A computational method for the analysis and prediction of protein:phosphopeptide‐binding sites

Phosphopeptide‐binding domains, including the FHA, SH2, WW, WD40, MH2, and Polo‐box domains, as well as the 14‐3‐3 proteins, exert control functions in important processes such as cell growth, division, differentiation, and apoptosis. Structures and mechanisms of phosphopeptide binding are generally diverse, revealing few general principles. A computational method for analysis of phosphopeptide‐binding domains was therefore developed to elucidate the physical and chemical nature of phosphopeptide binding, given this lack of structural similarity. The surfaces of nine phosphopeptide‐binding proteins, representing seven distinct classes of phosphopeptide‐binding modules, were discretized, and encoded with information about amino acid identity, surface curvature, and electrostatic potential at every point on the surface in order to identify local surface properties enriched in phosphoresidue contact sites. Cross‐validation indicated that propensities corresponding to this enrichment calculated from a subset of the training data could be used to predict the phosphoresidue contact site on proteins not used in training with no false negative results, and with few unconfirmed positive predictions. The locations of phosphoresidue contact sites were then predicted on the surfaces of the checkpoint kinase Chk1 and the BRCA1 BRCT repeat domain, and these predictions are consistent with recent experimental evidence.