Toward prediction of binding affinities between the MHC protein and its peptide ligands using quantitative structure-affinity relationship approach.

It is important but challenging to determine the binding specificity of MHC-peptide interactions accurately and to predict their binding affinity quantitatively. In this paper, we discuss the application of an effective amino acid descriptor to model and predict the binding affinities between the MHC protein and its peptide ligands. This amino acid descriptor was derived from 23 electronic properties, 37 steric properties, 54 hydrophobic properties and 5 hydrogen bond properties of coded amino acids using principal component analysis (PCA), called the divided physicochemical property scores (DPPS). The DPPS descriptor was used to characterize a set of mouse MHC (H-2K(K)) binding peptides, and genetic algorithm-partial least square (GA-PLS) models were then constructed. In analyses, these models were statistically consistent with previous reports and molecular graphics exhibition. Hydrophobic interactions and hydrogen bonds were important to antigen recognition and presentation, especially exerting effects on anchor residues of peptides.