can mean 1, 2, 3, 4, 5, 6, 7, 8, or more.

[0044] As used herein, the term "and / or," when used between two or more elements, is understood to mean either one of the elements or any combination of the elements. In other words, "and / or" includes "and" as well as "or." For example, A and / or B includes A, B, and A+B. A, B, and / or C includes A, B, C, and any combination of the elements, such as A+B, A+C, B+C, and A+B+C. More elements defined with "and / or" are understood in a similar manner and include any one of the elements and any combination of the elements.

[0045] As used herein, "antigen" refers to a molecule that, upon entering the body, can elicit an acquired immune response of the body, which can involve the production of antibodies, or specific immunogenically active cells, or both. Those skilled in the art will appreciate that any macromolecule, including almost all proteins or peptides, can serve as an antigen. Further, the antigen can be derived from recombinant or genomic DNA or RNA.

[0046] As used herein, the term "wild type" indicates that the sequence is naturally occurring and has not been modified by man, including naturally occurring mutants.

[0047] As used herein, the term "neoantigen" is an antigen having at least one change that makes it different from the corresponding wild type parent antigen, for example, the change is a tumor cell mutation. The neoantigen can include a polypeptide sequence or a nucleotide sequence.

[0048] As used herein, the term "variation" is a difference between a nucleic acid of a subject and a reference human genome used as a control, which includes genetic mutations and genetic recombination. The mutation can include a point mutation caused by a single base change, or a deletion, duplication, and insertion of multiple bases.

[0049] As used herein, the term "tumor neoantigen" is a neoantigen present in a tumor cell or tissue of a subject but not present in a corresponding normal cell or tissue of the subject.

[0050] As used herein, the term "tumor" generally refers to any new pathological growth of tissue. The tumor of the present application can be benign or malignant. The tumor of the present application can be a tumor of any stage, for example, very early stage tumor, early stage tumor, intermediate stage tumor, late stage tumor. The tumor of the present application includes solid tumors and blood tumors. The solid tumors include, but are not limited to, squamous cell carcinoma, adenocarcinoma, basal cell carcinoma, renal cell carcinoma, breast ductal carcinoma, soft tissue sarcoma, osteosarcoma, melanoma, small cell lung cancer, non-small cell lung cancer, lung adenocarcinoma, peritoneal cancer, hepatocellular carcinoma, gastrointestinal cancer, gastric cancer, pancreatic cancer, neuroendocrine cancer, glioblastoma, cervical cancer, ovarian cancer, liver cancer, bladder cancer, brain cancer, hepatoma, breast cancer, colon cancer, colorectal cancer, endometrial cancer or uterine cancer, esophageal cancer, salivary gland cancer, kidney cancer, liver cancer, prostate cancer, vulvar cancer, thyroid cancer, neuroblastoma, or head and neck cancer; the blood tumors include, but are not limited to, leukemia, lymphoma, myeloma, acute myelogenous leukemia, chronic myelogenous leukemia, acute lymphoblastic leukemia, chronic lymphoblastic leukemia, hairy cell leukemia, Hodgkin's lymphoma, non-Hodgkin's lymphoma, mantle cell lymphoma, or multiple myeloma.

[0051] As used herein, the term "tumor tissue" generally refers to a sample of any tissue from a tumor in a subject, including any tissue of any solid tumor and / or non-solid tumor in a subject.

[0052] As used herein, the term "normal tissue" refers to any tissue that is substantially healthy and does not show any significant symptoms or signs of disease (e.g., the tissue is not cancerous), and includes all normal somatic tissues. The normal tissue used in the present application can be from normal solid tissue (e.g., bone and bone marrow), soft tissue (e.g., organs, muscle, fat, and skin), or body fluid (e.g., blood, plasma, serum, urine, peritoneal fluid, cerebrospinal fluid, pleural fluid, and amniotic fluid). Preferably, the normal tissue used in the present application is from normal peripheral blood.

[0053] As used herein, the term "tumor purity" refers to the proportion of tumor cells in a sample relative to all cells.

[0054] As used herein, the term "hydrophobic" refers to a phenomenon or property of water-insoluble molecules in an aqueous solution to exclude water molecules around them and to aggregate together so as to minimize the contact between the water molecules and the water-insoluble molecules, thereby reducing the interaction between the water molecules and the water-insoluble molecules and, in turn, reducing the energy of the system to reach a relatively stable state.

[0055] As used herein, the term "hydrophobic value" refers to a measure of the hydrophobicity of a substance, with a larger value indicating a higher affinity of the substance to an organic phase and a stronger hydrophobicity.

[0056] As used herein, the term “MHC” refers to a group of tightly linked genes that determine whether transplanted tissues are identical, are closely associated with immune responses, and are mainly composed of MHC class I molecules and MHC class II molecules. “MHC class I molecules” refer to a dimer of MHC class I alpha chain and beta2 microglobulin chain, and “MHC class II molecules” refer to a dimer of MHC class II alpha chain and MHC class II beta chain. The MHC of humans is called the human leukocyte antigen (HLA) complex.

[0057] As used herein, the term “MHC-I type variant polypeptide fragment”, “MHC-I type tumor neoantigen peptide” refers to a variant polypeptide fragment or tumor neoantigen peptide that is specifically recognized by MHC class I molecules.

[0058] As used herein, the term “MHC-II type variant polypeptide fragment”, “MHC-II type tumor neoantigen peptide” refers to a variant polypeptide fragment or tumor neoantigen peptide that is specifically recognized by MHC class II molecules.

[0059] As used herein, the term “pMHC” refers to an MHC molecule (MHC class I or MHC class II) that includes a peptide bound in a MHC peptide binding pocket that is recognized in the art. In some cases, the MHC molecule can be a membrane-bound protein expressed on the surface of a cell. In other cases, the MHC molecule can be a soluble protein that lacks a transmembrane region or a cytosolic region.

[0060] As used herein, the term “pMHC affinity” refers to the binding affinity between a particular antigen and a particular MHC allele.

[0061] As used herein, the term “pMHC ligand elution value” reflects the likelihood of eluting a ligand after a particular antigen binds to a particular MHC allele.

[0062] As used herein, the term “antigen processing score” reflects a score of how well an antigen is processed during antigen processing.

[0063] As used herein, the term “antigen presentation score” reflects a score of how well an antigen is presented during antigen processing.

[0064] As used herein, the term “allele” refers generally to a pair of genes that are located at the same position on a pair of homologous chromosomes that control contrasting traits, which is a form of a gene, or a form of a genetic sequence, or a form of a protein.

[0065] As used herein, the term “reference allele” refers to a sequence of one or more nucleotides at a chromosomal locus that is either the major allele at that chromosomal locus in a population of the species (e.g., a “wild type” sequence), or a predefined allele in a reference genome of the species.

[0066] The term "variant allele" as used herein refers to one or more nucleotide sequences at a chromosomal locus that is not the major allele at that chromosomal locus in the population of the species (e.g., is not the "wild type" sequence) nor is it a predefined allele in a reference sequence construct for the species (e.g., a reference genome or set of reference genomes). In certain instances, sequence variants that are found in the population of the species that do not affect the change in the encoded protein, or that result in amino acid substitutions that do not substantially affect the function of the encoded protein, are not variant alleles.

