A method and apparatus for identifying transcriptionally-derived tumor neoantigens

Abstract

The application discloses a method and device for identifying tumor neoantigens of transcriptional origin, which comprises: aligning RNA-seq data of a patient, splicing transcripts and predicting transcript coding ability of the aligned results; obtaining and annotating splicing events of the transcripts; extracting and quantifying exons from the spliced transcripts; integrating tumor-specific transcripts generated by specific splicing events and exons, and predicting tumor neoantigens. The method improves the identification strategy of tumor-specific transcripts, increases the identification of single-exon transcripts, optimizes the process of identifying tumor-specific transcripts and predicting their translation into proteins, and can efficiently identify tumor-specific transcripts and their derived neoantigen peptides from RNA-seq data of a patient, thereby providing strong technical support for the development of tumor neoantigen vaccines.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of tumor immunotherapy, in particular to a method and device for identifying tumor neoantigens from aberrant transcription. BACKGROUND

[0002] In the field of tumor immunotherapy, neoantigens have attracted extensive attention as specific targets for immune responses. Tumor neoantigens are polypeptide fragments generated from tumor-specific mutations or aberrant transcription, which can be presented to the cell surface by major histocompatibility complex (MHC) molecules, specifically recognized and activated by T cell receptors (TCRs) to activate anti-tumor immune responses. Personalized neoantigen vaccines based on genetic mutations have shown great potential in clinical trials. However, in low-mutation-load tumors, the number of neoantigens is limited, which greatly limits their clinical application. For example, the median number of mutation neoantigens in brain cancer patients is only about 9, far lower than the number of 20 neoantigen strings required for conventional vaccines. In addition, only about 1.6% of somatic mutations can generate immunogenic neoantigens. These factors have significantly limited the clinical application of tumor neoantigen vaccines.

[0003] In recent years, research has found that antigens generated from aberrant transcription (such as tumor-specific transcripts) can effectively make up for the shortage of mutation-derived neoantigens. The number of variable splicing-derived neoantigens in a single tumor sample (≈1.7 per sample) far exceeds that of single nucleotide variants (SNVs) (≈0.6 per sample), providing sufficient candidate targets for vaccines. In addition, tumor neoantigens from transcription have significant cross-tumor type and patient sharing. For example, Darwin et al. found that GNAS new splicing events are widely expressed in glioma, mesothelioma, prostate cancer, and liver cancer, providing potential targets for developing universal vaccines. Existing tools such as ASJA, IRIS, and SNAF predict neoantigens by identifying splicing events. Among them, ASJA, based on STAR alignment and StringTie transcript assembly, can accurately identify known and unannotated splicing events. Li et al. used ASJA to identify splicing events in pan-cancer and constructed a database, RJunBase, providing users with rich data resources.

[0004] However, the existing technology system still faces key bottlenecks in identifying tumor-specific transcript-derived neoantigens. The first problem is how to accurately define tumor specificity. Existing analysis methods (such as IRIS) often combine RNA-seq data from about 30 normal tissues (from the GTEx database) into a whole reference set for differential analysis. However, normal tissues themselves also have specific transcripts, which can interfere with the identification of tumor specificity. For example, when a transcript is highly expressed in a specific normal tissue (such as brain tissue), its overall expression mean may be diluted by the low expression of other tissues. Therefore, the detection of this transcript in tumor samples may be mistakenly determined as "tumor-specific". In addition, existing analysis methods mainly focus on the identification of splicing events, completely ignoring single-exon variants. However, experiments have shown that single-exon transcripts (such as MSTRG.28868.1) can also produce immunogenic polypeptides that specifically bind to HLA:32:01. The presence of cancer-specific exons in brain tumors suggests that we need to increase the analysis dimension at the exon level. Moreover, existing translation prediction strategies only simulate open reading frames (ORF) near the splicing site, ignoring the full-length characteristics of the transcript and its actual translation potential. This leads to a significant increase in false positive rate, thereby increasing the cost of clinical verification.

[0005] The above technical bottlenecks seriously restrict the clinical application value of transcription-derived neoantigens. Therefore, it is urgent to develop new strategies that integrate exon feature analysis, translation potential evaluation, and multi-dimensional normal tissue control to improve the identification efficiency of tumor-specific transcripts and their derived neoantigens. SUMMARY

[0006] The present application provides a method for identifying transcription-derived tumor neoantigens, which identifies tumor-specific transcripts by integrating splicing events and exons generated by abnormal transcription, and predicts tumor neoantigens from the protein sequences generated by the transcripts. This method improves the identification strategy of tumor-specific transcripts, increases the identification of single-exon transcripts, and optimizes the process of identifying tumor-specific transcripts and predicting their translation into proteins. This method can efficiently identify tumor-specific transcripts and their derived neoantigen peptides from patient RNA-seq data, providing strong technical support for the development of tumor neoantigen vaccines.

[0007] The application also provides a device for identifying tumor neoantigens from transcription, which comprises a pretreatment module, a tumor-specific splicing event module, a tumor-specific exon module and a tumor-specific transcript integration and neoantigen prediction module.

[0008] The technical scheme provided by the application is as follows:

[0009] A method for identifying tumor neoantigens from transcription comprises the following steps:

[0010] Aligning RNA-seq data of a patient, transcript splicing of the aligned result and transcript coding capacity prediction;

[0011] Obtaining splicing events of transcripts and annotating the same;

[0012] Extracting and quantifying exons from the spliced transcripts, and calculating the expression level of the exons by the following formula:

[0013] e_ncov=(e_cov / unique_mapped_read_of_STAR)*1e7

[0014] Wherein, e_cov represents exon coverage, and unique_mapped_read_of_STAR is the number of uniquely mapped reads of STAR after RNA-seq alignment;

[0015] Integrating tumor-specific transcripts generated by splicing events and exons, and predicting tumor neoantigens.

[0016] Further, aligning RNA-seq data of a patient by using STAR, transcript splicing of the aligned result by using StringTie, and evaluating the coding potential of the transcripts by using CPC2 and CPAT;

[0017] Obtaining splicing events of transcripts by using ASJA tool and annotating the same;

[0018] Extracting and quantifying exons from the spliced transcripts by using self-written scripts, and calculating the expression level of the exons by the following formula:

[0019] e_ncov=(e_cov / unique_mapped_read_of_STAR)*1e7

[0020] Wherein, e_cov represents the exon coverage obtained from the StringTie result, unique_mapped_read_of_STAR is the number of reads uniquely mapped to the sample after the RNA-seq is aligned by STAR;

[0021] By predicting the coding potential of the tumor-specific transcripts of the splicing events and exons by CPAT and CPC2, the peptide segment windowing of the multi-exon transcripts ORF and the single-exon ORF with translation potential generated by specific splicing events or exons is performed in 8-11, and the MHC class I neoantigen prediction and screening are performed.

