Deciphering HLA motifs across HLA peptidomes correctly predicts neo-antigens and identifies allostery in HLA specificity

摘要: The precise identification of Human Leukocyte Antigen class I (HLA-I) binding motifs plays a central role in our ability to understand and predict (neo-)antigen presentation infectious diseases cancer. Here, by exploiting co-occurrence HLA-I alleles across ten newly generated as well forty publicly available in-depth HLA peptidomics datasets, we show that can rapidly accurately identify map them their corresponding without any priori knowledge specificity. Our novel approach uncovers new for several up now had no known ligands. HLA-ligand predictors trained on such data substantially improve neo-antigen predictions four melanoma two lung cancer patients, indicating unbiased are ideal silico (neo-)antigens. further reveal allosteric modulation specificity unravel the underlying mechanisms protein structure analysis, mutagenesis vitro assays.