Pan-prediction of major histocompatibility complex class II–restricted epitopes across species via an AlphaFold-based quantification scheme

Predicting major histocompatibility complex class II (MHC-II)–restricted epitopes is challenging; however, AlphaFold (AF) may provide a structure-based pan-prediction solution. In this study, we have established the new tool AF-prediction (AF-pred) with a clear standard for quantitative prediction results of MHC-II–restricted epitopes. We validated AF-pred using MHC-II molecules from humans, pigs, cattle, and bats, with this validation including large-scale in silico analyses with known immunopeptidome datasets, as well as in vitro binding assays and crystallographic characterization of newly predicted epitopes. When compared to sequence-based tools heavily trained with the human immunopeptidome, AF-pred demonstrates advantages in the prediction of cross-species MHC-II binding patterns. Using unreported bat MHC-II structures, we analyzed the prediction capability, logic, and limitations of AF-pred. In addition, we explored the impact of the AF algorithm iterations on the prediction of MHC-II–restricted epitopes. These results demonstrate that AF-pred is capable of cross-species prediction of MHC-II–restricted epitopes, which is conducive to the development of novel T-cell epitope vaccines and the advancement of the “One Health” initiative. • AlphaFold prediction (AF-pred) can predict major histocompatibility complex (MHC) class II-restricted epitopes. • The structure-based tool AF-pred exhibits distinct advantages in cross-species prediction. • AF-pred supporting T-cell epitope vaccine development and "One Health" implementation.