[0022] Further, the RNA-seq data of the patient is aligned by using STAR, the aligned result is spliced by StringTie, and the coding potential of the transcript is predicted by CPAT and CPC, including the following steps:

[0023] The sample, bam_dir, genomeDir, and genomeFastaFiles parameters in star.sh are modified by using STAR;

[0024] The sample, bam_dir, and gtf parameters in 02_stringtie.sh are modified by using StringTie;

[0025] The result / coding directory is created, and the refs, stringtie_dir, and outdir_coding parameters in coding_pred.sh are modified as needed;

[0026] The reference file for splicing event and exon annotation is created, and the reference file includes the known splicing event file, the bed format file of the known splicing event, the bed format file of the known exon, and the gtf, out_dir, refs_junction_file, refs_junction_file_bed, refs_exon_file_bed, refs_genome, and gffread parameters in setup.sh are modified by the user according to the actual situation.

[0027] Further, the splicing events of the transcript are classified as ‘lowexpr’, ‘specific’, ‘Hfc’, and ‘nosig’, and the tumor-specific splicing events defined as ‘specific’ and ‘Hfc’ and all the information thereof are extracted and stored in a file named sp.junction.txt.

[0028] Further, the tumor specificity of the exons is defined, including 'lowexpr','specific', 'Hfc' and 'nosig', the tumor-specific exons defined as'specific' and 'Hfc' and all information thereof are extracted and stored in a file named sp.exon.txt.

[0029] Further, before the coding potential of the tumor-specific transcript defined by the splicing events and exons predicted by CPAT and CPC2, the following steps are further included:

[0030] With 3178 samples of 29 normal tissues in the GTEx database as the default control queue, the tumor-specific transcripts are screened by comparing the splicing events and exon expression differences between human tumor samples and the control queue.

[0031] Further, the splicing events of the transcripts are obtained and annotated by the ASJA tool, specifically:

[0032] The parameters in batch_junction.sh are modified, including sample, gtf, stringtie_dir, coding, refs_junction_file_bed, refs_exon_file_bed, outdir, min_expr, maxfc, minfc, wherein min_expr is the threshold of the minimum expression of the tumor-specific splicing event, maxfc is the fold change of the splicing event compared with the maximum expression value of each normal tissue, minfc is the fold change of the splicing event compared with the median expression value in each normal tissue, and the default control normal tissue file of human is GETX_adjTCGA.txt, which is controlled by the tcga_junc parameter.

[0033] Further, the exons are extracted and quantified from the spliced transcripts by the self-written script, specifically:

[0034] Modify the parameters in batch_Exon.sh, including sample, STAR_dir, stringtie_dir, coding, out_dir, genecode_sort_exon, gencode_v29_annotation_junctions_filtered, min_expr, maxfc, minfc, where min_expr is the threshold of the minimum expression of tumor-specific exon, maxfc is the fold change of the exon compared to the maximum expression value of each normal tissue, minfc is the fold change of the exon compared to the median expression value of each normal tissue, the default control normal tissue of human includes 29 normal tissues and TCGA paracancerous tissues, and the exon reference file of multi-exon transcript is controlled by gtexnormal_path, gtexnormal_1, gtexnormal parameters, where gtexnormal_path is the path of the reference file of the file, gtexnormal_1 is the name of the first normal tissue, and the name of the remaining normal tissue is input by the gtexnormal parameter.

[0035] Further, the tumor-specific transcripts with translation potential are screened by CPAT and CPC2 results, the ORF of the multi-exon transcript generated by specific splicing events or exons and the ORF of the single exon with translation potential are subjected to 8-11 peptide segment windowing, MHC class I neoantigen prediction and screening, specifically:

[0036] modify the parameters in batch_jun_exon_neo.sh, including sample, junction_path, exon_path, coding_path, hla_path, out_dir, RPAY, exonpos, intronpos, uniport, netmhc, genetype, fitter_rps_hla_IG_ref, tpm_cutoff, UTR5_len, wherein the junction_path, exon_path, coding_path parameters are directories for storing tumor-specific splicing events, tumor-specific exons and transcript coding potential respectively; exonpos and intronpos point to the bed files of known exons and introns; Uniport points to the modified and sorted protein sequences downloaded from uniport; netmhc points to the execution file and its directory of netMHCpan; hla_path points to the HLA typing file of the sample; RPAY, fitter_rps_hla_IG_ref, tpm_cutoff, UTR5_len are respectively the files and thresholds for filtering the integrated tumor-specific transcripts, in turn, the gene list file with the gene ID as _PAR_Y at the end, the ribosome, HLA and immunoglobulin gene list file, the transcript expression value and the length of 5' UTR.

[0037] The application also provides a device for identifying transcription-derived tumor neoantigens, which adopts the method for identifying transcription-derived tumor neoantigens described above, and comprises:

[0038] a preprocessing module, which is used for aligning the RNA-seq data of a patient, transcript splicing and HLA typing prediction of the aligned results;

[0039] a tumor-specific splicing event module, which is used for obtaining and annotating the splicing events of transcripts;

[0040] a tumor-specific exon module, which is used for extracting and quantifying exons from the spliced transcripts;

[0041] a tumor-specific transcript integration and neoantigen prediction module, which is used for integrating the tumor-specific transcripts generated by the tumor-specific splicing events and exons, and predicting tumor neoantigens.

[0042] Compared with the prior art, the application has the beneficial effects that:

[0043] 1. High-precision neoantigen identification

[0044] The method provided by the application accurately identifies splicing events in the sample in combination with the ASJA tool, and introduces a self-developed algorithm to comprehensively analyze the exons in the transcript, especially single exons. In screening tumor-specific transcription events, the method compares with each normal tissue, effectively avoiding the interference of normal tissue-specific transcripts. In addition, the method not only uses the CPC2 and CPAT tools to evaluate the coding potential of the transcript, but also comprehensively considers the 5'UTR length, the integrity of the start codon (ATG) and the stop codon (TAA / TGA / TAG), effectively excluding the interference of non-coding transcripts, and improving the accuracy of neoantigen prediction.

