Leveraging Information Across HLA Alleles/Supertypes Improves HLA-Specific Epitope Prediction

We present a model for predicting HLA-specific CTL epitopes. In contrast to almost all other work in this area, we train a single model on epitopes from all HLA alleles and supertypes, yet retain the ability to make epitope predictions for specific HLA alleles. We are therefore able to leverage data across all HLA alleles and/or their supertypes, automatically learning what information should be shared and also how to combine allele-specific, supertype-specific, and global information in a principled way. We show that this leveraging can improve prediction of epitopes having HLA alleles with known supertypes, and dramatically increases our ability to predict epitopes having alleles which do not fall into any of the known supertypes. Our model, which is based on logistic regression, is simple to implement and understand, is solved by finding a single global maximum, yet performs on par with (to our knowledge) the best published results.