Abstract B75: Entire landscape of epitopes from all possible missense mutations in human coding sequences

Abstract

Abstract Recent advancements in cancer immunotherapy and next-generation sequencing (NGS) technologies have created promising opportunities for precision cancer medicine. Immune checkpoint blockade agents that enhance the ability of endogenous T cells to destroy cancer cells have demonstrated efficacy in a variety of human malignancies and underscore the importance of the endogenous immune system in controlling tumor growth in at least a large proportion of patients. Antigen-driven processes, in particular, are believed to be crucial to anticancer immune responses. Certain somatic mutations in cancer genomes result in amino acid alterations, which can generate, in turn, tumor-specific immunogenic peptides called neoantigens. The presentation by the MHC molecule is one of the main factors to determine whether mutated peptides can elicit immune response or not. Therefore, it will give us useful insight to explore the immunogenicity of all possible mutated peptides derived from coding sequences. To address this, we made a list of all mutated peptides and estimated their binding affinities to the 86 MHC class 1 molecules and >5,000 MHC class 2 genotypes using NetMHC 4.0 and NetMHCII 3.2, respectively. All possible nucleotide changes at every single position are introduced in all consensus coding sequences (CCDS). In total, there are ~159 million possible nucleotide changes in ~31,000 autosomal transcripts from ~ 19,000 genes, which comprise about 53 M nucleotides. 123 M (~77%) out 159 M alteration are missense mutations. More than 500,000 distinctive mutated peptides in length of 9 and 15 amino acids were derived from these missense mutations. Machine learning algorithms, called NetMHC 4.0 and NetMHCII 3.2, estimated the binding affinity of all mutated peptides that are not found in wild type to all MHC molecules, respectively. Based on the predicted binding affinity, 0.6% ~ 1.0% of them are strong binder to one of the MHC molecules while 1.7% ~ 2.5% of them are weak binder to one of the MHC class I molecules and similar trend was observed in class 2 as well. HLA-A0201 has the most epitopes binding. For example, KRAS can generate collectively epitopes that bind to 60 different MHC molecules including KRAS D12V, which is very hot spot, that make epitopes that binds to 30 MHC molecules. We made a portal called “Immune Explorer” at http://genomeportal.stanford.edu/immune. This portal enables users to explore the potential immunogenic epitopes from genes of their interest such as KRAS. Users can search by HLA genotypes as well. By this way, it is easy to obtain the list of mutations that bind to the HLA of their interest. In addition, we mapped the frequencies of missense mutations observed in TCGA and COSMIC. This web interface allows researchers to address important questions regarding their immunotherapy or cancer vaccine. Citation Format: HoJoon Lee, Stephanie Greer, Hanlee p Ji. Entire landscape of epitopes from all possible missense mutations in human coding sequences [abstract]. In: Proceedings of the AACR Special Conference on Tumor Immunology and Immunotherapy; 2018 Nov 27-30; Miami Beach, FL. Philadelphia (PA): AACR; Cancer Immunol Res 2020;8(4 Suppl):Abstract nr B75.