Neoantigen clonal architecture influences the efficacy of immunotherapy. The longitudinal dynamics of clonal neoantigen immunodominance over a canvas of heterogeneous subclonal neoantigens is critical to response durability, but challenging to characterize clinically and experimentally. We developed a computational model of neoantigen evolution that longitudinally tracks the "presentome" - that is, a cell's aggregate neoantigen repertoire and HLA status - of simulated non-small cell lung cancer (NSCLC) prior to and during treatment with checkpoint blockade. Using well-established neoantigen immunogenicity parameters, we first recapitulated the clinically observed presentomes of early-stage NSCLC tumors. After validating simulated neoantigen burden in an independent cohort, we simulated response to PD-1 blockade in silico. In addition to recapitulating known phenomena, such as the correlation of pre-treatment clonal neoantigen burden with response to PD-1 blockade, we found several additional salient features of response, including a profound clonal collapse among completely responding tumors and an insensitivity of response to increased oligoclonality of anti-neoantigen-presenting cell cytotoxicity. Our work can be adapted to longitudinally model the neoantigen architecture of other cancers, modalities of immunotherapy, and preclinical tumor models, ultimately informing cognizant strategies for future therapeutic development.