32. pVACsplice: A Computational tool for predicting and prioritizing alternative splicing neoantigens

Abstract

Splicing neoantigens represent a rich yet underexplored source of tumor-specific targets for immunotherapy. Tumors exhibit increased mis-splicing events compared to normal tissues, which in turn create diverse isoforms that encode novel peptides. These peptides, especially ones derived from frameshifts, are highly distinct from self-antigens, hence presenting an opportunity for enhanced immune recognition.

Though neoantigens arising from somatic single-point mutations in coding regions have been widely targeted by cancer therapies, other neoantigen sources, including alternative splicing neoantigens haven't received the same amount of attention. Here, we develop pVACsplice, a tool that predicts and prioritizes cis-splicing associated neoantigen candidates. pVACsplice takes alternative transcripts as input, translates them into altered peptides, then constructs neoantigens of user-defined sizes. It then estimates binding affinities of neoepitopes with user-input MHC alleles, and prioritizes candidates based on various criteria (binding affinity, solubility, transcript quality, and more).

We then utilize pVACsplice to explore the splicing neoantigen landscape of a Small Cell Lung Cancer (SCLC) cohort. We find numerous neojunctions and neoantigen candidates associated with genes frequently mutated in this malignancy.