[0067] The term "allele frequency" as used herein is the frequency of an allele (or variant of a gene) relative to all alleles of the gene, which can be expressed as a fraction or a percentage. Allele frequencies are often associated with a particular genomic locus because genes are typically located at one or more loci.

[0068] The term "variant allele frequency," "VAF" as used herein is the frequency of a variant or mutant allele relative to all alleles of the gene, which can be obtained by dividing the number of times a variant or mutant allele is observed (e.g., the number of reads supporting a candidate variant allele) by the total number of times the position is sequenced (e.g., the total number of reads covering the candidate site).

[0069] The term "somatic variant" or "somatic mutation" as used herein is a mutation that occurs in a non-germline cell of an individual. Somatic DNA mutations, including single nucleotide variants (SNVs) and insertions and deletions (INDELs), are a major source of aberrant antigens. Single Nucleotide Variants (SNVs) are variations in a single nucleotide in a DNA sequence. There are three common patterns of single nucleotide variants: substitution, deletion, and insertion. In the substitution pattern, a single nucleotide at a certain position in the genome is changed to another nucleotide; in the deletion pattern, a nucleotide at a certain position in the genome is deleted; and in the insertion pattern, a single nucleotide at a certain position in the genome is duplicated. Insertion and deletion (INDEL) refers to the insertion or deletion of a small fragment of sequence at a certain position in the genome, usually less than 50 bp in length. It usually refers to the insertion or deletion of DNA fragments of different sizes in the genome, and many of them occur in important positions such as intron regions, even exon regions, promoter regions, etc.

[0070] As used herein, the term "copy number variation", "SNV" refers to somatic copy number variation events that occur only in diseased cells and are defined as a change in the copy number of maternal and / or paternal alleles of a region of the diseased (tumor) genome relative to a matched normal genome, wherein the substitution preferably affects a genomic region spanning about 1 kb or longer.

[0071] As used herein, "genome" refers to the sum of all genetic material of an organism. The genetic material includes DNA or RNA.

[0072] As used herein, the term "exome" is a subset of the genome that encodes proteins. The exome can be the entire set of exons of the genome.

[0073] As used herein, the term "Whole Exome Sequencing (WES)" refers to a genomic analysis method that performs high-throughput sequencing of all exonic regions in the genome.

[0074] As used herein, the term "whole transcriptome sequencing" is a sequencing technology that can reveal the presence and quantity of RNA in a biological sample at a given moment, analyzing the constantly changing cellular transcriptome. Transcriptome broadly refers to the collection of all transcribed products in a cell under a certain physiological condition, including messenger RNA, ribosomal RNA, transfer RNA and non-coding RNA; narrowly refers to the collection of all mRNA transcripts. Transcript refers to one or more mature mRNA available for encoding proteins formed by transcription of a gene.

[0075] As used herein, the term "Next-generation sequencing (NGS)" is also known as high-throughput sequencing, which introduces reversible termination ends, thus realizing sequencing by synthesis, determining DNA sequence by capturing special markers carried by newly added bases during DNA replication.

[0076] As used herein, the term "third-generation sequencing" is also known as single molecule sequencing, which does not require PCR amplification during sequencing, realizing the individual sequencing of each DNA molecule. Representative third-generation sequencing technologies / platforms are, for example, SMRT technology of Pacific Biosciences and nanopore technology of Oxford Nanopore Technologies.

[0077] As used herein, the term "gene annotation" or "annotation" refers to high-throughput annotation of biological functions of all genes in a genome using bioinformatics methods and tools, such as searching and defining genomic elements on a DNA sequence by various methods from de novo, homology, structural definition, etc., to obtain information of their location, sequence, structure, function, etc.

[0078] As used herein, the term "Lasso" refers to least absolute shrinkage and selection operator algorithm, LASSO regression is also called Lasso regression, which is a compressed estimate. Lasso regression prevents overfitting and solves the problem of serious collinearity by performing L1 regularization, i.e., adding a penalty term equal to the absolute value of the coefficient magnitude, so that some coefficients become zero and are eliminated from the model. In other words, LASSO can be used to reduce the number of independent variables of a linear model.

[0079] As used herein, the term "cross validation" or "k-fold cross validation" is used to test the accuracy of an algorithm. The data set is divided into k parts, and k-1 parts are used as training data and 1 part is used as validation data in turn to test. Each test will give the corresponding accuracy (or error rate). The average of the accuracy (or error rate) of the results of k times is used as an estimate of the accuracy of the algorithm. Generally, multiple k-fold cross validations (such as k times k-fold cross validation) are required, and the average is taken as an estimate of the accuracy of the algorithm.

[0080] As used herein, the term "receiver operating characteristic curve" or "ROC" refers to a graphical curve that shows the performance of a binary classifier system as its discrimination threshold is varied. For any given method, the curve can be created by plotting the true positive rate against the false positive rate at various threshold settings. In addition, the ROC curve can determine the value or expected value of any unknown parameter as long as at least one of the three parameters (e.g., sensitivity, specificity, and threshold setting) is provided. The unknown parameter can be determined using a curve suitable for the ROC curve. The term "AUC" refers to the area under the receiver operating characteristic curve. An AUC close to 1 indicates that the test is highly sensitive and has high specificity, while an AUC close to 0.5 indicates that the test is neither sensitive nor specific.

[0081] As used herein, the term "subject" encompasses a cell, tissue or organism, human or non-human, whether in vivo, ex vivo, or in vitro, male or female. A subject includes a mammal, including a human. The term "mammal" encompasses both human and non-human, and includes, but is not limited to, humans, non-human primates, canids, felines, murines, bovids, equids, and suids. Preferably, the subject is a human.

[0082] The first aspect of the present application provides a method for constructing a tumor neoantigen peptide score model, comprising: obtaining whole exome sequencing data of tumor tissue and normal tissue, whole transcriptome sequencing data of tumor tissue and immunogenicity results of variant polypeptide fragments of a plurality of subjects; processing the sequencing data to obtain score features corresponding to each variant polypeptide fragment; screening the score features using a Lasso regression model to obtain score key features and their regression coefficients, and establishing a tumor neoantigen peptide score model.

[0083] In some embodiments, the normal tissue can be derived from all normal somatic tissues, preferably from normal peripheral blood.

[0084] In some embodiments, the sequencing data can be obtained by any suitable sequencing method, such as next-generation sequencing (NGS) technology, third-generation sequencing technology, etc., preferably NGS technology. Third-generation sequencing methods can replace NGS technology in the future to speed up the sequencing step of the present method.

[0085] In some embodiments, the immunogenicity results of the variant polypeptide fragments can be obtained by any suitable immunological experiment that can identify the immunogenicity of the variant polypeptide fragments. A non-limiting example is to test the immunogenicity of the variant polypeptide fragments by an enzyme-linked immunospot (ELISPOT) assay.

