A meta-analysis of experimentally validated neo-epitopes: patterns, biases, and opportunities.

Cancer cells harbor somatic mutations that generate novel amino acid sequences that are absent in the self-proteome. These mutation-derived cancer-specific peptides are defined as "neo-peptides". Neo-peptides eliciting immune responses, i.e. immunogenic neo-peptides, are defined as "neo-epitopes". Given their relevance to cancer immunotherapy, we conducted a meta-analysis to examine how experimental evidence informs our understanding of neo-epitopes. Our study is the largest reported to date. Using the cancer epitope database and analysis resource (CEDAR), we analyzed over 16,000 neo-peptides tested in more than 20,000 T cell assays across 180 studies. We found that validated neo-epitope frequencies varied across cancer types, with the highest rates in skin and lung and the lowest in colorectal cancer. Neo-epitopes were enriched in driver genes such as TP53 and KRAS. However, testing frequency correlated with mutation prevalence, revealing bias toward recurrent mutations. Despite the high sequence similarity among RAS family members, validated neo-epitope overlap was minimal, challenging pan-RAS strategies. Shared neo-epitopes across cancer types are rare, with only 16 validated in more than one cancer type. While most assays involved HLA class I, class II alleles presented a higher proportion of validated neo-epitopes. Specific alleles, including HLA-B*40:01 and HLA-DRB1*11:01, were enriched for neo-epitopes, whereas others, like HLA-A*02:01, were enriched for non-immunogenic neo-peptides. Finally, amino acid substitutions that altered hydrophobicity or charge were more common in neo-epitopes. Together, these findings define key features of neo-epitopes, expose methodological and biological biases in the literature, and highlight opportunities to improve the selection and prioritization of neo-epitopes for cancer immunotherapy.