The presenting HLA determines fidelity of SARS-CoV-2 spike protein epitope prediction.

Early in the COVID-19 pandemic, multiple studies used prediction methods to identify candidate peptides to be included in prospective vaccines. While subsequent studies identified epitopes from convalescent and vaccinated subjects, few studies have compared the predicted to identified epitopes. Here we used three methods to predict SARS-CoV-2 spike protein helper T cell epitopes and compared the results to experimentally determined peptide binding as well as published epitopes. The correspondence between the predicted and experimental binding results and published epitopes depended more on the HLA being investigated than the prediction method used. Lastly, these observations were used to predict peptides which bind to the most HLAs. These peptides were previously identified and predicted to maintain HLA binding in the current variants of interest. This study highlights which prediction methods and conditions lead to the most reliable prediction results which would be of great interest for improving the design of future vaccines.