[0045] 2. Multi-dimensional transcriptome analysis

[0046] Existing researches mainly focus on the influence of splicing abnormalities on tumor transcripts, ignoring the abnormalities of single exons in the transcription and protein levels. The device provided by the application performs in-depth analysis of the transcriptome from the exon level through the tumor-specific exon module. The module not only annotates each exon with unique genomic coordinates, but also quantifies its expression level through a standardized expression formula to ensure the comparability of large-scale clinical samples. Combined with the splicing events identified by ASJA, the device comprehensively analyzes the transcriptome abnormalities from the dimensions of splicing and exons, breaks through the limitations of traditional methods which only rely on splicing events, and significantly improves the identification ability of tumor-specific neoantigens. In addition, the device provides detailed annotations of splicing events and exons, allowing users to deeply analyze the biological functions of tumor-specific transcripts and their potential neoantigen sources, providing reliable data support for subsequent immunotherapy target screening and personalized vaccine design.

[0047] 3. Parameter self-definable and adaptive

[0048] The method provided by the application supports users to flexibly adjust parameters according to research needs, including splicing event expression threshold (default > 5), exon expression threshold (default > 8), and threshold for individual comparison of each normal tissue (median fold change and maximum expression fold change). Through parameter optimization, different tumor types (such as brain tumors or lung cancer) and species (human or mouse) can be adapted, significantly improving the sensitivity and specificity of neoantigen identification.

[0049] 4. Full-process automation and efficient calculation

[0050] The application uses multi-thread parallel computing technology to realize full-process automation from data preprocessing to neoantigen prediction. By seamlessly integrating analysis tools such as ASJA, CPAT, CPC2, and NetMHCpan with self-developed algorithms, the device significantly improves the calculation efficiency while ensuring the analysis accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 Structure diagram of the device for identifying tumor-specific transcripts in an embodiment of the present application;

[0052] Figure 2 Diagram for identification and quantification of exons in tumor-specific exon module in an embodiment of the present application;

[0053] Figure 3 Diagram for analysis method and process of tumor-specific splicing events and exons in an embodiment of the present application;

[0054] Figure 4 Diagram for tumor-specific transcript integration and neoantigen prediction by tumor-specific transcript integration and neoantigen prediction module in an embodiment of the present application;

[0055] Figure 5 Diagram for identification of tumor-specific transcript-derived neoantigens in 140 pancreatic cancer samples in an embodiment of the present application, wherein (A) tumor-specific transcripts with coding potential and the resulting neo-peptide sequences in each sample; (B) the circular diagram shows the number of tumor-specific transcript-derived neoantigens identified in each tumor sample, the average affinity of the transcript-derived neoantigens to HLA, the number of tumor-specific transcripts that generate neoantigens, and the number of mutation-derived neoantigens;

[0056] Figure 6 Diagram for analysis of tumor-specific transcripts in pancreatic cancer proteomics in an embodiment of the present application, wherein (A) the number of tumor-specific transcripts expressed in the protein spectrum in each sample; (B) visualization of tumor-specific splicing events of the PDIA3 gene; (C) protein spectrum visualization of polypeptides generated by tumor-specific splicing events of the PDIA3 gene;

[0057] Figure 7 Diagram for correlation of tumor-specific transcript-derived neoantigens and mutation-derived neoantigens in an embodiment of the present application, wherein (A) distribution of the ratio of tumor-specific transcript-derived neoantigens to mutation-derived neoantigens in each tumor sample; (B-C) correlation of tumor-specific transcript-derived neoantigen load, mutation-derived neoantigen load, and patient overall survival under the condition of the optimal grouping threshold;

[0058] Figure 8 Diagram for identification of tumor-specific transcripts and their neoantigens in Panc02 model in an embodiment of the present application, wherein (A) is a result diagram of the analysis flowchart; (B) tumor neoantigens and their corresponding tumor-specific transcription events for in vivo experimental verification;

[0059] Figure 9Figure for in vivo experiment procedure and immune effect evaluation of tumor neoantigen in embodiments of the present application, wherein (A) is a schematic diagram of in vivo experiment procedure of tumor neoantigen, including vaccination time points (5th day, 12th day) and tumor growth monitoring scheme; (B) is a comparison of tumor volumes of mice in each group on the 35th day, showing tumor size differences of tumor-specific transcript neoantigen vaccine group, mutant neoantigen vaccine group and control group; (C) is a dynamic change curve of tumor volumes of mice at different time points (7th day to 33rd day), reflecting the vaccine treatment effect; (D-E) are detection of neoantigen reactive CD8+ T cell and CD4+ T cell reaction intensity by intracellular factor labeling (IFNγ) and fluorescence activated cell sorting (Fluorescence-Activated Cell Sorting, FACS).

[0060] The reference signs are as follows:

[0061] 1 - pretreatment module, 2 - tumor-specific splicing event module, 3 - tumor-specific exon module, 4 - tumor-specific transcript integration and neoantigen prediction module. DETAILED DESCRIPTION

[0062] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the embodiments described below are part of the embodiments of the present application, rather than all the embodiments of the present application.

[0063] Reference Figure 1 The present application provides a method for identifying transcription-derived tumor neoantigens, comprising the following steps:

[0064] Step S1, aligning RNA-seq data of a patient, transcript splicing and transcript coding ability prediction are performed on the aligned results;

[0065] Step S2, obtaining splicing events of transcripts and annotating the same;

[0066] Step S3, extracting and quantifying exons from the spliced transcripts, and the exon expression level is calculated by the following formula:

[0067] e_ncov = (e_cov / unique_mapped_read_of_STAR)*1e7

[0068] Wherein, e_cov represents exon coverage, and unique_mapped_read_of_STAR is the number of uniquely aligned reads of RNA-seq to the sample after alignment;

[0069] Step S4, integrating tumor-specific splicing events and tumor-specificity generated by exons, and predicting tumor neoantigens.

[0070] The method for identifying tumor neoantigens of transcriptional origin provided by the application is developed based on Perl, Python and Shell languages, relies on multiple Python packages and local functions, including BioPython, re, ASJA, CPAT, CPC2, NetMHCpan and bedtools. In addition, the alignment and transcript splicing of RNA-seq data need to be completed using STAR and StringTie tools. Before using the method, it is ensured that all necessary basic packages and tools have been installed. The method integrates splicing event analysis (based on the ASJA tool) and exon-level analysis (based on a self-developed exon quantification algorithm), breaking through the limitations of traditional methods relying only on splicing events, and realizing the simultaneous detection of splicing events, single-exon (s.exon) and multi-exon (m.exon) variations.

