CN111755067A - A screening method for tumor neoantigens - Google Patents

Abstract

The invention discloses a method for screening tumor neoantigens, aiming at solving the problem that the existing method can not screen high-quality tumor neoantigens from multiple angles. The method comprises the following specific steps: step one, selecting a variation type of a tumor; step two, RNA variation analysis is carried out on the tumor; step three, carrying out MHC molecule analysis on normal blood and tumor; step four, RNA expression analysis is carried out on the tumor; fifthly, carrying out variation annotation on the tumor; analyzing the variation driving gene of the tumor; seventhly, predicting the HLA molecule binding affinity of the tumor; step eight, analyzing the mutation frequency of the tumor; step nine, comprehensively scoring the tumor neoantigens; step ten, analyzing the synthesis difficulty of the tumor neoantigen; step eleven, comprehensively selecting the final tumor neoantigen. The invention optimally combines tumor mutation analysis and tumor expression prediction tumor specific antigen, so that the analysis process is more efficient and accurate.

Description

A screening method for tumor neoantigens

technical field

The invention relates to the field of tumor immunity, in particular to a screening method for tumor neoantigens.

Background technique

Tumor-specific antigens (tumor-specific antigens, abbreviated TSAs) refers to antigens specific to tumor cells, also known as neoantigens. Tumor-specific antigens were proposed in the first half of the last century. Later, with the development of molecular biology and the in-depth understanding of the molecular function of major histocompatibility complex (MHC), Boon et al. , The complex of specific peptides produced by tumors and MHC molecules can be recognized by T cells such as CD8+ or CD4+. Subsequent studies have recognized that these antigens that can be recognized by T cells are derived from the genomic variation of tumors and expressed as tumor-specific peptides (neo-epitopes), which are defined as neoantigens. Unlike tumor-associated antigens, tumor-specific antigens are only present in tumor cells.

In July 2017, the British scientific journal "Nature" published the results of two independent clinical phase I trials at the same time. By sequencing the DNA and RNA of tumor cells, we searched for neoantigens (neoantigens) specifically expressed by tumor cells due to gene mutations, and then constructed Personalized tumor vaccines are injected back into the body to activate immune cells and kill tumor cells with the above-mentioned antigens. This is the first cancer vaccine study to be successful in clinical trials.

The prediction methods of tumor neoantigens that have been published so far mainly include EpiToolKit and Epi-Seq. However, EpiToolKit only starts from mutations, does not consider the depth and coverage of sequencing data, and does not consider the quality of mutations in terms of data quality, so it cannot judge the quality of the obtained neoantigens. In addition, EpiToolKit does not consider the expression abundance and the expression of neoantigens, which will cause false positive predictions and cannot screen high-quality neoantigens. Many mutations at the DNA level are not expressed, and on average, 50% of the mutations may not be expressed, which may cause false positives in predicting neoantigens. Moreover, the expression of the mutation can be high or low, and the higher the expression, the stronger the overall immunogenicity. In addition, EpiToolKit does not consider the comparison between mutant peptides and normal peptides. For high-quality neoantigens, the affinity of mutant peptides is generally higher than that of normal peptides. The lack of such comparison in EpiToolKit will also result in high-quality neoantigen screening. false positive.

Epi-Seq only predicts tumor-specific antigens from tumor expression data, and predicts neoantigens from expression data, which will also cause false positives. On the one hand, it is easy to cause false positives due to the influence of RNA editing; on the other hand, because RNA sequencing is reverse-transcribed from cDNA and then sequenced, this process will also introduce a large number of false positives; on the other hand, tumorcDNA VS There are many false positives in germline DNA detection methods. The above factors lead to more false positives for the new 66 antigens obtained by Epi-Seq.

Therefore, there is currently no method that can directly screen high-quality tumor neoantigens from multiple perspectives based on the results of sequencing comparisons, and people have been conducting related research.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a screening method for tumor neoantigens to solve the problems raised in the above background art.