[0086] In some embodiments, before obtaining whole exome sequencing data of tumor tissue and normal tissue, whole transcriptome sequencing data of tumor tissue and immunogenicity results of variant polypeptide fragments of a plurality of subjects, the raw data of the sequencing data is also pre-processed to obtain sequencing data, which includes trimming sequencing adapter sequences, trimming sequencing tag sequences, removing sequences with poor sequencing quality, etc. The pre-processing can be performed by any suitable sequencing data quality control software. A non-limiting example includes using Trimmomatic software to pre-process the raw data of the sequencing data.

[0087] In some embodiments, wherein the processing the sequencing data, obtaining the scoring features corresponding to each variant polypeptide fragment comprises: aligning the whole exome sequencing data of the tumor tissue and normal tissue to a reference genome to obtain DNA alignment results; aligning the whole transcriptome sequencing data of the tumor tissue to the reference genome to obtain RNA alignment results; analyzing to obtain somatic mutation sites in the tumor tissue relative to the normal tissue according to the DNA alignment results, and predicting the HLA typing of the subject; calculating variant polypeptide fragment features according to variant polypeptide fragments with known immunogenicity results; obtaining expression features of somatic mutations corresponding to the variant polypeptide fragments according to the somatic mutation sites, the DNA alignment results and the RNA alignment results; predicting the binding features of variant polypeptide fragments and HLA molecules based on the variant polypeptide fragments and the HLA typing results.

[0088] In some embodiments, the aligning the whole exome sequencing data of the tumor tissue and normal tissue to a reference genome to obtain DNA alignment results, and the aligning the whole transcriptome sequencing data of the tumor tissue to the reference genome to obtain RNA alignment results can employ any suitable sequence alignment tool and alignment algorithm. In one non-limiting example, in the aligning the whole exome sequencing data of the tumor tissue and normal tissue to a reference genome to obtain DNA alignment results, the Burrows-Wheeler Transform algorithm is employed; in the aligning the whole transcriptome sequencing data of the tumor tissue to the reference genome to obtain RNA alignment results, the STAR algorithm is employed.

[0089] In some embodiments, the reference genome is a human reference genome, for example, the optional human reference genome hg38, hg19, etc.

[0090] In some embodiments, after obtaining the DNA alignment results and / or RNA alignment results, further comprising preprocessing the DNA alignment results and / or RNA alignment results, the preprocessing comprising deduplication, adding group information, and recalculating base quality, etc., the preprocessing can be performed by any suitable quality control software. One non-limiting example comprises employing GATK software to preprocess the DNA alignment results and / or RNA alignment results.

[0091] In some embodiments, the analyzing somatic mutation sites in tumor tissue relative to normal tissue according to DNA alignment results can employ any suitable somatic mutation analysis software. In a non-limiting example, the analyzing somatic mutation sites in tumor tissue relative to normal tissue according to DNA alignment results can specifically comprise: analyzing somatic mutation sites in tumor tissue using Mutect2 and / or Strelka and / or Varscan and / or lofreq and / or Muse and / or Scalpel and / or Sniper and / or Vardict and / or Somaticseq software according to exome sequencing alignment results of tumor tissue and normal tissue. Preferably, all of the above software are used for analysis respectively.

[0092] In some embodiments, the predicting HLA typing of the subject according to DNA alignment results can employ any suitable HLA typing prediction software. In a non-limiting example, the predicting HLA typing of the subject according to DNA alignment results can specifically comprise: analyzing the genotype of MHC class I molecules of the tumor tissue of the subject using polysolver software and / or HLA-VBSeq software and / or OptiType software and / or HLA-HD software and / or xHLA software, wherein preferably all of the above software are used for analysis respectively, and the genotype predicted by the most number of software simultaneously can improve the prediction accuracy; analyzing the genotype of MHC class II molecules of the tumor tissue of the subject using HLA-VBSeq software and / or HLA-HD software and / or xHLA software, wherein preferably all of the above software are used for analysis respectively, and the genotype predicted by the most number of software simultaneously can improve the prediction accuracy.

[0093] In some embodiments, the calculating features of the variant polypeptide fragments according to known immunogenicity results at least comprises obtaining the length of the variant polypeptide fragments according to the amino acid sequence of the variant polypeptide fragments, and calculating the hydrophobic value of the variant polypeptide fragments, wherein the hydrophobic value of the variant polypeptide fragments can be obtained by any suitable calculation method in the art. In a non-limiting example, the hydrophobic value of the variant polypeptide fragments is calculated based on the following rules:

[0094] Rule 1: The hydrophobic value of each amino acid at different positions in the sequence is shown in the following table:

[0095] Table 1: Amino acid position factor

[0096]

[0097]

[0098] Rule two: adjacent position factor correction. For strong hydrophobic amino acids Y, F, L, I, V, W, if there is a basic amino acid H, R and K in the front (N-terminal) adjacent position, the strong hydrophobic amino acid hydrophobic value is reduced by 0.05, 0.1, 0.3, 0.15, 0.2, 0.05 respectively. For the peptide chain containing RL peptide segment, the hydrophobic value of the peptide chain is reduced by 0.3 for each occurrence of the RL peptide segment.

[0099] Rule three: hydrophobic cluster factor correction. When the strong hydrophobic amino acid set Y (W, F, L, I and Y) and the weak hydrophobic amino acid set X (M, V and A) are arranged in the following combinations, the peptide chain hydrophobic value H is reduced to different degrees respectively.

[0100] Table 2: Hydrophobic cluster factor correction

[0101]

[0102]

[0103] Rule four: hydrophobic value H segment correction. The final hydrophobic value of the peptide chain is corrected according to the following rules:

[0104]

[0105] In some embodiments, the expression characteristics of the somatic variation corresponding to the variant polypeptide fragment are obtained according to the somatic variation site, DNA alignment result and RNA alignment result, at least comprising: obtaining the frequency of the variant allele at the DNA level in the tumor tissue sample according to the DNA alignment result; analyzing the somatic genomic copy number variation of the tumor tissue and calculating the tumor purity; obtaining the total number of sequencing fragments covering the variation site in the transcriptome sequencing, the number of sequencing fragments of the variant allele in the transcriptome sequencing according to the RNA alignment result, and calculating the frequency of the variant allele at the RNA level and the expression amount of the gene where the variation is located; and calculating the corrected frequency of the variant allele at the DNA level according to the frequency of the variant allele at the DNA level and the tumor purity.

[0106] In some embodiments, the frequency of the variant allele at the DNA level in the tumor tissue sample is obtained by calculating the proportion of the number of sequencing fragments of the variant allele in the exome sequencing in the total number of sequencing fragments covering the variation site in the exome sequencing, and the number of sequencing fragments of the variant allele in the exome sequencing and the total number of sequencing fragments covering the variation site in the exome sequencing can be obtained by using any suitable bam data extraction software, such as but not limited to bam-readcount software.

[0107] In some embodiments, the analyzing somatic genomic copy number variations of the tumor tissue can be performed using any suitable software for detecting genomic copy number variations. In a non-limiting example, the analyzing somatic genomic copy number variations of the tumor tissue is performed using cnvkit software.

[0108] In some embodiments, the tumor purity is calculated based on somatic variations and somatic genomic copy number variations of the tumor tissue, and any suitable software for analyzing tumor purity can be used to calculate the tumor purity. In a non-limiting example, the tumor purity is calculated using ABSOLUTE software.