[0071] The present application takes 3178 samples of 29 normal tissues in the GTEx database as the default control queue. For the analysis of specific splicing events and exons of mouse tumor models, this study downloaded the RNA-seq of 19 normal mouse tissues published by PRJNA375882, PRJNA859788 and PRJNA435886 from the SRA public database, as well as the unpublished normal mouse liver, as controls. The splicing events and exons of the control samples were obtained in the same way as the tumor samples.

[0072] By comparing the splicing events and exon expression differences between human tumor samples and control queues, tumor-specific transcripts are screened. The specific screening criteria are as follows:

[0073] Tumor-specific splicing events: the default expression level is greater than 5, and meets any of the following conditions:

[0074] 1) The splicing event has no expression in any control tissue;

[0075] 2) The expression of the splicing event in the tumor is at least 10 times the maximum expression value of any control tissue, and is 5 times the median expression amount of any control tissue, and the frequency in any control tissue is less than 0.9;

[0076] 3) The frequency of the splicing event in any control tissue is less than 0.01, and is 5 times the median expression amount of any control tissue.

[0077] Tumor-specific exons: The same threshold and condition as splicing events were used, with the additional requirement that the minimum expression value of the exon is above 8. Due to the lower accuracy of RNA-seq in aligning the UTRs of genes, the following rules were made to compare the first or last exon of a transcript: for the first exon of a transcript, it was considered comparable if the donor splice site in tumor and control was the same; for the last exon, it was considered comparable if the acceptor splice site in tumor and control was the same. Through the above strict screening criteria, the tumor-specific transcription events identified meet the high reliability standard.

[0078] In step S1, the STAR (version 2.7.8) was used to align the RNA-seq data of the patient, and the aligned results were subjected to transcript splicing by StringTie (version 2.2.1), and the coding potential of the transcript was evaluated by CPC2 (version 0.1) and CPAT (version 3.0.4) under the default settings.

[0079] CPAT needs to construct a model based on species-specific coding sequences, so the present application obtains the CDS sequences of non-coding genes and coding genes from GRCh38 by the gffread tool, and uses make_hexamer_tab.py and make_logitModel.py to construct the CPAT prediction model. For mouse data, the same steps were used to construct the reference file, and the genome sequence was replaced by GRCm38_68. CPC2 uses a universal algorithm and does not need to be adapted to species, and the present application uses the default parameters of this tool to predict the coding ability of the transcript.

[0080] The definition of a coding transcript needs to meet the following conditions at the same time:

[0081] 1) The CPAT score of the transcript of human origin is greater than 0.5, and the CPAT score of the transcript of mouse origin is greater than 0.44;

[0082] 2) The translation initiation codon is ATG, and the termination codon is TAA, TGA or TAG;

[0083] 3) The transcript is identified as a protein coding by CPC2.

[0084] In step S2, ASJA tool is used to identify and annotate splicing events from RNA-seq data. ASJA is based on STAR alignment and StringTie assembled transcript data, which can comprehensively describe the genomic location and normalized expression level of all splicing events. This tool is not only suitable for the identification of known splicing events, but also can effectively find unannotated splicing events. To clarify the source of splicing events, the present application uses bedtools intersect-s command to compare the splicing sites with the known exon and intron genomic locations in the GENCODE database. The intersection of splicing sites with exons is denoted as "E", with introns as "I", and with intergenic regions as "intergenic". If the splicing event is not annotated in GENCODE, the splicing event is added with an 'AS' label. To simplify the classification, the present application defines the splicing event type as: "E-E" (known splicing event), "E-E-AS" (exon-related abnormal splicing), "E-E-IAS" (exon-intron splicing), while "E-I-AS", "I-E-AS", "I-I-AS" and "I-I-IAS" are classified as intron-related splicing events, and the rest are classified as intergenic region splicing events. If a splicing event produces multiple transcripts, the present application defaults to select the transcript with the highest CPAT predicted coding score or the highest expression (TPM) as the representative transcript. This procedure is compatible with human and mouse data analysis.

[0085] In step S3, exons are the core components of mature RNA, and different combinations of exons can generate new protein sequences, so the transcriptome variation analysis at the level of exons is of great significance. Compared with splicing events, exon analysis can not only identify exons in multi-exon transcripts (m. exons), but also effectively identify exons in single-exon transcripts (s. exons). In order to obtain the genomic location and expression level of the exon, the present application extracts the genomic location information of the exon from the transcripts assembled by StringTie. For single-exon that can be translated into protein, the start and end points of its ORF are taken as the genomic location; for non-coding single-exon, the genomic location defined from the transcript is retained. The expression level of the exon is calculated by the following formula: e_ncov = (e_cov / unique_mapped_read_of_STAR)*1e7. Wherein, e_cov represents the exon coverage (obtained from StringTie results), and unique_mapped_read_of_STAR is the number of uniquely mapped reads to the sample after RNA-seq is aligned by STAR. If the number of uniquely mapped reads cannot be obtained, the sum of e_cov is used instead. By this method, the present application can accurately quantify the expression level of the exon and ensure the comparability between samples. If an exon is the component of multiple transcripts, the transcript with the highest expression (TPM) is selected as the representative transcript of the exon. This process is compatible with human and mouse data analysis.

[0086] In step S3, the genomic location and coverage of the exon of the transcript spliced by StringTie are identified, and each exon is given unique location information and standardized expression level.

[0087] In step S4, tumor-specific transcripts generated by tumor-specific splicing events and exons are integrated, and tumor neoantigens are predicted. The input files of this step are sp.junction.txt and sp.exon.txt and HLA typing file. Tumor neoantigens are predicted by screening transcripts with coding ability and ORF containing specific splicing events or exons. At the same time, this step can also provide other screening conditions for users to choose, such as filtering ribosomal genes and HLA class I and II genes and immunoglobulin genes, low expression transcripts (such as TPM < 1) and limiting the length of 5UTR. Finally, the remaining tumor-specific transcripts are converted into amino acid sequences, and the amino acid sequences containing 11 amino acids upstream of the specific splicing event to the end of the coding region and the HLA typing of the sample are taken as the output file of NetMHCpan to predict antigens. Antigens meeting the threshold screening requirements are defined as tumor neoantigens of the sample.

[0088] The application also provides a device for identifying tumor neoantigens derived from transcription, which is based on the above-mentioned method for identifying tumor neoantigens derived from transcription and at least comprises the following four modules.

[0089] A preprocessing module 1, which is used in combination with a processing module to align the RNA-seq data of a patient, splice the aligned results, and perform HLA typing prediction.