To achieve the above object, the present invention provides the following technical solutions:

A screening method for tumor neoantigens, the specific steps are as follows:

Step 1, select the variant type of tumor somatic cells;

Step 2, performing RNA mutation analysis on tumor somatic cells;

Step 3: Perform major histocompatibility complex (MHC) molecular analysis on normal blood and tumor somatic cells respectively;

Step 4: Perform RNA expression analysis on tumor somatic cells;

Step 5: Variation annotation on tumor somatic cells;

Step 6, analyze the mutation driving genes of tumor somatic cells;

Step 7: Predict the binding affinity of human leukocyte antigen (HLA) molecules on tumor cells;

Step 8, analyze the mutation frequency of tumor somatic cells;

Step 9, comprehensively score and sort the candidate tumor neoantigens;

Step ten, analyze the synthesis difficulty of candidate tumor neoantigens;

Step 11: Select the final tumor neoantigen based on the results of Step 9 and Step 10.

As a further solution of the present invention: the mutation types of the tumor somatic cells in step 1 include DNA point mutations, insertion deletion mutations and frameshift mutations of the tumor somatic cells.

As a further scheme of the present invention: in step 3, Optitype, xHLA and seq2HLA software are used to perform molecular analysis on normal blood and tumor somatic cells, and the results of the three are combined to determine the HLA typing results of tumor somatic cells.

As a further solution of the present invention: the variation annotation in step 5 includes the variation annotation of point mutation in tumor somatic variation, the variation annotation of indel mutation and the variation annotation of frameshift mutation.

As a further scheme of the present invention: step 6 is to analyze the point mutation, indel mutation and frameshift mutation in the tumor somatic cells to analyze the mutation-driven gene.

As a further scheme of the present invention: in step 7, the HLA molecule of the tumor somatic cells is subjected to HLA molecular analysis according to the HLA molecule type of the tumor somatic cells, the predicted mutant peptide obtained in the mutant peptide prediction step, and the wild-type peptide sequence corresponding to the predicted mutant peptide. Binding affinity prediction.

As a further scheme of the present invention: in step 9, the scoring and ranking are based on the MHC affinity of tumor somatic cells, the expression abundance of tumor somatic cells and the degree of contrast of wild-type peptides, the mutation frequency of tumor somatic cells, and whether the tumor somatic cells are RNA mutations and whether it is a tumor driver gene.

As a further scheme of the present invention: in step ten, the synthesis difficulty of candidate tumor neoantigens is analyzed according to molecular weight, isoelectric point, electrostatic charge at pH 7, average hydrophilicity and hydrophilic residue ratio.

Compared with the prior art, the beneficial effects of the present invention are:

The invention optimizes the combination of tumor mutation analysis and tumor expression prediction tumor specific antigen, so that the analysis process is more efficient and accurate;

The invention is not only aimed at human gene data, but also adds a mouse analysis module, so that the application range of neoantigen prediction analysis is wider;

The invention starts from reading the fastq file, automatically generates the result with one key, optimizes the joint call of the intermediate file merging in the big data processing, adopts the multi-task distributed processing to greatly improve the analysis efficiency, and reduces the requirement of the biological big data analysis hardware , which makes the tumor mutation analysis results more accurate, further improves the accuracy of subsequent treatment, and has a positive application prospect.

Description of drawings

为3时的曲线图。Fig. 1 is the hypothesis in Example 2 of the screening method for tumor neoantigens

The graph when it is 3.

FIG. 2 is a graph of transcriptome expression levels generated when Expression_TMP is set to 4.4 in Example 2 of the screening method for tumor neoantigens.

FIG. 3 is a graph of the wild-type peptide chain affinity score calculated when the Normal_score is 7.6 in Example 2 of the screening method for tumor neoantigens.

Detailed ways

The technical solution of the present patent will be described in further detail below in conjunction with specific embodiments.

Example 1

A screening method for tumor neoantigens, the specific steps are as follows:

1. Steps of tumor somatic mutation selection

The internationally recognized GATK tumor cell somatic mutation detection software and commercial software were used to conduct DNA point mutations, indel mutations and frameshift mutations in tumor somatic cells from the whole-exon next-generation sequencing results of tumor somatic cell samples and normal blood samples. For detection, mutations with a high mutation frequency detected by various detection software are selected as candidate mutations; at the same time, mutation analysis is performed on the results of tumor somatic cell transcriptome sequencing;

2. Steps of tumor somatic RNA variation analysis

Combining the somatic mutation of tumor somatic DNA and tumor somatic RNA mutation, the tumor somatic mutation is finally determined.

