Abstract B080: Improved neoantigen vaccine selection by combining prediction of pMHC presentation and T-cell epitopes

Abstract

Abstract The OpenVax group has helped initiate two neoantigen vaccine clinical trials (NCT02721043, NCT02721043) at Mount Sinai based on a simple multiplicative ranking criterion which assigns equal weight to expression and predicted Class I MHC binding affinity of mutated peptides (1). This poster seeks to better ground our ranking method for selecting the contents of neoantigen vaccines in several sources of immunological data. We built a better model of MHC-I presentation on the cell surface by relating RNA expression and MHC affinity to pMHC ligands identified with mass spectrometry (2). Secondly, we trained a model of overall T-cell immunogenicity whose primary input is the predicted pMHC presentation score of any peptide-MHC combination, alongside other features such as similarity to the self proteome. This model is trained on T-cell response data deposited in the Immune Epitope Database (3). Lastly, we assembled a small dataset of peptide sequences used in neoantigen vaccine trials (1,4,5), which are labeled by whether they achieved a CD8+ or CD4+ T-cell response. This dataset allows us to explore several hypotheses about the relationship between immunogenic response and sequence similarity to both the self proteome and pathogenic proteomes.