A structure-based algorithm to predict potential binding peptides to MHC molecules with hydrophobic binding pockets

Binding of peptides to MHC class I molecules is a prerequisite for their recognition by cytotoxic T cells. Consequently, identification of peptides that will bind to a given MHC molecule must constitute a central part of any algorithm for prediction of T-cell antigenic peptides based on the amino acid sequence of the protein. Binding motifs, defined by anchor positions only, have proven to be insufficient to ensure binding, suggesting that other positions along the peptide sequence also affect peptide-MHC interaction. The second phase of prediction schemes therefore take into account the effect of all positions along the peptide sequence, and are based on position- dependent-coefficients that are used in the calculation of a peptide score. These coefficients can be extracted from a large ensemble of binding sequences that were tested experimentally, or derived from structural considerations, as in the algorithm developed by us recently. This algorithm uses the coordinates of solved complexes to evaluate the interactions of peptide amino acids with MHC contact residues, and results in a peptide score that reflects its binding energy. Here we present our analysis for peptide binding to four MHC alleles (HLA-A2, HLA-A68, HLA-B27 and H-2K^b), and compare the predictions of the algorithm to experimental binding data. The algorithm performs successfully in predicting peptide binding to MHC molecules with hydrophobic binding pockets but not when MHC molecules with hydrophilic, charged pockets are considered. For MHC molecules with hydrophobic pockets it is demonstrated how the algorithm succeeds in distinguishing binding from non-binding peptides, and in high ranking of immunogenic peptides within all overlapping same-length peptides spanning their respective protein sequences. The latter property of the algorithm makes it a useful tool in the rational design of peptide vaccines aimed at T- cell immunity.