[0090] A tumor-specific splicing event module 2, which is used to obtain and annotate the splicing events of transcripts.

[0091] A tumor-specific exon module 3, which is used to extract and quantify exons from the spliced transcripts.

[0092] A tumor-specific transcript integration and neoantigen prediction module 4, which is used to integrate the tumor-specific transcripts generated by tumor-specific splicing events and exons, and predict tumor neoantigens.

[0093] The device for identifying tumor neoantigens derived from transcription provided by the application can identify tumor-specific transcripts and the new antigens generated by the tumor-specific transcripts of any tumor patient through the sequential operation of the four modules.

[0094] As shown in Figure 2 , the tumor-specific exon module 3 assigns a unique location information and a normalized expression level to each exon by identifying the exon genomic position and coverage of the StringTie spliced transcript.

[0095] The analysis method and process of tumor-specific splicing events and exons by the device are shown in Figure 3 , wherein A: constructing the expression matrix of splicing events or exons of tumor tissue and control tissue; B: judging the specificity of splicing events and exons by comparison with each control tissue.

[0096] As shown in Figure 4 , the tumor-specific transcript integration and neoantigen prediction module 4 integrates the tumor-specific transcripts defined by splicing events and exons, and screens the tumor-specific transcripts with coding potential by CPAT and CPC2 results, performs 8-11 peptide windowing on the multi-exon transcripts ORF generated by specific splicing events or exons and the ORF of single-exon with translation potential, and performs MHC class I neoantigen prediction and screening.

[0097] The device performs the preliminary data preparation through the preprocessing module 1, including the alignment of RNA-seq data, the splicing of transcripts, etc. These results are used as the input of the subsequent three modules, so as to perform the analysis of the subsequent three modules to obtain tumor neoantigens.

[0098] The task of the tumor-specific splicing event module 2 is to identify the splicing events of the sample through ASJA and annotate the splicing events thereof. In this module, the representative transcripts defining the splicing events, and the prediction results of the coding ability of the transcripts by CPAT and CPC2 are also included. Finally, the tumor specificity of each splicing event is defined by strict threshold screening, including 'lowexpr','specific', 'Hfc' and 'nosig'. Finally, the tumor-specific splicing events defined as'specific' and 'Hfc' and all their information are extracted and stored in a file named sp.junction.txt.

[0099] The task of the tumor-specific exon module 3 is performed synchronously by the device. This module can identify the exons of the sample and assign each exon (including single exon) a unique genomic location and its expression level. At the same time, each exon is annotated, and the representative transcript of the exon is defined, and the prediction results of the coding ability of the transcripts by CPAT and CPC2 are integrated. Finally, the tumor specificity of each exon is defined by strict threshold screening, including 'lowexpr','specific', 'Hfc' and 'nosig'. Finally, the tumor-specific exons defined as'specific' and 'Hfc' and all their information are extracted and stored in a file named sp.exon.txt.

[0100] The task of the module 4 of tumor-specific transcript integration and neoantigen prediction is to integrate the transcriptome-specific events obtained through splicing events and exons, and to predict tumor neoantigens. The input files of this module are sp.junction.txt and sp.exon.txt and HLA typing file. This module performs tumor neoantigen prediction by screening the transcripts with coding ability and ORF covering specific splicing events or exons. At the same time, this module also has other screening conditions for user selection, such as filtering ribosomal genes and HLA class I and II genes and immunoglobulin genes, low expression transcripts (such as TPM<1) and limiting the length of 5UTR. Finally, the remaining tumor-specific transcripts are converted into amino acid sequences, and the amino acid sequences containing 11 amino acids upstream of the specific splicing event to the end of the coding region and the HLA typing of the sample are used as the output file of NetMHCpan to predict antigens. The antigens meeting the threshold screening requirements are defined as tumor neoantigens of the sample.

[0101] Embodiment 1

[0102] The embodiment provides a method for identifying tumor neoantigens of transcriptional origin, and the method is implemented as follows:

[0103] Step S1, aligning RNA-seq data of a patient, transcript splicing of the aligned results, and transcript coding ability prediction

[0104] S1.1 aligning RNA-seq data of a patient

[0105] The user modifies the sample, bam_dir, genomeDir, and genomeFastaFiles parameters in star.sh according to the actual situation.

[0106] Run the command: sh star.sh

[0107] S1.2 transcript splicing

[0108] The user modifies the sample, bam_dir, and gtf parameters in 02_stringtie.sh according to the actual situation.

[0109] Run the command: sh 02_stringtie.sh

[0110] S1.3 transcript coding ability prediction

[0111] Create the result / coding directory, and the user modifies the refs, stringtie_dir, and outdir_coding parameters in coding_pred.sh according to the actual situation.

[0112] Run the command: sh coding_pred.sh

[0113] S1.4 preparation of reference files

[0114] The user needs to create reference files for splice event and exon annotation, including a known splice event file, a bed format file of the known splice event, a bed format file of the known exon, and the like. The user modifies the gtf, out_dir, refs_junction_file, refs_junction_file_bed, refs_exon_file_bed, refs_genome, and gffread parameters in setup.sh according to the actual situation.

[0115] Run the command: sh setup.sh

[0116] Step S2, get the splicing events of the transcript and annotate them, the specific operation and command are as follows:

[0117] The user needs to modify the parameters in batch_junction.sh, including sample, gtf, stringtie_dir, coding, refs_junction_file_bed, refs_exon_file_bed, outdir, min_expr, maxfc, minfc. Among them, min_expr is the minimum expression threshold of tumor-specific splicing events, maxfc is the fold change of splicing events compared with the maximum expression value of each normal tissue, and minfc is the fold change of splicing events compared with the median expression value of each normal tissue, and their default thresholds are 5, 10, 5 respectively. The default control normal tissue file of human is GETX_adjTCGA.txt (including 29 normal tissues and TCGA paracancer tissues), which is controlled by the tcga_junc parameter. The results of tumor-specific splicing events will be stored in the following file result / Junction / sample / sp.junction.txt

[0118] The command for running human tumor tissue samples is: sh batch_junction.sh

[0119] The command for running mouse tumor cell samples is: sh batch_junction_mus.sh

