Prediction of cancer neoepitopes needs new rules.

Tumors are immunogenic and the non-synonymous point mutations harbored by tumors are a source of their immunogenicity. Immunologists have long been enamored by the idea of synthetic peptides corresponding to mutated epitopes (neoepitopes) as specific "vaccines" against tumors presenting those neoepitopes in context of MHC I. Tumors may harbor hundreds of point mutations and it would require effective prediction algorithms to identify candidate neoepitopes capable of eliciting potent tumor-specific CD8+ T cell responses. Our current understanding of MHC I-restricted epitopes come from the observance of CD8+ T cell responses against viral (vaccinia, lymphocytic choriomeningitis etc.) and model (chicken ovalbumin, hen egg lysozyme etc.) antigens. Measurable CD8+ T cell responses elicited by model or viral antigens are always directed against epitopes possessing strong binding affinity for the restricting MHC I alleles. Immense collective effort to develop methodologies combining genomic sequencing, bioinformatics and traditional immunological techniques to identify neoepitopes with strong binding affinity to MHC I has only yielded inaccurate prediction algorithms. Additionally, new evidence has emerged suggesting that neoepitopes, which unlike the epitopes of viral or model antigens have closely resembling wild-type counterparts, may not necessarily demonstrate strong affinity to MHC I. Our bearing need recalibration.