MultiMHC: a multiscale learning framework integrating sequence and residue contact topology for peptide-MHC binding prediction.

MultiMHC predicts peptide–MHC binding by fusing sequence context with residue contact topology, achieving state-of-the-art–level performance and particularly strong ranking-based results.

• Why it matters: Many pMHC predictors lean mainly on sequence features and only partially reflect structural determinants of recognition, limiting upstream identification of T-cell epitopes for immunotherapy.
• What they did: MultiMHC uses a recurrent neural network for global sequence context and a graph-based residue contact representation from known pMHC interfaces, then fuses both to score peptides of varying lengths.
• The result: On the Tumor Neoantigen Selection Alliance dataset, MultiMHC binding scores enrich experimentally validated T-cell epitopes among top-ranked candidates, supporting use as a binding-based filtering step.

AI summary of the abstract · may contain errors · not medical advice · read the paper.