[0120] Step S3, extract and quantify exons from spliced transcripts

[0121] Users need to modify the parameters in batch_Exon.sh, including sample, STAR_dir, stringtie_dir, coding, out_dir, genecode_sort_exon, gencode_v29_annotation_junctions_filtered, min_expr, maxfc, minfc. Among them, min_expr is the minimum expression threshold of tumor-specific exon, maxfc is the fold change of exon compared with the maximum expression value of each normal tissue, and minfc is the fold change of exon compared with the median expression value of each normal tissue, and their default thresholds are 8, 10, 5 respectively. The default control normal tissue of human includes 29 normal tissues and TCGA paracancer tissues, and the exon reference file of multi-exon transcript is controlled by gtexnormal_path, gtexnormal_1, gtexnormal parameters, among which gtexnormal_path is the path of the reference file of the file, gtexnormal_1 is the name of the first normal tissue, and the name of the remaining normal tissue is transmitted by the gtexnormal parameter. The default reference file of single-exon of normal tissue is 'TCGA_adj_GTEX_SE_exon.txt', which includes single-exon expressed in 29 normal tissues and TCGA paracancer tissues, and is controlled by gtexnomral_SE parameter. The results of tumor-specific exon will be stored in the file result / exon / sample / sp.exon.txt

[0122] The running command of human tumor tissue sample is: sh batch_Exon.sh

[0123] The running command of mouse tumor cell sample is: sh batch_exon_mus.sh

[0124] Step S4, tumor-specific transcript integration and neoantigen prediction

[0125] The user needs to modify the parameters in batch_jun_exon_neo.sh, including sample, junction_path, exon_path, coding_path, hla_path, out_dir, RPAY, exonpos, intronpos, uniport, netmhc, genetype, fitter_rps_hla_IG_ref, tpm_cutoff, UTR5_len. Among them, the junction_path, exon_path, coding_path parameters are the directories storing tumor-specific splicing events, tumor-specific exons and transcript coding potential respectively; exonpos and intronpos point to the bed files of known exons and introns; Uniport points to the modified and sorted protein sequences downloaded from uniport; netmhc points to the execution file and its directory of netMHCpan; hla_path points to the HLA typing file of the sample, which includes 2 columns, respectively, the sample name and its corresponding typing (4-digit); RPAY, fitter_rps_hla_IG_ref, tpm_cutoff, UTR5_len are the files and thresholds for filtering the integrated tumor-specific transcripts, respectively, the gene list file with gene ID as _PAR_Y at the end, the ribosome, HLA and immunoglobulin gene list file, the transcript expression value (default TPM>1) and the length of 5'UTR (default length 1000). The tumor neoantigen and its corresponding transcript results will be stored in the following file outdir / sample / interget.ej_pep_ic500.txt.

[0126] Human tumor tissue sample running command: sh batch_jun_exon.sh

[0127] Mouse tumor cell sample running command: sh batch_jun_exon_mus.sh

[0128] The final file intergate.ej_pep_ic500.txt containing the results of running the human tumor tissue samples and mouse tumor cell line samples contains the following information: genomic position (ID) of the tumor-specific splice event or exon, expression level of the tumor-specific splice event or exon (Expr), genomic annotation of the tumor-specific splice event or exon (Pos), type annotation of the tumor-specific splice event or exon (Class), tumor-specific transcript ID (STRINGTID), length of the tumor-specific splice event or exon (ej_length), position of the tumor-specific splice event or exon in the transcript (ej_pos), gene name annotation of the tumor-specific transcript (Gene), expression level of the tumor-specific transcript (TPM), CPAT predicted score of the tumor-specific transcript (CPAT), stop codon of the CPAT predicted tumor-specific transcript (STOP), start and end positions of the ORF of the CPC predicted tumor-specific transcript (cpcORF_Start, cpcORF_end), coding type of the CPC predicted tumor-specific transcript (cpclabel), start and stop codons of the CPC predicted tumor-specific transcript (CPCst, CPCed), expression of the tumor-specific splice event or exon in each normal control tissue, tumor-specific splice event or exon specific type (Type), tumor neoantigen sequence produced by the tumor-specific transcript (pep).

[0129] The directory structure contained in the method is as follows:

[0130]

[0131]

[0132] Example 2

[0133] This example uses human pancreatic cancer patient tumor tissues and mouse pancreatic cancer Panc02 tumor cell line to evaluate the effectiveness of the method provided by the present application for identifying transcription-derived tumor neoantigens.

[0134] BAM files of 140 pancreatic cancer patients and 39 cancer adjacent tissues RNA-seq data in CPTAC cohort were downloaded from GDC portal (https: / / portal.gdc.cancer.gov / ) using gdc-client in step S1. Then, the command sh 02_stringtie.sh was run to perform transcript assembly using StringTie with GENCODE as the reference annotation. To obtain the coding potential of each transcript, the command sh coding_pred.sh was run to predict each transcript and obtain the ORF start and stop codons and the start and stop positions of the translated protein.

[0135] HLA typing prediction was performed for each sample in the pancreatic cancer cohort using arcasHLA software (version 3.9). The BAM files of RNA-seq were processed by the extract, genotype, partial and merge functions of arcasHLA in sequence to obtain the final MHC class-I haplotype. To ensure compatibility with the subsequent NetMHCpan prediction, only the HLA types that were reliably predicted by NetMHCpan were retained for further analysis.

[0136] Specific splicing events and exons in pancreatic cancer tumor tissues were identified by step S2 and step S3, and 3178 samples of 29 tissues in GTEx database and 39 cancer adjacent tissue samples constituted the control cohort. Tumor-specific splicing events were obtained by running the command sh batch_junction.sh, with the threshold settings as follows:

[0137] Splicing events with expression greater than 5 and no expression in normal samples;

[0138] Splicing events with expression greater than 5, expression frequency less than 0.01 in normal samples, and median expression fold change greater than 5;

[0139] Splicing events with expression greater than 5, maximum expression fold change greater than 10 in normal samples, median expression fold change greater than 5, and expression frequency less than 0.9.

[0140] Tumor-specific exons were obtained by running the command sh batch_Exon.sh, using the same thresholds as above, but with a minimum expression threshold of 8.

[0141] The tumor-specific transcripts defined by tumor-specific splicing events and exons were integrated in a single sample unit by executing the sh batch_jun_exon.sh command, and their coding sequences were extracted for tumor neoantigen prediction. The specific screening criteria are as follows: the CPAT prediction score threshold is set to 0.5, and there is no restriction on the TPM value and 5'UTR length of the transcript, while retaining the genes with gene ID ending with _PAR_Y, ribosomal genes, HLA genes and immunoglobulin genes, and the other parameters are set to default. For each patient, the binding affinity of the 8-11 amino acid peptide to its HLA molecule is predicted by NetMHCpan, and other parameters are set to default.