3. MHC molecular analysis steps

The HLA class I molecules of the whole-exon next-generation sequencing of tumor somatic cell samples and normal blood samples were analyzed by HLA typing software Optitype; the HLA typing software xHLA was used to analyze tumor somatic cell samples and normal blood samples, respectively. Whole-exome next-generation sequencing results were analyzed for HLA class I molecules and HLA class II molecules; the HLA typing software seq2HLA was used to analyze the HLA class I molecules and HLA class II molecules of the tumor somatic transcriptome sequencing results; Results The HLA typing results were finally determined, and the consistency of the samples was confirmed from the three results.

4. Tumor Somatic RNA Expression Analysis

Transcriptome expression analysis was performed on the results of tumor somatic transcriptome sequencing to determine genes and transcriptional TPM (Transcripts Per Million) values.

5. Variation annotation steps

Annotation of point mutations, indels, and frameshift mutations in tumor somatic mutations from genome mutation to transcriptome correspondence to amino acid mutations; TMB (tumor burden) analysis of tumor somatic mutations;

6. Steps of tumor somatic driver gene analysis

The tumor somatic mutation-driven gene analysis was performed with reference to the COSMIC tumor database for point mutations, indel mutations, and frameshift mutations in tumor somatic variants.

7. HLA molecular affinity prediction steps

Including the HLA molecular type of the tumor somatic cell sample obtained in the MHC molecular identification step, the mutant predicted peptide obtained in the mutant peptide prediction step, and the wild-type peptide sequence corresponding to the mutant predicted peptide as the MHC type I and MHC type II affinity The input of the prediction software predicts the affinity level of the mutant peptide with the MHC type I and MHC type II genes, respectively. Using computer neural network NNAlign algorithm binding affinity and MS elution ligand data to predict the mutated peptides of MHC class I molecules.

8. Analysis steps of tumor somatic mutation frequency

Including the use of tumor mutation frequency analysis software to detect the frequency of the mutation of tumor somatic cells accounting for this gene locus in all DNA, the higher the mutation frequency, the greater the percentage of tumor somatic cells.

9. Comprehensive scoring and sorting steps for candidate tumor neoantigens

Including the score of each mutation predicted peptide in the candidate tumor neoantigen according to the influence factors such as MHC affinity, antigen expression abundance and wild-type peptide comparison, tumor somatic mutation frequency, whether RNA mutation, whether it is tumor somatic driver gene, etc. , according to the score from high to low, and select the one with the highest score as the tumor neoantigen.

10. Analysis steps of the difficulty of peptide synthesis

With the predicted mutant peptide as the center, the peptide chains with a total length of 30 positions were filled to the left and right. According to the analysis software for the difficulty of peptide synthesis, the molecular weight, isoelectric point, electrostatic charge at pH 7, average hydrophilicity were calculated from the analysis software. And the ratio of hydrophilic residues to analyze the difficulty of tumor candidate neoantigen synthesis.

11. Final selection steps for candidate tumor neoantibodies

The final synthetic tumor neoantigen was selected according to the score in the ninth step and the difficulty of peptide synthesis in the tenth step.

Example 2

A screening method for tumor neoantigens, the specific steps are as follows:

1. Steps of tumor somatic mutation selection

1.1. Use relevant reagents to perform whole-exon next-generation sequencing on the tumor tissue and blood of the sample, and the sequencing depth is 200X and 100X respectively;

1.2. Use OpenGene's fastp software to perform comprehensive quality analysis on the sequencing data in 1.1. If the Q20 is less than 98% or the Q30 is less than 90% or the GC ratio is abnormal, the quality of the sequencing data is considered unqualified, and the neoantigen analysis is stopped. Filters out reads whose quality is too low, too short, or with too many Ns.

1.3. Perform BWA-MEM alignment on fastq clean data. Determine the sample type. If it is a human sample, select the GRCh38 human reference genome to compare the tumor tissue sequencing data and blood sequencing data; if it is a mouse sample, select the GRCm38 reference genome to compare the tumor tissue sequencing data and blood sequencing data. .

1.4. Further statistics on sequencing quality. The Phred score of each sequencing cycle in the tumor tissue and blood data after 1.3 was calculated separately, and the Q value of each sequencing cycle was required to be above 30, otherwise the neoantigen analysis was stopped. Calculate the sequencing depth of tumor tissue and blood data after step 1.3, respectively, and stop the neoantigen analysis if the sequencing depth is lower than 200X and 100X, respectively.

