“…Despite advances in neoantigen characterization, it has been difficult to use neoantigens as predictive biomarkers of immunotherapy response due to several reasons: a) Neoantigen profiles are mostly unique to each patient possibly because the majority of neoantigens originate from passenger mutations [14,16,17,34], b) Neoantigen repertoires might evolve dynamically over time due to genomic instability of the tumors and selection pressures exerted by the immune system; Verdegaal et al for example have shown loss of immunogenic neoantigens in sequential lesions obtained from two melanoma patients [34], c), The accuracy of computational models utilized to construct neoantigen profiles have not been broadly tested; these models may need to be specifically trained on large datasets of tumor isolated neoantigens that are experimentally determined by techniques such as mass spectrometry [35], and d) The neoantigens, albeit predicted in high numbers, might not be visible to the immune system due to disruptions in the antigen processing/presentation machinery. Giannakis et al , for example have shown that colorectal tumors with high neoantigen burden and high immune infiltration (hence favorable anti-tumor immune environment) also harbored immune-evasive mutations, such as those at certain HLA allelic loci and B2M [36].…”