This new machine learning based prediction tool can accurately model the binding strength between highly polymorphic natural killer (NK) cell KIR receptors and their equally diverse HLA class I ligands, incorporating not only the receptor and HLA sequences, but also the modulating peptide presented between them.