Computational prediction of neoantigens: do we need more data or new approaches?

The capability of the immune system for self-recognition has recently become the focus of interest for a broadening research community due to the impressive clinical results achieved with immune checkpoint inhibitors and other immunotherapies in the treatment of advanced metastatic carcinoma [1]. A significant portion of this clinical efficacy is attributed to the fact that the genomic instability of cancer generates mutation-derived peptides (neopeptides) that are not present in normal cells. A subset of these neopeptides may be neoantigens (also called neoepitopes) which are recognized as alien by cytotoxic T cells. Ideally, a cell producing neoantigens should be eliminated by the immune system. However, tumors are able to inhibit this immune response by activating various checkpoint mechanisms; these can be overcome by the now widely used therapeutic agents of checkpoint inhibitors such as anti PD-L1 or anti-CTLA-4 antibodies. There is increasing evidence that such neoantigen-driven immune responses are responsible for the significant clinical response shown to immune checkpoint inhibitors in at least one specific type of tumors, microsatellite instable cancer [2, 3].