[0109] In some embodiments, the obtaining the total number of sequencing reads covering the variant site in the transcriptome sequencing and the number of sequencing reads of the variant allele in the transcriptome sequencing based on the RNA alignment results can be performed using any suitable bam data extraction software. In a non-limiting example, the bam-readcount software is used to obtain the total number of sequencing reads covering the variant site in the transcriptome sequencing and the number of sequencing reads of the variant allele in the transcriptome sequencing. In some embodiments, the variant allele frequency at the RNA level is obtained by calculating the proportion of the number of sequencing reads of the variant allele in the transcriptome sequencing in the total number of sequencing reads covering the variant site in the transcriptome sequencing. In some embodiments, the expression level of the gene in which the variant is located can be calculated using any suitable gene expression analysis software. In a non-limiting example, the expression level of the gene in which the variant is located in the tumor tissue is calculated using RSEM software, and the expression level of the gene in which the variant is located can be represented in any suitable manner, which can be selected by a person skilled in the art according to actual needs, such as RPKM (reads per kilobase per million), FPKM (fragments per kilobase per million) or TPM (transcripts per kilobase per million), and FPKM is preferably used.

[0110] In some embodiments, the corrected variant allele frequency at the DNA level is corrected by the tumor purity and the variant allele frequency at the DNA level, so as to obtain the corrected variant allele frequency at the DNA level in the tumor cell. In a non-limiting example, the corrected variant allele frequency at the DNA level is obtained by calculating the ratio of the variant allele frequency at the DNA level to the tumor purity (if the tumor purity value is greater than 1, it is calculated as 1).

[0111] In some embodiments, the predicting the binding characteristics of the variant polypeptide fragment and the HLA molecule based on the variant polypeptide fragment and the HLA typing result at least includes predicting the pMHC affinity value and the pMHC ligand elution value using any suitable MHC affinity prediction software.

[0112] In some embodiments, for the MHC-I type variant polypeptide fragments, the binding characteristics of the variant polypeptide fragments to HLA molecules further at least include antigen processing scores and antigen presentation scores.

[0113] In a non-limiting example, based on the variant polypeptide fragments and HLA typing results, the predicted binding characteristics of the variant polypeptide fragments to HLA molecules at least include calculating the pMHC affinity values, pMHC ligand elution values and antigen processing scores and antigen presentation scores of the MHC-I type variant polypeptide fragments using netMHCpan and MHCflurry software, respectively; calculating the pMHC affinity values, pMHC ligand elution values of the MHC-II type variant polypeptide fragments using netMHCIIpan software.

[0114] In some embodiments, since the variant polypeptide fragments that cannot effectively bind to MHC molecules will greatly interfere with the modeling process, before using the Lasso regression model to screen the score characteristics, it further includes preliminarily filtering the variant polypeptide fragments according to the binding characteristics of the variant polypeptide fragments to HLA molecules, and the preliminary filtering standard is: for MHC-I type variant polypeptide fragments, the pMHC affinity value predicted by MHC-I type affinity prediction software is not higher than 500nM; for MHC-II type variant polypeptide fragments, the pMHC affinity value predicted by MHC-II type affinity prediction software is not higher than 200nM.

[0115] In a non-limiting example, the preliminary filtering standard is: for MHC-I type variant polypeptide fragments, the pMHC affinity value predicted by netMHCpan is not higher than 500nM; for MHC-II type variant polypeptide fragments, the pMHC affinity value predicted by netMHCIIpan is not higher than 200nM.

[0116] In some embodiments, considering that there are great differences between the data obtained by MHC-I class molecules and MHC-II class molecules, when using the Lasso regression model to screen the score characteristics, obtain the key score characteristics and their regression coefficients, and establish the tumor neoantigen peptide scoring model, the same modeling method is used to establish the tumor neoantigen peptide scoring model for MHC-I type variant polypeptide fragments and MHC-II type variant polypeptide fragments, respectively.

[0117] In some embodiments, before using the Lasso regression model to screen the score characteristics, it further includes normalizing the values of the score characteristics, which can use any applicable normalization method, such as Z-score normalization, logarithmic normalization, etc., and preferably uses Z-score normalization processing, so that all score characteristic values satisfy the (0, 1) standard normal distribution.

[0118] In some embodiments, the Z-score normalization formula is:

[0119]

[0120] wherein Value * is the normalized score feature value, value is the original score feature value, μ is the estimated population mean of the score feature, and σ is the estimated population standard deviation of the score feature.

[0121] In one specific example, the estimated population mean and population standard deviation of each score feature are shown in the following table:

[0122] Table 3: Estimated population mean and population standard deviation of each score feature for constructing MHC-I type tumor neoantigen peptide score model

[0123]

[0124]

[0125] Table 4: Estimated population mean and population standard deviation of each score feature for constructing MHC-II type tumor neoantigen peptide score model

[0126]

[0127] The estimated population mean and population standard deviation of each score feature described above are not intended to limit the specific parameters when calculating the Z-score. Those skilled in the art can understand that for different reference samples and data processing methods, the values of each score feature and its estimated population mean and estimated population standard deviation will have slight changes, but the overall trend will not change.

[0128] In some embodiments, wherein the Lasso regression model selects the optimal penalty parameter based on the cross-validation method, and determines the score key features and their regression coefficients for establishing the tumor neoantigen peptide score model according to the optimal penalty parameter, the optimal penalty parameter can be selected according to actual needs, for example, the penalty parameter when the regression model reaches the minimum mean square error or the penalty parameter corresponding to the most slowly rising average area under the curve of the receiver operating characteristic curve, preferably the penalty parameter corresponding to the most slowly rising average area under the curve of the receiver operating characteristic curve as the optimal penalty parameter, at this time the model ensures under-fitting and avoids over-fitting.

[0129] In some embodiments, wherein the tumor neoantigen peptide score model is expressed in the form of prediction probability:

[0130]

[0131] wherein,

[0132] MScore =∑(βi×Xi)+β0

[0133] Xi is the value of the normalized score key feature, βi is its regression coefficient, and β0 is the intercept term.

[0134] In some embodiments, the score key features of the MHC-I tumor neoantigen peptide scoring model after dimensionality reduction by Lasso regression model include variant allele frequency at RNA level, pMHC affinity value, pMHC ligand elution value, antigen processing score, and antigen presentation score; and the score key features of the MHC-II tumor neoantigen peptide scoring model include variant polypeptide fragment length, variant allele frequency at RNA level, pMHC affinity value, and pMHC ligand elution value.

[0135] The second aspect of the present application provides a tumor neoantigen peptide scoring model, which is constructed by the method of the first aspect of the present application.