[0142] Prediction of pancreatic cancer patient mutation neoantigens

[0143] The annotated VCF files of CPTAC pancreatic cancer patients' whole exome sequencing analysis were downloaded from the GDC portal, and the gene expression data (TPM) were also downloaded at the same time. The mismatch mutations were screened, and the mutations with sequencing depth greater than 5 and TPM greater than 10 were selected for mutation neoantigen prediction. The mutant protein sequence was obtained using snpEff.jar. Through a custom script, the mutant amino acid and its upstream and downstream 11 amino acids (a total of 23 amino acids) were included in the neoantigen prediction. NetMHCpan was used for prediction with the same settings and thresholds as described above.

[0144] Alignment and transcript splicing of Panc02 cell RNA-seq

[0145] After quality control (QC) of the original RNA-seq data of Panc02, the sh star.sh command was used to align the reads to the mouse reference genome GRCm38.91 (downloaded from the Ensembl database). Subsequently, the sh 02_stringtie.sh command was run to splice the aligned data into transcripts using StringTie, with Mus_musculus.GRCm38.91 as the reference annotation. To obtain the coding potential of each transcript, we ran the sh coding_pred.sh command to predict each transcript and obtain the ORF start and stop codons and the start and end of the translated positions.

[0146] Tumor neoantigen prediction process of Panc02 cell line tumor-specific transcript-derived tumor neoantigens

[0147] Specificity splicing events and exons in Panc02 cell line were identified by comparing with control samples from SRA projects PRJNA375882, PRJNA859788, PRJNA435886 and unpublished normal mouse liver RNA-seq. The thresholds for tumor-specific splicing events were set as follows:

[0148] Splicing events with expression level greater than 5 and no expression in normal samples;

[0149] Splicing events with expression level greater than 5, expression frequency less than 0.01 in normal samples and median expression fold change greater than 5;

[0150] Splicing events with expression level greater than 5, maximum expression fold change greater than 10 in normal samples, median expression fold change greater than 5 and expression frequency less than 0.9.

[0151] The identification of tumor-specific exons used the same thresholds, but the minimum expression threshold was set to 8.

[0152] Panc02 tumor-specific splicing events and exons defined tumor-specific transcripts were combined by sh batch_jun_exon_mus.sh command, and their coding sequences were extracted for tumor neoantigen prediction, with CPAT prediction score threshold set to 0.44. NetMHCpan was used to predict the binding affinity of 8-11 amino acid peptides to H-2-Db and H-2-Kb. The rest used the default settings.

[0153] Pancreatic cancer patient tumor-specific transcript-derived neoantigen prediction results

[0154] Through tumor-specific splicing site identification, exon quantification analysis and neoantigen transformation prediction, a median of 30 tumor-specific transcripts with coding potential can be identified in each patient, and 143 new polypeptide sequences can be derived from them Figure 5 A). After strict screening, a median of 31 (range: 1-302) high-confidence tumor-specific transcript-derived neoantigens can be identified in each patient, which is significantly higher than the number of mutation-derived neoantigens (median value is 17, range: 0-100). Notably, the average HLA binding affinity of transcript-derived neoantigens is 89.2 nM, suggesting that they have stronger immunogenic potential Figure 5 B).

[0155] To verify the reliability of tumor-specific transcripts, the corresponding protein mass spectrometry data of each pancreatic cancer patient was analyzed, and the results showed that 17.3% of the tumor-specific transcripts identified in each sample were supported by mass spectrometry data Figure 6A). A typical example is the MHC-I restricted antigenic peptide FIQENMVMM of the PDIA3 gene (binding affinity: 32 nM), which is translated from a transcript generated by an exon skipping event Figure 6 B-C). The results demonstrate that the screened transcripts of the application have biological functional activity, and their degradation products can participate in the immune response as real tumor antigens.

[0156] Further comparison with mutation-derived neoantigens found that the transcription-derived neoantigens have a clinical compensatory effect. The transcription-derived neoantigens and mutation-derived neoantigens are significantly negatively correlated in the distribution of the number ratio Figure 7 A). In pancreatic cancer patients with low load of mutation-derived neoantigens (0 mutation-derived neoantigens) (n = 3), the method can identify a median of 18 transcription-derived neoantigens. This finding shows that when the genomic mutation level is insufficient, the application can provide key target supplements for individualized vaccine design through in-depth analysis at the transcriptome level. In addition, based on Kaplan-Meier survival analysis, the median overall survival (OS) of patients with high load of transcription-derived neoantigens is significantly better than that of the low load group (Log-rank P = 0.025, Figure 7 B), while the load of mutation-derived neoantigens has no significant correlation with OS (P = 0.41) Figure 7 C). The results suggest that in patients with high load of transcription-derived neoantigens, more tumor-specific antigens are presented by MHC-I molecules, activating CD8+ T cell-mediated anti-tumor immune responses, thereby effectively inhibiting tumor growth and spread.

[0157] Prediction and experimental verification of tumor neoantigens of Panc02 cell line

[0158] Based on the above analysis process, 297 specific splicing events or exons were identified in Panc02 cells, including 81 coding transcripts. Forty-three coding transcripts of new ORFs generated by tumor-specific splicing events and exons were retained, which were predicted to show binding (SB or WB) with H2-Kb / H2-Db. Finally, 93 tumor neoantigens were identified from 21 tumor-specific transcripts defined by 35 splicing events or exons Figure 8 A).

[0159] In the Panc02 mouse pancreatic cancer model, 5 tumor-specific transcription-derived tumor neoantigens were screened for experimental verification Figure 8 B). To evaluate their therapeutic effect, mice were subcutaneously injected with Panc02 to form tumors, and on the 5th and 12th days, they were inoculated with a vaccine composed of 5 tumor neoantigen sequences, while the control group of mice was not treated Figure 9 A). At the same time, 17 mutation-derived neoantigens of Panc02 were also synthesized into vaccines and tested in mice.

[0160] The experimental results show that, compared with the control group, the tumor volume of the tumor-specific transcript-derived neoantigen vaccine group and the mutant neoantigen vaccine group is significantly reduced (P<0.05), and 66% (4 / 6) of the mice completely regressed on the 35th day Figure 9 B). Notably, the effect of the tumor-specific transcript-derived neoantigen vaccine is better than that of the mutant neoantigen vaccine group. Through the restimulation of the in vitro neoantigen, the proliferation of the reactive CD8+T cell population is successfully induced Figure 9 D). Among them, the reaction against the R17-derived neoantigen (an unannotated single exon) is particularly significant Figure 9 D). In addition, the experimental data show that all tumor-specific transcript-derived neoantigens can induce the proliferation of neoantigen-reactive CD4+T cells, among which the R16-derived neoantigen (an abnormal splicing of an intronic region) exhibits particularly strong activity Figure 9 E).