1.5. The degree of repetition of the culling after step 1.3.

1.6. Realign the data after step 1.5 according to the known indel information in 1000G.

1.7. Using the existing mutation database, build a model to generate a recalibration table. The quality scores of the bases are then corrected according to this model.

1.8. Using MuTect, Mutect2 for tumor somatic mutation analysis. Combine the results of the two variants.

2. Analysis of RNA Variation in Tumor Tissue

2.1. Perform RNA-seq on tumor tissue, and the sequencing cluster is 60M

2.2. Use FastQC software to verify the quality of RNA-seq data. If the quality is unqualified, the neoantigen analysis will be stopped.

2.3. Alignment using STAR software. Determine the sample type. If it is a human sample, select the GRCh38 human reference genome to compare the tumor tissue sequencing data with the blood sequencing data. If it is a mouse, select the GRCm38 reference genome to compare the tumor tissue sequencing data with the blood sequencing data.

2.4. Further statistics on sequencing quality. The Phred score of each sequencing cycle in the tumor tissue and blood data after 2.3 was calculated separately, and the Q value of each sequencing cycle was required to be above 30, otherwise the neoantigen analysis was stopped.

2.5. The degree of repetition for the rejection after step 2.3.

2.6. Use the split reads strategy to discover new junctions.

2.7. Realign the data after step 2.6 according to the known indel information in 1000G.

2.8. Using the existing variant database, build a model to generate a recalibration table. The quality scores of the bases are then corrected according to this model.

2.9. Variation analysis using HaplotypeCaller. In the HaplotypeCaller command, the emit_conf parameter is set to 30, the call_conf parameter is set to 25, and the ploidy transfer parameter is set to 4.

3. MHC molecular analysis steps

3.1. Use the HLA typing software Optitype to analyze the HLA class I molecules of the whole-exome next-generation sequencing results of the sample tumor tissue and normal blood respectively;

3.2. Use the HLA typing software xHLA to analyze the HLA class I molecules and HLA class II molecules of the sample tumor tissue and normal blood by whole-exome next-generation sequencing;

3.3. Use the HLA typing software seq2HLA to analyze the HLA class I molecules and HLA class II molecules of the tumor tissue transcriptome sequencing results;

3.4. Combine the results of 3.1, 3.2 and 3.3 to finally determine the HLA typing results. If the difference between the results of the three is very large, the alarm will be withdrawn and the neoantigen analysis will be stopped.

4. Tumor Tissue RNA Expression Analysis

4.1. Compare the tumor tissue transcriptome sequencing results to determine the sample type. If it is a human sample, select the grch38_tran human reference genome to compare the tumor tissue sequencing data and blood sequencing data. If it is a mouse, select the grcm38_tran reference genome to compare the tumor tissue sequencing data with the blood sequencing data.

4.2. Sort the bam file output from 4.1. with samtools.

4.3. Use stringtie to calculate transcriptome expression.

4.4. Extract the TMP (Transcripts Per Million) value of the transcriptome from the gtf file generated in 4.3.

5. Variation annotation steps

5.1. Use VEP software to annotate point mutations, indel mutations, and frameshift mutations in tumor somatic variants. Determine the sample type. If it is a human sample, select the GRCh38 human reference genome to compare the tumor tissue sequencing data and blood sequencing data; if it is a mouse sample, select the GRCm38 reference genome to compare the tumor tissue sequencing data and blood sequencing data. .

5.2. Use vcf2maf software to convert vcf format to maf format.

5.2. Screening tumor somatic mutations to exclude Intron, 5'UTR, 3'UTR, IGR, 5'Flank, 3'Flank, RNA, and lincRNA types of mutations, and non-dbsnp mutations, calculate TMB (tumor load) analysis.

6. Tumor driver gene analysis steps

The tumor somatic mutation-driven gene analysis was performed with reference to the COSMIC tumor database for point mutations, indel mutations, and frameshift mutations in tumor somatic variants.

7. HLA molecular affinity prediction steps

Including the HLA molecular type of tumor samples obtained in the MHC molecular identification step, the mutant predicted peptide obtained in the mutant peptide prediction step, and the wild-type peptide sequence corresponding to the mutant predicted peptide as the MHC type I and MHC type II affinity prediction software input to predict the affinity level of mutant peptides to MHC type I and MHC type II genes, respectively. Using computer neural network NNAlign algorithm binding affinity and MS elution ligand data to predict the mutated peptides of MHC class I molecules.