[0136] In a specific embodiment, the scoring model for evaluating MHC-I variant polypeptide fragments is as follows:

[0137] MScore = 0.3934947 × RNA_VAF * -0.314251084 × netMHCpan_ic50_neo * +0.475021684 × netMHCpan_el_score_neo * +0.017763407 × MHCflurry_processing_score_neo * +0.488476145 × MHCflurry_presentatin_score_neo * -2.3431054

[0138]

[0139] The scoring model for evaluating MHC-II variant polypeptide fragments is as follows:

[0140] MScore = 0.297529845 × Length * +0.165504699 × RNA_VAF * -0.449870092 × netMHCpan_ic50_neo * +0.43987096 × netMHCpan_el_score_neo * -7.79007716

[0141]

[0142] In a specific embodiment, in order to facilitate use, the present application also provides a tumor neoantigen peptide scoring model comprising a normalization step, which can be used to calculate the tumor neoantigen prediction probability using the original scoring feature value, at this time the MHC-I type tumor neoantigen peptide scoring model is as follows:

[0143]

[0144] The MHC-II type tumor neoantigen peptide scoring model is as follows:

[0145]

[0146] The third aspect of the present application provides the use of the tumor neoantigen peptide scoring model of the second aspect of the present application in screening tumor neoantigen peptides.

[0147] The fourth aspect of the present application provides a method for screening tumor neoantigen peptides, the method comprising: obtaining whole exome sequencing data of tumor tissue and normal tissue of a subject to be tested, whole transcriptome sequencing data of the tumor tissue; processing the sequencing data to obtain variant polypeptide fragments, and extracting scoring features corresponding to each variant polypeptide fragment; inputting the scoring features corresponding to each variant polypeptide fragment as input data into the scoring model constructed according to the method of the first aspect of the present application or the scoring model of the second aspect of the present application, calculating the prediction probability of each variant polypeptide fragment constituting a tumor neoantigen peptide; according to the prediction probability from high to low, selecting the variant polypeptide fragment with high prediction probability of constituting a tumor neoantigen peptide as a predicted tumor neoantigen peptide.

[0148] In some embodiments, wherein the processing of the sequencing data, obtaining the variant polypeptide fragments, and extracting the scoring features corresponding to each variant polypeptide fragment comprise: aligning the whole exome sequencing data of the tumor tissue and the normal tissue with the reference genome to obtain DNA alignment results; aligning the whole transcriptome sequencing data of the tumor tissue with the reference genome to obtain RNA alignment results; analyzing the somatic mutation sites in the tumor tissue relative to the normal tissue according to the DNA alignment results, and predicting the HLA typing of the subject; obtaining the variant polypeptide fragments according to the somatic mutation sites in the tumor tissue relative to the normal tissue; predicting the binding of the variant polypeptide fragments to MHC-I class molecules or to MHC-II class molecules based on the variant polypeptide fragments and the HLA typing results; and extracting the scoring features in the MHC-I type tumor neoantigen peptide scoring model or the MHC-II type tumor neoantigen peptide scoring model according to the results of the binding of the variant polypeptide fragments to MHC-I class molecules or to MHC-II class molecules.

[0149] In some embodiments, the obtaining of the variant polypeptide fragments according to the somatic mutation sites in the tumor tissue relative to the normal tissue can be performed using any suitable software in the art. In a non-limiting example, the obtaining of the variant polypeptide fragments according to the somatic mutation sites in the tumor tissue relative to the normal tissue can be performed using the customProDB software.

[0150] In some embodiments, wherein the extracting of the score features in the MHC-I type tumor neoantigen peptide scoring model or the MHC-II type tumor neoantigen peptide scoring model according to the results of the binding of the variant polypeptide fragments to the MHC class I molecules or the binding of the variant polypeptide fragments to the MHC class II molecules at least comprises: for the variant polypeptide fragments binding to the MHC class I molecules, obtaining the variant allele frequency at the RNA level according to the somatic mutation sites and the RNA alignment results, and obtaining the pMHC affinity value, the pMHC ligand elution value, the antigen processing score and the antigen presentation score according to the variant polypeptide fragments and the HLA typing results; for the variant polypeptide fragments binding to the MHC class II molecules, obtaining the variant polypeptide fragment length according to the variant polypeptide fragments, obtaining the variant allele frequency at the RNA level according to the somatic mutation sites and the RNA alignment results, and obtaining the pMHC affinity value and the pMHC ligand elution value according to the variant polypeptide fragments and the HLA typing results.

[0151] In the present application, the method for constructing a tumor neoantigen peptide scoring model of the present application can be presented in a systematic manner, and the system is configured to implement the method for constructing a tumor neoantigen peptide scoring model of the present application. The tumor neoantigen peptide scoring model construction system can be used to construct a tumor neoantigen peptide scoring model. For example, the tumor neoantigen peptide scoring model construction system can include a data acquisition module, a data processing module and a scoring model training module. The data acquisition module is used to acquire whole exome sequencing data of tumor tissues and normal tissues, whole transcriptome sequencing data of tumor tissues and immunogenicity results of variant polypeptide fragments of a plurality of subjects; the data processing module is used to process the sequencing data to obtain score features corresponding to each variant polypeptide fragment; the scoring model training module is used to screen the score features using a Lasso regression model to obtain score key features and their regression coefficients, and to establish a tumor neoantigen peptide scoring model.

[0152] In the present application, the screening method of the tumor neoantigen peptide of the present application can be presented in a systematic manner, and the system is configured to implement the screening method of the tumor neoantigen peptide of the present application. For example, the screening system of the tumor neoantigen peptide can include a data acquisition module, a data processing module, a scoring module and an output module. The data acquisition module is used to obtain whole exome sequencing data of tumor tissue and normal tissue of a test subject, whole transcriptome sequencing data of the tumor tissue; the data processing module is used to process the sequencing data, obtain variant polypeptide fragments, and extract scoring features corresponding to each variant polypeptide fragment; the scoring module is used to input the scoring features corresponding to each variant polypeptide fragment as input data into the scoring model constructed according to the method of the first aspect of the present application or the scoring model of the second aspect of the present application, and calculate the prediction probability of each variant polypeptide fragment constituting a tumor neoantigen peptide; the output module is used to sort the prediction probability from high to low, and select the variant polypeptide fragment with high prediction probability of constituting a tumor neoantigen peptide as a predicted tumor neoantigen peptide.

[0153] In some embodiments, the screening system of the tumor neoantigen peptide of the present application can further include a sequencing module configured to perform whole exome sequencing of tumor tissue, whole exome sequencing of normal tissue, and / or whole transcriptome sequencing of the tumor tissue on the test subject.

[0154] In some embodiments, the tumor neoantigen peptide scoring model construction system and / or the screening system of the tumor neoantigen peptide of the present application can further include a sequencing data preprocessing module configured to preprocess whole exome raw sequencing data of tumor tissue and normal tissue and whole transcriptome raw sequencing data of the tumor tissue, and the preprocessing of the raw sequencing data includes trimming sequencing adapter sequences, trimming sequencing tag sequences, removing sequences with poor sequencing quality, etc.; and / or an alignment result preprocessing module configured to preprocess DNA alignment results and RNA alignment results, and the preprocessing of the alignment results includes deduplication, adding group information and recalculating base quality, etc.

[0155] In the present application, the method of constructing a tumor neoantigen peptide scoring model or the screening method of a tumor neoantigen peptide of the present application can also be presented in the form of a computing device, which includes a memory configured to store one or more programs, and a processing unit coupled to the memory and configured to execute the one or more programs to cause the device to perform the method of constructing a tumor neoantigen peptide scoring model or the screening method of a tumor neoantigen peptide of the present application.

