The ability to predict B cell epitopes from antigen sequences is critical for biomedical research and many clinical applications. However, despite substantial efforts over the past 20 years, the performance of even the best B cell epitope prediction software is still modest. Based on the idea of T-B reciprocity, BepiTBR is a B cell epitope prediction model that demonstrates improved performance by incorporating prediction of nearby CD4+ T cell epitopes close to the B cell epitopes.