Locating the Long-range Contact Centers with B-factors and Hydrophobic Cores

It is well known that contact map is one key to representing a specific protein structure. To simplify the complex problem of the contact map prediction, instead we predict the long-range contact centers centered at groups of neighboring contacts. Here, we propose a novel contact prediction method that combines two SVM predictors; each one contains input information from sequence profile, from evolutionary rate, and from prediction of secondary structure. One SVM predictor relies on the local lowest B-factor (attribute for representing the thermal motion of a given atom in protein), which can directly proportional to a local maximal contact density around one residue and in this case this site also may be a clustering centre of long-range contacts between residues. Another SVM predictor is based on hydrophobic cores, each one of which may be considered as a set of neighboring long-range contacts between residues. Incorporated these two SVM predictors by a fusion technique, the clustering centers of long-range contacts can be accurately predicted.