[0156] Figure 6A schematic diagram of a computing device 500 suitable for implementing embodiments of the present disclosure is shown. The device 500 can be a device for implementing a method of the first or third aspects of the present application. As shown, the device 500 comprises a central processing unit (CPU) 501 which can perform various suitable actions and processes in accordance with computer program instructions stored in a read only memory (ROM) 502 or computer program instructions loaded from a storage unit 508 into a random access memory (RAM) 503. Various programs and data used by the device 500, in operation, can also be stored within the RAM 503. The CPU 501, ROM 502 and RAM 503 are connected to each other by a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0157] Various components of the device 500 are connected to the I / O interface 505, including an input unit 506, an output unit 507, a storage unit 508, and the processing unit 501 performs various methods and processes described above, for example, a method of the first or third aspects of the present application. For example, in some embodiments, a method of the first or third aspects of the present application can be implemented as a computer software program stored on a machine readable medium, such as the storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded onto the RAM 503 and executed by the CPU 501, one or more operations of a method of the first or third aspects described above can be performed. Alternatively, in other embodiments, the CPU 501 can be configured to perform one or more actions of a method of the first or third aspects of the present application by any other suitable means, for example, by means of firmware.

[0158] Furthermore, and alternatively, the present disclosure can be a method, an apparatus, a system, and / or a computer program product. The computer program product can include a computer readable storage medium (or media) having stored thereon instructions that, when executed by a machine, cause the machine to perform the methods of the present application of constructing a tumor neoantigen peptide scoring model or the method of screening tumor neoantigens.

[0159] A computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0160] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0161] Computer readable program instructions for carrying out operations of the present disclosure can be in assembly code, instruction set architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform various aspects of the present disclosure.

[0162] Generally, the various example embodiments of the present disclosure can be implemented in hardware or special-purpose circuits, software, firmware, logic, or any combination thereof. Some aspects can be implemented in hardware, while other aspects can be implemented in firmware or software which can be executed by a controller, microprocessor or other computing device.

[0163] The computer readable program instructions or computer program product for executing various embodiments of the present disclosure can also be stored in the cloud, and when needed, the user can access the computer readable program instructions stored in the cloud through mobile Internet, fixed network or other network, so as to implement the technical solutions disclosed in the various embodiments of the present disclosure.

[0164] Examples

[0165] The example adopts the above method to establish a tumor neoantigen peptide scoring model, and uses the model to predict the tumor neoantigen peptide of the test subject, and the specific process is as follows:

[0166] 1. Collect whole exome and whole transcriptome sequencing data from the subject, and use Trimmomatic software to pretreat the original sequencing data, which includes trimming sequencing adapter sequences, trimming sequencing tag sequences, and removing sequences with poor sequencing quality, etc. The pretreated sequencing data is used for subsequent alignment.

[0167] 2. Using BWA software (whole exome) and STAR software (whole transcriptome) to align the pre-processed sequencing data of the subjects to the selected reference genome to obtain DNA alignment and RNA alignment results; and using GATK software to quality control the alignment results, which includes deduplication, adding group information, and recalculating base quality, etc. The quality controlled sequencing data is used for subsequent tumor cell variant analysis.

[0168] 3. Analyzing the whole exome sequencing data of the tumor tissue and normal tissue of the subject and the whole transcriptome data of the tumor tissue, using Mutect2, Strelka, Varscan, lofreq, Muse, Scalpel, Sniper, Vardict, Somaticseq software to obtain somatic variant sites of the tumor tissue, using customProDB software to translate nucleotide variants into variant polypeptide fragments containing amino acid variants, and further analyzing the sequencing data to obtain at least the following scoring features:

[0169] (1) the length of the variant polypeptide fragment (length);

[0170] (2) the hydrophobic value of the variant polypeptide fragment (hy_value);

[0171] (3) the allele frequency of the somatic variant site at the DNA level (DNA_VAF), the number of sequencing fragments containing the somatic variant site allele in exome sequencing and the total number of sequencing fragments covering the somatic variant site in exome sequencing are obtained by bam-readcount software, and the proportion of the number of sequencing fragments containing the somatic variant site allele in exome sequencing in the total number of sequencing fragments covering the somatic variant site in exome sequencing is calculated to obtain;

[0172] (4) the corrected allele frequency of the somatic variant site at the DNA level (adjust_vaf), which is obtained by calculating the ratio of the variant allele frequency at the DNA level to the tumor purity (if the tumor purity value is greater than 1, it is calculated as 1);

[0173] (5) the number of sequencing fragments containing the somatic variant site allele in transcriptome sequencing (RNA_AD), which is obtained by bam-readcount software;

[0174] (6) the total number of sequencing fragments covering the somatic variant site in transcriptome sequencing (RNA_Depth), which is obtained by bam-readcount software;

[0175] (7) Allele frequency of somatic mutation site at RNA level (RNA_VAF), obtained by calculating the proportion of sequencing fragments containing allele of somatic mutation site (RNA_AD) in total sequencing fragments covering somatic mutation site in transcriptome sequencing (RNA_Depth);

[0176] (8) Expression level of gene containing somatic mutation, represented by FPKM, obtained by gene expression analysis software RSEM;

[0177] (9) Purity value of tumor (purity), obtained by purity analysis software ABSOLUTE.

[0178] 4. Based on whole exome sequencing data of normal tissue from the source of the subject, analyze the genotype of MHC class I molecules using Polysolver, OptiType, xHLA, HLA-VBSeq and HLA-HD software, and analyze the genotype of MHC class II molecules using HLA-VBSeq, HLA-HD and xHLA.

[0179] 5. Based on the results of variant polypeptide fragments and HLA typing, analyze the binding characteristics of mutant polypeptides and HLA molecules, specifically including at least the following scoring characteristics:

[0180] (1) Affinity value of binding between variant polypeptide fragments and HLA molecules predicted by netMHCpan_ba (netMHCpan_ic50_neo), obtained by netMHCpan_ba (MHC class I molecules) or netMHCIIpan_ba (MHC class II molecules) software;

[0181] (2) Elution value of binding between variant polypeptide fragments and HLA molecules predicted by netMHCpan_el (netMHCpan_el_score_neo), obtained by netMHCpan_el (MHC class I molecules) or netMHCIIpan_el (MHC class II molecules) software;

[0182] (3) Affinity value of binding between variant polypeptide fragments and HLA molecules predicted by MHCflurry (MHCflurry_affinity_neo), obtained by MHCflurry software;

[0183] (4) Score of correct processing of variant polypeptide fragments predicted by MHCflurry (MHCflurry_processing_score_neo), obtained by MHCflurry software;

[0184] (5) The score of the predicted neoantigenic peptide fragment that is effectively presented (MHCflurry_presentation_score_neo) is obtained by MHCflurry software.

[0185] 6. Preliminary filtering of the neoantigenic peptide fragments, only the neoantigenic peptide fragments that can bind to HLA molecules are retained, because most of the predicted neoantigenic peptide fragments cannot effectively bind to HLA molecules and will cause great interference to the modeling process. The definition of the neoantigenic peptide fragments that can bind to HLA molecules is as follows:

[0186] (1) For MHC-I type neoantigenic peptide fragments, the affinity value between the MHC-I class molecules and the neoantigenic peptide fragments predicted by the netMHCpan method is required to be not higher than 500nM;

[0187] (2) For MHC-II type neoantigenic peptide fragments, the affinity value between the MHC-II class molecules and the neoantigenic peptide fragments predicted by the netMHCIIpan method is required to be not higher than 200nM.

