Abstract
Cancer immunotherapy is a promising avenue for treating patients who exhaust their current standard lines of care. These therapies work by promoting anti-cancer adaptive immunity. The process of adaptive immunity relies largely upon T cell-mediated recognition of tumor specific or associated peptides presented upon the Major Histocompatibility Complex (MHC). Given longstanding limitations in profiling MHC presented peptides, several computational tools have been developed to instead predict peptide MHC binding. Here, we leverage recent technical advances in proteomics to generate deep immunopeptidomics data from 20 cancer, including 14 glioblastoma, tissue samples and utilize this data to identify tumor antigens. Immunopeptidomics profiling was performed on a Bruker timsTOF Ultra mass spectrometer platform, yielding an average of 19,283 peptides/sample, and a total of 106,079 unique peptides from 12,209 genes across the cohort. Peptides from multiple sources can drive immunogenic recognition, including tumor associated antigens (TAAs), mutated tumor specific antigens (mTSAs), and aberrantly expressed tumor specific antigens (aeTSAs). Here, we built a workflow for personalized immunopeptidomics that incorporates germline variation, variant phasing, and clonality to perform searches for mTSAs, TAAs, and aeTSAs. Initial analyses applied to samples from glioblastoma patients revealed several TAAs abundantly present on patient specific MHC alleles. Next, this work will be applied to aeTSAs and mTSAs to identify a candidate set of high-quality targets for immunotherapy. In addition, this data will be used to comprehensively assess the accuracy of peptide MHC binding prediction tools run on the matched transcriptomes and genomes in this cohort.