ImmunoStruct: a multimodal neural network framework for immunogenicity prediction from peptide-MHC sequence, structure, and biochemical properties.

Epitope-based vaccines are promising therapeutic modalities for infectious diseases and cancer, but identifying immunogenic epitopes is challenging. The vast majority of prediction methods only use amino acid sequence information, and do not incorporate wide-scale structure data and biochemical properties across each peptide-MHC. We present ImmunoStruct, a deep-learning model that integrates sequence, structural, and biochemical information to predict multi-allele class-I peptide-MHC immunogenicity. By leveraging a multimodal dataset of ∼27,000 peptide-MHCs, we demonstrate that ImmunoStruct improves immunogenicity prediction performance and interpretability beyond existing methods, across infectious disease epitopes and cancer neoepitopes. We further show strong alignment with assay results for a set of SARS-CoV-2 epitopes, as well as strong performance in peptide-MHC-based cancer patient survival prediction. Overall, this work also presents a new architecture that incorporates equivariant graph processing and multimodal data integration for the long standing task in immunotherapy.