Somatic mutations in cancer cells lead to the production of neoantigens: patient- and tumor-specific peptides that are capable of inducing T cell recognition. Recent clinical trials have established that, when introduced in a vaccine, these neoantigens can stimulate anti-tumor immune responses. The path to producing such a personalized vaccine begins with sequencing a patient?s tumor, identifying candidate somatic mutations and then computationally predicting which neoepitopes will be most effective at stimulating a T-cell response. This prediction step should ideally assess a complex interplay of factors, including the type of somatic mutation, the patient?s class I and II HLA alleles, peptide processing, peptide transport, peptide-MHC binding and many co- factors of immune recognition and signaling. The best current approaches focus almost entirely on a single factor (peptide-MHC binding) and have only a 16-43% success rate in predicting immunogenic peptides. To address this challenge we will develop pVACtools, an informatics toolkit for comprehensive identification, characterization, and clinical application of neoantigens. This tool will be the first to support all major neoepitope sources including insertions, deletions, transcript isoforms, gene fusions, peptides from normally non-coding regions, and B cell or T cell rearrangements (BCRs/TCRs). We will also integrate analysis of Class I and II peptide-MHC binding. All tools will be developed to support foundational pre-clinical work in animal models of immunotherapy. Furthermore, we will test several specific hypotheses relating to new predictors of immunogenicity. To elucidate these factors and enhance prioritization of neoantigens we will create the first open access database of experimentally and clinically validated neoantigens. Using these data we will address the question of what peptide-intrinsic and patient-specific features determine the therapeutic potential of a neoantigen. To validate their translational potential, we will apply our neoantigen tools to clinical trials involving checkpoint blockade drugs and personalized cancer vaccines. We will develop a visualization interface that facilitates clinical review and selection of neoantigen candidates for several vaccine delivery platforms. These tools will be used to perform analysis of >200 cases from ongoing vaccine trials to evaluate their performance and address key outstanding immunobiology questions including: (a) the importance of particular neoantigen sources in specific cancer types, (b) the importance of accurately determining HLA mutation/expression, (c) the significance of having both MHC class I and II restricted peptides in a vaccine, (d) how to identify specific neoepitope/TCR pairings, and (e) how neoantigens contribute to mechanisms of resistance to immunotherapies. These tools will thus enable fundamental studies of T cell biology, lead to more effective personalized cancer vaccine designs, and support better prediction of response to checkpoint blockade therapy. Finally, based on these experiences and in collaboration with our team of clinical vaccine trial leaders, we will develop detailed guidelines and training materials for neoantigen analysis.
Neoantigens are tumor specific antigens arising from somatic mutations that, when identified accurately, can be high value targets for cancer immunotherapy. pVACtools is an open source, open access suite of informatics tools that enables identification, prioritization, and clinical application of neoantigens. We propose to significantly extend and adapt pVACtools to support all major sources of neoantigens, integrate analysis of MHC Class I and II binding, discover new factors that influence immunogenicity, and use this knowledge to better prioritize neoantigen candidates. Successful execution will allow a broad constituency of users to answer pressing cancer immunology questions and design more effective personalized cancer vaccines.
