Antigen Quality: Dissimilarity To The Self-Proteome As A Novel Determinant Of Immunogenicity

Neoantigens are increasingly appreciated as prognostic biomarkers and direct targets for cancer immunotherapy. Current analysis methods predict neoantigens by identifying protein coding tumor mutations and using algorithms to identify mutant peptides with predicted binding affinity to a patient’s MHC alleles. This affinity-centric approach has had limited success; neoantigens predicted this way have inconsistent correlation with clinical benefit from immunotherapy, and the rate of immunogenicity when empirically assayed is less than 2%. Recently, additional properties orthogonal to MHC affinity have been proposed as metrics of neoantigen quality that may correlate with immunogenicity and responses to immunotherapy. These quality metrics apply rational immunological filters to bulk neoantigen predictions to further enrich for immunogenic peptides. To better characterize clinically relevant neoantigens, I developed an open-source neoantigen analysis package in R, antigen.garnish, including an ensemble MHC affinity prediction approach that improved precision. I also assessed neoantigen quality metrics such as differential agretopicity, viral homology, and the novel metric of dissimilarity to the self-proteome to identify the most likely immunogenic peptides from clinical data, and found that these quality metrics correlated with clinical outcomes when bulk neoantigen quantity did not. Furthermore, the novel dissimilarity to the self-proteome metric was enriched amongst immunogenic peptides from known pathogen sequences, autoantigen sequences, and neoantigens from three sources: historical literature review, tandem-minigene screening, and vaccine clinical trials. To alter the tumor-immune interface of immunologically “cold” pancreatic ductal adenocarcinoma, I pharmacologically increased the mutational burden of a murine cell line, which rendered tumor growth sensitive to endogenous T cells. Using this software package, I was able to identify novel acquired high quality neoantigens that could be therapeutically exploited with vaccines or cellular therapies in this tumor type where immunotherapy has been unsuccessful. Neoantigen quality analysis, as implemented in the publically available antigen.garnish repository on GitHub, is therefore an orthogonal approach to MHC affinity for predicting CD8 T cell immunogenicity in antigens from all sources. Incorporating neoantigen quality analysis into target selection for vaccines and cellular therapies in the future may accelerate their development.