Abstract 6268: JANUS: A unified deep learning framework for predicting peptide binding to MHC I and II alleles

The accurate prediction of peptide binding to major histocompatibility complex (MHC) molecules is crucial for advancing cancer immunotherapy, particularly in the design of personalized cancer vaccines and in identifying antigens which may be predictive of immunotherapy-related adverse events. Existing computational models for prediction of peptide binding are generally restricted to a single MHC class. We hypothesize that a unified model capable of predicting peptide binding to both MHC I or MHC II alleles can lead to a synergistic improvement in model performance and thereby facilitate advances in cancer neoantigen discovery. Challenges to this approach include finding a universal encoding for both MHC I and MHC II candidate peptide binders as well as addressing data imbalances arising from training with disparate datasets. Using publicly available datasets including from the Immune Epitope Database (IEDB), we developed a novel deep learning framework JANUS (Joint Allele-specific Neural prediction of MHC I/II-binding Universal Sequences) that leverages biological sequence embeddings from a protein large language model and adapts our loss function to resolve data imbalance issues. The model utilizes a feed-forward multi-layer perceptron with dropout layers for regularization. Ultimately, this model explores whether a unified approach demonstrates competitive performance compared to current single class models, particularly regarding increased predictive accuracy for rare or infrequent alleles, or in scenarios with sparse data. Citation Format: Alan Perez-Rathke, Justin Balko, Jens Meiler. JANUS: A unified deep learning framework for predicting peptide binding to MHC I and II alleles [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 6268.