Pr14 Emergence Of Quality Metrics For Neoantigens: Dissimilarity To The Self-Proteome As A Novel Determinant Of Immunogenicity.

Abstract Fewer than 2% of computationally predicted neoantigens are immunogenic, a proven barrier to the efficacy of neoantigen vaccines. Currently, strong MHC affinity is dominant among neoantigen selection criteria for targeted therapies. To enrich for neoantigens with the greatest likelihood of immunogenicity, we developed an analytic method to parse neoantigen quality through rational biologic criteria and applied this method across five clinical datasets for 318 cancer patients. We assessed absolute MHC affinity, pathogen homology, and a novel analysis of dissimilarity to the self-proteome. We hypothesized that sequences with high dissimilarity to wild-type sequences would be more likely to perceived as “non-self” by immunosurveilling T cells. In a validation dataset, dissimilarity enriched for immunogenic peptides with an odds ratio of 34.9 (95% CI: 30.9 to 39.5, p < 0.001). Pathogen homology enriched 3.9-fold (95% CI: 3.6 to 4.3, p < 0.001), whereas selection on the basis of strong MHC-binding affinity alone enriches only 1.4-fold (95% CI: 1.2 to 1.5, p < 0.001). In patient tumors, these newly classified high-dissimilarity neoantigens were rare (1.18% of predicted MHC binders), did not overlap with the other metrics of neoantigen quality, and were enriched for hydrophobic sequences. While conventionally predicted neoantigens did not correlate with survival in two trials of PD-1 checkpoint therapy in patients with non-small cell lung cancer, high-dissimilarity neoantigens, even independent of MHC affinity, were predictive (HR - 3.29, 95% CI: 1.50 – 7.22, p = 0.014 and 3.40, 95% CI: 1.37 – 8.47, p = 0.021). Furthermore, we found that dissimilar peptides were enriched among known immunogenic neoantigens reported in multiple independent settings, including vaccine clinical trials (p = 0.001 – 0.023). This enrichment was not explained by a correlation with reported MHC binding affinities. Dissimilarity also enriched for immunogenicity in a dataset of human autoantigens, tumor-associated antigens, and cancer-testis antigens (odds ratio: 10.44, p < 2.2e-16). Dissimilarity to the self-proteome is therefore an orthogonal approach to MHC affinity for predicting CD8 T cell immunogenicity in both neoantigens and self-antigens. Overall, we show that neoantigen quality, including our novel analysis of dissimilarity to the self-proteome, identifies immunogenic peptides from conventionally predicted neoantigens for which the immunogenic “hit rate” is currently low and facilitates the more rapid development of neoantigen vaccines for cancer immunotherapy. Our pipeline, antigen.garnish, is open source and available on GitHub. This abstract is also being presented as Poster B30. Citation Format: Lee P. Richman, Andrew J. Rech, Robert H. Vonderheide. Emergence of quality metrics for neoantigens: Dissimilarity to the self-proteome as a novel determinant of immunogenicity [abstract]. In: Proceedings of the AACR Special Conference on Tumor Immunology and Immunotherapy; 2019 Nov 17-20; Boston, MA. Philadelphia (PA): AACR; Cancer Immunol Res 2020;8(3 Suppl):Abstract nr PR14.