Data from High-Throughput Prediction of MHC Class I and II Neoantigens with MHCnuggets

• DOI: 10.1158/2326-6066.c.6550177
• Publication Date: 2023-04-05T02:22:34Z
• Authors: Xiaoshan M. Shao; Rohit Bhattacharya; Justin Huang; I.K. Ashok Sivakumar; Collin Tokheim; Lily Zheng; Dylan Hirsch; Benjamin Kaminow; Ashton Omdahl; Maria Bonsack; Angelika B. Riemer; Victor E. Velculescu; Valsamo Anagnostou; Kymberleigh A. Pagel; Rachel Karchin

<div>Abstract<p>Computational prediction of binding between neoantigen peptides and major histocompatibility complex (MHC) proteins can be used to predict patient response cancer immunotherapy. Current predictors focus on <i>in silico</i> estimation MHC affinity are limited by low predictive value for actual peptide presentation, inadequate support rare alleles, poor scalability high-throughput data sets. To address these limitations, we developed MHCnuggets, a deep neural network method that predicts peptide–MHC binding. MHCnuggets common or alleles class I II with single architecture. Using long short-term memory (LSTM), accepts variable length is faster than other methods. When compared methods integrate MHC-bound (HLAp) from mass spectrometry, yields 4-fold increase in positive independent HLAp data. We applied 26 types The Cancer Genome Atlas, processing 26.3 million allele–peptide comparisons under 2.3 hours, yielding 101,326 unique predicted immunogenic missense mutations (IMM). Predicted IMM hotspots occurred 38 genes, including 24 driver genes. load was significantly associated increased immune cell infiltration (<i>P</i> < 2 × 10<sup>−16</sup>), CD8<sup>+</sup> T cells. Only 0.16% IMMs were observed more patients, 61.7% derived mutations. Thus, describe its performance characteristics demonstrate utility sets representing multiple human cancers.</p></div>