7.1. Determine the sample type. If it is a human sample, select the cDNA reference sequence of GRCh38 and the peptide sequence of GRCh38; if it is a mouse sample, select the cDNA reference sequence of GRCm38 and the peptide sequence of GRCm38. Using the results of vcf2maf in 5.2, for SNP mutation, insertion and deletion type mutations, the wild-type peptide chain and mutant peptide chain corresponding to the predicted length are searched; for frameshift mutations, the wild-type peptide chain corresponding to the predicted length is searched according to the cDNA sequence and the peptide chain reference sequence. peptide chains and mutant peptide chains.

7.2. Use netMHCpan-4.0 to analyze the HLA-I molecules generated in step 3. Turn on netMHCpan-4.0 and integrate the parameters of binding affinity (BA) and mass spectrometry data (MS) to obtain more information from two different perspectives. Firstly, a class of MHC data in the IEDB database is used to conduct necessary screening. The model training is to use the data of affinity (BA) and mass eluted ligand (MS eluted ligand), and integrate the information of the two data through the method of artificial neural network. Based on the NNAlign framework, the affinity values and lengths of peptides that are predicted to bind to specific MHC molecules are added. The NetMHCpan-4.0 approach improves prediction accuracy in tumor neoantigens, validated eluting ligands (ELs), and T cell immune epitopes. Using netMHCpan-4.0 to predict the affinity of HLA class I molecules with the 8-15-long wild-type peptide chain and the 8-15-long mutant peptide chain generated in 7.1 for prediction and scoring.

7.3. Use netMHCIIpan-3.2 to analyze the HLA-II molecules generated in step 3. Predict the affinity of HLA class I molecules with the 8-15-long wild-type peptide chain and the 8-15-long mutant peptide chain generated in 7.1 for prediction and scoring.

7.4. The HLA-I and class II molecules generated in step 3 were analyzed by netMHC, and the affinity threshold was set at 500 nm. Predict the affinity of HLA class I molecules with the 8-15-long wild-type peptide chain and the 8-15-long mutant peptide chain generated in 7.1 for prediction and scoring.

7.5. The HLA-I and class II molecules generated in step 3 were analyzed by NetMHCcons, and the affinity threshold was set at 500 nm. Predict the affinity of HLA class I molecules with the 8-15-long wild-type peptide chain and the 8-15-long mutant peptide chain generated in 7.1 for prediction and scoring.

7.6. Take the median of the affinity scores for the HLA-I molecules of the above three softwares, and use the Chinese value to score the final affinity.

8. Tumor mutation frequency analysis steps

Including the use of tumor mutation frequency analysis software to detect the frequency of tumor mutation in this gene locus in all DNA, the higher the mutation frequency, the greater the percentage of tumor cells. Read the mutation frequency from the VCF file. If it is the analysis result of Mutect software, read the FA field in the VCF file, and if it is Mutect2, read the AF field in the VCF file.

9. Comprehensive scoring and sorting steps for candidate tumor neoantigens

It includes scoring each mutation predicted peptide in the candidate tumor neoantigen according to the influence factors such as MHC affinity, antigen expression abundance and wild-type peptide comparison, tumor mutation frequency, whether it is RNA mutation, whether it is a tumor driver gene, etc., according to the score Sort from high to low, and select those with higher scores as tumor neoantigens.

9.1. Formula 1

Neoantigen Score =

Calculate the neoantigen scores of the 8-15 peptide chains in 7.6, and sort all neoantigens in reverse order of scores.

9.2. Formula 2

Mutant_affinity_score calculation formula:

Mutant_affinity_score =Δ*(1+e ^{mutant_score*10-5} )

为肿瘤变异数量，如果为snp变异，则

为1；如果为插入或者删除变异，则

为具体的插入或者删除变异数量；如为移码变异则

为具体的移码数量。图1为假设

为3时的曲线。Mutant_score is the affinity score of the mutant peptide chain calculated in 7.6, which is converted into a natural logarithm with the affinity as an index. in

is the number of tumor mutations, if it is a snp mutation, then

is 1; if it is an insertion or deletion mutation, then

is the specific number of insertion or deletion variants; if it is a frameshift variant, then

is the specific number of code shifts. Figure 1 is a hypothetical

curve at 3.

