Abstract A116: A predictive model to identify immunogenic neoepitopes in patients with melanoma treated with CTLA-4 blockade

Abstract

Abstract Recent studies have shown that CD8+ T cells that recognize neoantigens presented by tumor cells, but the target specificity of these tumor infiltrating T cells are largely unknown. Developing a predictive model that identifies immunogenic neoepitopes presented by major histocompatibility complex (MHC) class I molecules is critical to identify potential vaccine targets and to improve prediction of response to immune checkpoint inhibitors. We hypothesize that the immunogenic heterogeneity of tumors may explain, in part, differences in clinical response that individual patients derive to checkpoint blockade therapy. We have recently shown that the relative hydrophobicity of T cell receptor (TCR) contact residues is a hallmark of immunogenic MHC class I epitopes, and that this may represent a biochemical basis by which T cells discriminate immunogenic epitopes within the background of self-peptides. Here, we extend this analysis using candidate MHC class I tumor neopeptides from patients with melanoma. Somatic missense mutations were identified from recently published data on whole-exome sequencing of tumors from 11 patients who derived a long-term benefit and 14 patients who derived a minimal or no benefit from ipilimumab or tremelimumab. Mutational burden ranged from 1 to 1,304 mutations per tumor. More than 3,200 potential MHC class I neopeptides were identified based on a predicted binding algorithm (binding affinity, ≤500 nM). Our preliminary results indicate significant differences in hydrophobicity at the anchor residues P2 and P9 (P < 0.001 for both), reflecting MHC binding differences. In addition, there is an increase in hydrophobicity at the TCR contact residues P4 and P7 (P < 0.01 for both) in long-term responders. The overall frequency of mutations in neoepitopes is significantly different at the TCR contact residues P4 and P7 between long-term responders and minimal or no responders (P < 0.01), suggesting genetic immunoediting of these tumors. Based on these results, we have developed a probabilistic Bayesian-based model to predict immunogenic neoepitopes. The model is being applied on the overall pool of neopeptides from each individual patient with melanoma. The sets of predicted immunogenic neoepitopes may serve as potential therapeutic vaccine candidates for patients with melanoma. Citation Format: Diego Chowell, Sri Krishna, Alexandra Snyder, Makarov Vladimir, Timothy A. Chan, Karen S. Anderson. A predictive model to identify immunogenic neoepitopes in patients with melanoma treated with CTLA-4 blockade. [abstract]. In: Proceedings of the CRI-CIMT-EATI-AACR Inaugural International Cancer Immunotherapy Conference: Translating Science into Survival; September 16-19, 2015; New York, NY. Philadelphia (PA): AACR; Cancer Immunol Res 2016;4(1 Suppl):Abstract nr A116.