[0161] The above merely describes the most optimal specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any changes or replacements within the technical scope disclosed by the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for identifying transcriptionally derived tumor neoantigens, characterized in that, Includes the following steps: The patient's RNA-seq data were compared using STAR, and the results were used to splice transcripts using StringTie. The coding potential of the transcripts was assessed using CPC2 and CPAT. The splicing events of transcripts were obtained and annotated using the ASJA tool, specifically as follows: Configure the parameters in the batch_junction.sh script, including sample, gtf, stringtie_dir, coding, refs_junction_file_bed, refs_exon_file_bed, outdir, min_expr, maxfc, and minfc. Here, min_expr is the threshold for the lowest expression of tumor-specific splicing events, maxfc is the fold change of the splicing event compared to the maximum expression value in each normal tissue, and minfc is the fold change of the splicing event compared to the median expression value in each normal tissue. The default human control normal tissue file is GETX_adjTCGA.txt, controlled by the tcga_junc parameter. Exons were extracted and quantified from the spliced ​​transcripts using a self-written script, specifically as follows: The parameters in the batch_Exon.sh script are set, including sample, STAR_dir, stringtie_dir, coding, out_dir, genecode_sort_exon, gencode_v29_annotation_junctions_filtered, min_expr, maxfc, and minfc. Here, min_expr is the threshold for the lowest expression of tumor-specific exons, maxfc is the fold change of exons compared to the maximum expression value in each normal tissue, and minfc is the fold change of exons compared to the median expression value in each normal tissue. The default human control normal tissues include 29 normal tissues and TCGA adjacent normal tissues. The exon reference files constituting the multi-exon transcripts are controlled by the gtexnormal_path, gtexnormal_1, and gtexnormal parameters. gtexnormal_path is the path to the reference file, gtexnormal_1 is the name of the first normal tissue (defaulting to 'Adipose_Tissue'), and the names of the remaining normal tissues are passed in by the gtexnormal parameter. Exon expression levels are calculated using the following formula: e_ncov=(e_cov / unique_mapped_read_of_STAR)*1e7 Where e_cov represents the exon coverage obtained from the StringTie results, and unique_mapped_read_of_STAR is the number of reads that are uniquely mapped to the sample after the RNA-seq is aligned with STAR. Using 3178 samples from 29 normal tissues in the GTEx database as the default control cohort, tumor-specific transcripts were screened by comparing splicing events and exon expression differences between human tumor samples and the control cohort. Tumor-specific transcripts with translational potential were screened using CPAT and CPC2 results. ORFs of multi-exon transcripts generated by specific splicing events or exons and ORFs of single exons with translational potential were windowed with 8-11 peptide segments to predict and screen for MHC class I neoantigens.

2. The method for identifying transcription-derived tumor neoantigens according to claim 1, characterized in that, The process of using STAR to align the patient's RNA-seq data, splicing the aligned results using StringTie, and predicting the coding potential of the transcripts using CPAT and CPC includes the following steps: Use STAR to modify the sample, bam_dir, genomeDir, and genomeFastaFiles parameters in star.sh; Use StringTie to modify the sample, bam_dir, and gtf parameters in 02_stringtie.sh; Create the result / coding directory, and modify the refs, stringtie_dir, and outdir_coding parameters in coding_pred.sh as needed; Create reference files for splicing events and exon annotations. These reference files include known splicing event files, BED format files of known splicing events, and BED format files of known exons. Users can modify the gtf, out_dir, refs_junction_file, refs_junction_file_bed, refs_exon_file_bed, refs_genome, and gffread parameters in setup.sh according to their actual needs.

3. The method for identifying transcriptionally derived tumor neoantigens according to claim 1, characterized in that, The splicing events of the transcripts are classified as 'lowexpr', 'specific', 'Hfc', and 'nosig'. Tumor-specific splicing events defined as 'specific' and 'Hfc', along with all their information, are extracted and stored in a file named sp.junction.txt.

4. The method for identifying transcriptionally derived tumor neoantigens according to claim 3, characterized in that, The tumor specificity of the exons is defined, including 'lowexpr', 'specific', 'Hfc', and 'nosig'. The tumor-specific exons defined as 'specific' and 'Hfc', along with all their information, are extracted and stored in a file named sp.exon.txt.

5. The method for identifying transcriptionally derived tumor neoantigens according to claim 1, characterized in that, The process involves screening tumor-specific transcripts with translational potential using CPAT and CPC2 results, performing 8-11 peptide windowing on the ORFs of multi-exon transcripts generated by specific splicing events or exons, and on the ORFs of single-exon transcripts with translational potential, to predict and screen for MHC class I neoantigens. Specifically: The parameters in the batch_jun_exon_neo.sh script are set as follows: sample, junction_path, exon_path, coding_path, hla_path, out_dir, RPAY, exonpos, intronpos, uniport, netmhc, genetype, fitter_rps_hla_IG_ref, tpm_cutoff, and UTR5_len. The junction_path, exon_path, and coding_path parameters respectively store tumor-specific splicing events, tumor-specific exons, and transcript encoding. The directory for code potential; exonpos and intronpos point to the BED files containing known exons and introns; Uniport points to the corrected and organized protein sequences downloaded from Uniport; netmhc points to the executable file of netMHCpan and its directory; hla_path points to the HLA typing file of the sample; RPAY, fitter_rps_hla_IG_ref, tpm_cutoff, and UTR5_len are the files and thresholds for filtering the integrated tumor-specific transcripts, respectively, followed by the gene list file ending with _PAR_Y, the ribosome, HLA, and immunoglobulin gene list files, the transcript expression value, and the length of the 5'UTR.

6. An apparatus for identifying transcription-derived tumor neoantigens, employing the method for identifying transcription-derived tumor neoantigens as described in any one of claims 1-5, characterized in that, include: The preprocessing module is used to compare the patient's RNA-seq data, perform transcript splicing and HLA typing prediction on the comparison results; A tumor-specific splicing event module is used to acquire and annotate splicing events of transcripts. A tumor-specific exon module, wherein the tumor-specific exon module is used to extract and quantify exons from spliced ​​transcripts; A tumor-specific transcript integration and neoantigen prediction module is used to integrate splicing events and transcriptome-specific events obtained from exons, and to predict tumor neoantigens.

Images