9.3. Formula 3

Expression_score calculation formula:

Expression_score=tanh(expression_TMP)

Figure 2 shows the transcriptome expression level generated in Expression_TMP 4.4. When the transcriptome expression level reaches a certain level, the value of expression_score is 1.

9.4. Equation 4

Normal_affinity_score calculation formula:

Normal_affinity_score =1/(1+e ^normal_*10-5 )

Figure 3 shows the wild-type peptide chain affinity score calculated when the Normal_score is 7.6, which is converted into a natural logarithm with the affinity as an index and its reciprocal is used for calculation.

9.5. Formula 5

α=0.99* allele_frequency +0.9*TBM+0.1* in_RNA_mutant +

0.1* is_cancer_driven_genne

Allele_frequecy: Tumor mutation frequency calculated in 8 steps.

TBM: Tumor burden calculated in 5.2.

In_RNA_mutant: Whether the tumor mutation is in RNA mutation in step 2.9.

Is_cancer_driven_gene: Whether it is a tumor-driven gene in step 6.

10. Analysis steps of the difficulty of peptide synthesis

With the predicted mutant peptide as the center, the peptide chains with a total length of 25-30 positions are respectively filled out. The ease of synthesis of tumor candidate neoantigens was analyzed from the ratio of water and hydrophilic residues.

11. Final selection of candidate tumor neoantibodies

Select the final synthetic tumor neoantigen according to the score in step 9 and the difficulty of peptide synthesis in step 10

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the present invention. within the scope of protection. Any reference signs in the claims shall not be construed as limiting the involved claim.

In addition, it should be understood that although this specification is described in terms of embodiments, not each embodiment only includes an independent technical solution, and this description in the specification is only for the sake of clarity, and those skilled in the art should take the specification as a whole , the technical solutions in each embodiment can also be appropriately combined to form other implementations that can be understood by those skilled in the art.

Claims (7)

1. a screening method for tumor neoantigens, characterized in that the concrete steps are as follows: Step 1, select the variant type of tumor somatic cells; Step 2, performing RNA mutation analysis on tumor somatic cells; Step 3, carry out MHC molecular analysis on normal blood and tumor somatic cells respectively; Step 4: Perform RNA expression analysis on tumor somatic cells; Step 5: Variation annotation on tumor somatic cells; Step 6, analyze the mutation driving genes of tumor somatic cells; Step 7: Predict the binding affinity of HLA molecules on tumor somatic cells; Step 8, analyze the mutation frequency of tumor somatic cells; Step 9, comprehensively score and sort the candidate tumor neoantigens; Step ten, analyze the synthesis difficulty of candidate tumor neoantigens; Step 11: Select the final tumor neoantigen based on the results of Step 9 and Step 10. 2 . The method for screening tumor neoantigens according to claim 1 , wherein the mutation types of tumor somatic cells in the step 1 include DNA point mutations, indel mutations and frameshift mutations of tumor somatic cells. 3 . 3 . The method for screening tumor neoantigens according to claim 1 , wherein the mutation annotation in the step 5 includes the mutation annotation of point mutation in tumor somatic mutation, the mutation annotation of indel mutation and frameshift mutation. 4 . Variation annotations. 4. The method for screening tumor neoantigens according to claim 1 or 2, wherein the step 6 is to analyze point mutations, indel mutations and frameshift mutations in tumor somatic cells for mutation-driven gene analysis . 5 . The method for screening tumor neoantigens according to claim 1 , wherein in the seventh step, the predicted mutant peptides and the predicted mutant peptides obtained in the step of predicting HLA molecules of tumor somatic cells and mutant peptides are obtained. 6 . The corresponding wild-type peptide sequence of the segment was used to predict the binding affinity of HLA molecules to tumor somatic cells. 6. The method for screening tumor neoantigens according to claim 1 or 3, characterized in that, in the step 9, the scoring and ranking are based on the MHC affinity of tumor somatic cells, tumor somatic cell antigen expression abundance and wild-type peptide The degree of contrast, the mutation frequency of tumor somatic cells, whether the tumor somatic cells are RNA mutations, and whether it is a tumor driver gene. 7. The method for screening tumor neoantigens according to claim 1, wherein in step 10, according to molecular weight, isoelectric point, electrostatic charge at pH 7, average hydrophilicity and hydrophilic residues The synthesis difficulty of candidate tumor neoantigens was compared.

Images