Benchmarking the Foreign Antigen Space of Human Malignancies

Mutational load varies widely between malignancies and has been used as a proxy for the immunological foreignness of human cancers. However, without well-defined reference points it is difficult to determine which human tumors can be considered sufficiently foreign to the T-cell-based immune system. We established a neo-antigen prediction pipeline that processes single nucleotide variants, indels and gene fusion events and established its precision in identifying T-cell-recognized antigens. We used this pipeline to benchmark immunological foreignness of human cancers against pathogens for which T-cell control has been established. We demonstrate that up to 50% of tumors, spanning 25 sites of origin, are more foreign than these pathogen benchmarks. In addition, we demonstrate that the neo-antigen repertoire of treatment-naive tumors is not detectably influenced by immune editing. These data suggest that immunotherapeutic strategies that enhance activity of the endogenous T-cell compartment may be of value for a large fraction of human cancers.