Tumor Neoantigen Discovery With Proteogenomic Method

In tumor cells mutant genes would be transcribed into mRNA and be translated into corresponding mutant proteins.The abnormal protein sequences are cleaved into short peptides in tumor,then these abnormal peptides are presented on the cell surface by human leukocyte antigen(HLA)molecules,which are recognized by T cell receptor(TCR)as foreign antigens.Tumor-specific antigens could be differentially recognized by TCR as these sequences are unique to the tumor,named neoantigens.Based on highthroughput tumor genomic analysis,each missense mutation can potentially give rise to multiple neopeptides,resulting in a vast total number,only a few peptides can be recognized by TCR and elicit immune response.Specific identification of immunogenic candidate neoantigens is consequently a major challenge.In recent years,technologies in genomics and proteomics have been significantly improved,in the meantime some supportive bioinformatics and in silico HLA-binding prediction tools have been developed.However,previous HLA-binding prediction tools only predict neoantigens with genomic and transcriptome data,without considering proteomics data,which can be beneficial to high quality identification of neoantigens.With these considerations in mind,here we took state of the art key bioinformatics tools and screening methods int one workflow,which can greatly cut down the numbers of reliable neoantigen candidates and be beneficial to preclinical neoantigen studies.The workflow was constructed based on the genomics and proteomics data of Jurkat leukemia cell line.Firstly,based on transcriptome and genome data processing and annotating variants,we identified 9,817 missense mutations.Then a total of 36,835 HLA class Irestricted candidate neoantigens were predicted by NetMHCpan.655 candidate neoantigens were further filtered considering transcripts per million(TPM)to select only mutant peptides arising from expressed genes,and based on constructed mutant peptidome database and MaxQuant filtering software.These candidate neoantigens were evaluated using sequence similarities with the cross-reactive microbial peptides from IEDB that measure the ability of CD8+T cells to recognize candidate neoantigens.Our results showed that 313 neoantigens may be most likely identified by TCR.The workflow can greatly reduce the burden of experimental verification.To test this pipeline,we analyzed the high-throughput sequencing and mass spectrometry data in Mono-allelic Cells.First,we generated a customized personalized reference database based on RNAseq data.Then,we searched the raw MS data against this database,directly identified 9 peptide ligands harbouring mutations from eluted HLA class I peptides assigned to HLAB5701.In this study,the workflow was implemented in a software package for further potential application on personalized proteogenomic discovery of tumor neoantigens,named “ProGeo-neo”(https://github.com/kbvstmd/ProGeo-neo),which is currently the first integrated computational pipeline to predict and select neoantigens by taking proteomics data into consideration.The pipeline was constructed based on the genomics and proteomics data of Jurkat leukemia cell line but also suitable for predicting neoantigens in other solid tumors.Our proteogenomics workflow will likely become increasingly useful for cancer research.