Align Protein Surface Structures to Identify Evolutionally and Structurally Conserved Residues

Protein-protein interface underlies the protein protein interaction. Alanine mutation of protein-protein interface residues has shown that the distribution of binding free energy is not average among the interface residues. Actually, there are hot spots in the protein interfaces that contribute most binding energy. Here we provide a new method based on integer quadratic programming that systematically aligns protein surface structures shared by a set of proteins. This method incorporates protein sequence and structure data, and can correctly identify residues having evolutional and structural conservation between different proteins. It is sequence order independent, so can unravel the evolutional similarity between distant proteins. Furthermore, it can be used to predict hot spots with ROC area AUC=0.6. Compared with most hot spot prediction methods, our method does not need prior knowledge for the structure of protein complex or even the structure of the binding partner.