[0188] 7. The score model is constructed for MHC-I type neoantigenic peptide fragments and MHC-II type neoantigenic peptide fragments respectively, which specifically includes the following steps:

[0189] (1) The score features obtained in steps 3 and 5 are taken as the original parameter set, and the score features are subjected to Z-score normalization processing according to the following formula, and the estimated overall mean μ and the estimated overall standard deviation σ of each score feature are shown in the following table, to obtain the normalized score parameter Value * ;

[0190]

[0191] Table 5: The estimated overall mean and overall standard deviation of each score feature for constructing the MHC-I type tumor neoantigen peptide score model

[0192]

[0193] Table 6: The estimated overall mean and overall standard deviation of each score feature for constructing the MHC-II type tumor neoantigen peptide score model

[0194]

[0195] (2) All patient data is divided into five groups, four of which are training sets and one of which is a test set;

[0196] (3) Using the training set to fit the Lasso regression model, the intercept scaling is set to 1000, and the model is trained by 5-fold cross-validation and the best penalty parameter (also called regularization parameter, C) is found. The C value is arranged in a gradient from 0.00001 to 10;

[0197] (4) The model corresponding to the average value of the area under the curve (Area Under Curve, AUC) of the receiver operating characteristic (Receiver Operating Characteristic, ROC) curve of the subjects in the validation set rises most slowly is the best model;

[0198] (5) According to the trained best model, the coefficients of all scoring features are obtained to obtain the tumor neoantigen peptide scoring model.

[0199] According to the above steps, the scoring model is constructed for MHC-I type variant polypeptide fragments and MHC-II type variant polypeptide fragments, respectively. The AUC graph for screening the best MHC-I type tumor neoantigen peptide scoring model and the MHC-II type tumor neoantigen peptide scoring model is shown in Figure 1 and Figure 2 The selected best model is selected in the green box.

[0200] According to the trained best model, the MHC-I type tumor neoantigen peptide scoring model is as follows:

[0201] MScore = 0.3934947 x RNA_VAF * -0.314251084 x netMHCpan_ic50_neo * +0.475021684 x netMHCpan_el_score_neo * +0.017763407 x MHCflurry_processing_score_neo * +0.488476145 x MHCflurry_presentation_score_neo * -2.3431054

[0202]

[0203] The MHC-II type tumor neoantigen peptide scoring model is as follows:

[0204] MScore = 0.297529845 x Length * +0.165504699 x RNA_VAF * -0.449870092 x netMHCpan_ic50_neo* + 0.43987096 x netMHCpan_el_score_neo * - 7.79007716

[0205]

[0206] 8. After modeling is completed, the performance of the model is evaluated, and the specific steps are as follows:

[0207] (1) Score each variant polypeptide fragment of a specific subject in the test set by the scoring model, and sort them, and take the top 20 variant polypeptide fragments as tumor neoantigen peptides;

[0208] (2) In the top 20 tumor neoantigen peptides, the proportion of true positive tumor neoantigen peptides identified by the immune enzyme-linked spot test (ELISPOT) in the test set subjects is included in the predicted top 20 tumor neoantigen peptides to evaluate the accuracy of the model. The ELISPOT test is performed by a method well known to those skilled in the art.

[0209] The model accuracy evaluation results are as follows:

[0210] For MHC-I type tumor neoantigen peptides, the statistical result is plotted as shown in Figure 3 The proportion of true positive tumor neoantigen peptides identified by ELISPOT experiment included in the top 20 tumor neoantigen peptides is 30.8% to 100%, with an average of 79.2%.

[0211] For MHC-II type tumor neoantigen peptides, the statistical result is plotted as shown in Figure 4 The proportion of true positive tumor neoantigen peptides identified by ELISPOT experiment included in the top 20 tumor neoantigen peptides is 75% to 100%, with an average of 98%.

[0212] Since MHC-I type variant polypeptide fragments and MHC-II type variant polypeptide fragments are scored simultaneously in actual operation, the predicted scores of MHC-I type variant polypeptide fragments and MHC-II type variant polypeptide fragments are combined and sorted in actual operation, and the top 20 variant polypeptide fragments are selected as tumor neoantigen peptides, and the proportion of true positive tumor neoantigen peptides identified by ELISPOT experiment included therein is evaluated to evaluate the accuracy of the model. The results are shown in Figure 5 The proportion of true positive tumor neoantigen peptides identified by ELISPOT experiment included in the top 20 tumor neoantigen peptides is 30.8% to 88.9%, with an average of 67.1%.

[0213] The foregoing described embodiments of the application are merely descriptive and are not exhaustive or limited to the embodiments disclosed. Many modifications and variations of the described embodiments are possible in light of the above teachings. The choice of words in this document is intended to express the principle of the embodiments best and to explain the practical application or technical improvement in the art, or to enable other ordinary skilled persons in the art to understand the embodiments disclosed herein.

[0214] The above are only optional embodiments of the present application, and are not intended to limit the disclosure. The present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for constructing a tumor neoantigen peptide scoring model, the method comprising: We obtained whole-exome sequencing data of tumor and normal tissues from multiple subjects, whole-transcriptome sequencing data of tumor tissues, and immunogenicity results of variant peptide fragments. The sequencing data is processed to obtain the scoring features corresponding to each variant polypeptide fragment; The Lasso regression model was used to screen the scoring features, obtain the key scoring features and their regression coefficients, and establish a tumor neoantigen peptide scoring model.

2. The method according to claim 1, wherein the immunogenicity results of the variant polypeptide fragment are obtained by immunological experiments.

3. The method according to claim 1, wherein processing the sequencing data to obtain scoring features corresponding to each variant polypeptide fragment includes: The whole exome sequencing data of the tumor tissue and normal tissue were compared with the reference genome to obtain DNA alignment results. The whole transcriptome sequencing data of the tumor tissue was compared with a reference genome to obtain RNA alignment results; Based on the DNA alignment results, somatic mutation sites in tumor tissue relative to normal tissue are analyzed and the subject's HLA typing is predicted. Calculate the characteristics of the variant polypeptide fragments based on known immunogenicity results; Based on the somatic cell mutation sites, DNA alignment results, and RNA alignment results, the expression characteristics of somatic cell mutations corresponding to the mutant polypeptide fragments are obtained; Based on the mutant polypeptide fragments and HLA typing results, the binding characteristics of the mutant polypeptide fragments to HLA molecules are predicted.

4. The method according to claim 3, wherein the characteristics of the variant polypeptide fragment include at least the variant polypeptide fragment length and the variant polypeptide fragment hydrophobicity value.

5. The method according to claim 3, wherein the expression characteristics of the somatic variant corresponding to the variant polypeptide fragment include at least the frequency of variant alleles at the DNA level, the corrected frequency of variant alleles at the DNA level, the total number of sequencing fragments covering the variant site in transcriptome sequencing, the number of sequencing fragments of variant alleles in transcriptome sequencing, the frequency of variant alleles at the RNA level, the expression level of the gene containing the variant, and tumor purity.

