DinoMHC: An interpretable sequence-geometric framework improves MHC-I presentation prediction and transfers CD8+ T-cell immunogenicity

These are all training and benchmark datasets used in DinoMHC: MS.zip: Contains TRAIN.csv, TEST.csv, TEST_NON_OVERLAP.csv, and a folds/ directory containing the files for 3-fold cross-validation. IM.zip: Contains CD8T/, which includes the CD8+ T-cell epitope dataset, and MUT/, which contains the mutation dataset.