6. The method according to claim 3, wherein the binding characteristics of the variant polypeptide fragment to the HLA molecule include at least the pMHC affinity value and the pMHC ligand elution value.

7. The method according to claim 6, wherein for MHC-I type variant polypeptide fragments, the binding characteristics of the variant polypeptide fragments to HLA molecules further include at least an antigen processing score and an antigen presentation score.

8. The method according to claim 1, wherein before using the Lasso regression model to screen the scoring features, it further includes preliminary filtering of the variant polypeptide fragments based on the binding characteristics of the variant polypeptide fragments to HLA molecules, wherein the preliminary filtering criteria are: For MHC-I variant polypeptide fragments, the pMHC affinity value predicted by the MHC-I affinity prediction software is no higher than 500 nM; For MHC-II variant polypeptide fragments, the pMHC affinity value predicted by the MHC-II affinity prediction software is no higher than 200 nM.

9. The method according to claim 1, wherein the Lasso regression model is used to screen the scoring features to obtain key scoring features and their regression coefficients, and in establishing the tumor neoantigen peptide scoring model, tumor neoantigen peptide scoring models are established for MHC-I type variant polypeptide fragments and MHC-II type variant polypeptide fragments respectively.

10. The method of claim 1, wherein before using a Lasso regression model to screen the rating features, the method further includes normalizing the values ​​of the rating features.

11. The method of claim 1, wherein the Lasso regression model selects the optimal penalty parameter based on cross-validation, thereby determining the key scoring features and their regression coefficients for establishing the tumor neoantigen peptide scoring model.

12. The method according to claim 11, wherein the Lasso regression model selects the penalty parameter corresponding to the slowest increase in the average area under the curve as the optimal penalty parameter.

13. The method of claim 1, wherein the tumor neoantigen peptide scoring model is expressed as a predicted probability: in, MScore=∑(βi×Xi)+β0 Xi is the value of the normalized key feature of the score, βi is its regression coefficient, and β0 is the intercept term.

14. The method according to claim 13, wherein the key scoring features of the MHC-I type tumor neoantigen peptide scoring model include the frequency of variant alleles at the RNA level, pMHC affinity value, pMHC ligand elution value, antigen treatment score, and antigen presentation score; and the key scoring features of the MHC-II type tumor neoantigen peptide scoring model include the length of variant polypeptide fragments, the frequency of variant alleles at the RNA level, pMHC affinity value, and pMHC ligand elution value.

15. A tumor neoantigen peptide scoring model, wherein the model is constructed using the method described in any one of claims 1-14.

16. The application of the tumor neoantigen peptide scoring model of claim 15 in screening tumor neoantigen peptides.

17. A method for screening tumor neoantigen peptides, the method comprising: Obtain whole exome sequencing data of tumor tissue and normal tissue of the subject to be tested, and whole transcriptome sequencing data of tumor tissue; The sequencing data is processed to obtain variant polypeptide fragments, and scoring features corresponding to each variant polypeptide fragment are extracted. The scoring features corresponding to each variant polypeptide fragment are used as input data and input into the scoring model constructed according to any one of claims 1-14 or the scoring model according to claim 15 to calculate the predicted probability that each variant polypeptide fragment constitutes a tumor neoantigen peptide. Based on the predicted probabilities sorted from high to low, the variant polypeptide fragments with high predicted probabilities that constitute the tumor neoantigen peptide are selected as the predicted tumor neoantigen peptides.

18. The method of claim 17, wherein processing the sequencing data to obtain variant polypeptide fragments and extracting scoring features corresponding to each variant polypeptide fragment includes: The whole exome sequencing data of the tumor tissue and normal tissue were compared with the reference genome to obtain DNA alignment results. The whole transcriptome sequencing data of the tumor tissue was compared with a reference genome to obtain RNA alignment results; Based on the DNA alignment results, somatic mutation sites in tumor tissue relative to normal tissue are analyzed and the subject's HLA typing is predicted. Based on the somatic cell mutation sites in tumor tissue relative to normal tissue, mutated polypeptide fragments are obtained; Based on the mutant polypeptide fragment and HLA typing results, it is predicted whether the mutant polypeptide fragment will bind to MHC-I or MHC-II molecules. Based on the binding of the variant polypeptide fragment to MHC-I or MHC-II molecules, the scoring features in the corresponding MHC-I or MHC-II tumor neoantigen peptide scoring models are extracted.

19. The method according to claim 18, wherein extracting the scoring features from the corresponding MHC-I tumor neoantigen peptide scoring model or MHC-II tumor neoantigen peptide scoring model based on the result of the binding of the variant polypeptide fragment to MHC-I molecules or to MHC-II molecules includes: For MHC-I type variant polypeptide fragments, the extraction scoring features include at least obtaining the frequency of variant alleles at the RNA level based on the somatic variant sites and RNA alignment results, and obtaining pMHC affinity value, pMHC ligand elution value, antigen treatment score and antigen presentation score based on the variant polypeptide fragment and HLA typing results. For MHC-II variant polypeptide fragments, the extraction scoring features include at least obtaining the length of the variant polypeptide fragment based on the variant polypeptide fragment, obtaining the frequency of variant alleles at the RNA level based on the somatic mutation sites and RNA alignment results, and obtaining the pMHC affinity value and pMHC ligand elution value based on the variant polypeptide fragment and HLA typing results.

20. A tumor neoantigen peptide scoring model construction system, the system comprising the following modules: The data acquisition module is used to acquire whole exome sequencing data of tumor tissue and normal tissue from multiple subjects, whole transcriptome sequencing data of tumor tissue, and immunogenicity results of variant polypeptide fragments. The data processing module is used to process the sequencing data and obtain the scoring features corresponding to each variant polypeptide fragment; The scoring model training module is used to filter the scoring features using the Lasso regression model, obtain key scoring features and their regression coefficients, and establish a tumor neoantigen peptide scoring model.

21. A screening system for tumor neoantigen peptides, the system comprising the following modules: The data acquisition module is used to obtain whole exome sequencing data of tumor tissue and normal tissue of the subject to be tested, and whole transcriptome sequencing data of tumor tissue; The data processing module is used to process the sequencing data, obtain variant polypeptide fragments, and extract scoring features corresponding to each variant polypeptide fragment. The scoring module is used to take the scoring features corresponding to each variant polypeptide fragment as input data, input the scoring model constructed according to the system of claim 20 or the scoring model of claim 15, and calculate the predicted probability that each variant polypeptide fragment constitutes a tumor neoantigen peptide. The output module is used to sort the predicted probabilities from high to low and select the variant polypeptide fragments with high predicted probabilities that constitute the tumor neoantigen peptide as the predicted tumor neoantigen peptide.

22. A computing device, comprising: Memory, configured to store one or more programs; as well as A processing unit, coupled to the memory and configured to execute one or more programs to cause the device to perform the steps of the method of any one of claims 1-14 or 17-19.

23. A computer-readable storage medium storing machine-executable instructions that, when executed, cause a machine to perform the steps of the method according to any one of claims 1-14 or 17-19.

Images