Predicting enzyme functional surfaces and locating key residues automatically from structures.

Locating functionally important protein surfaces and identifying the catalytic site residues are critical for studying enzyme functions. Here, we present a method for predicting and characterizing catalytic sites of enzymes that is fold-independent. By extract atomic patterns of catalytic residues in surface pockets computed geometrically, we develop a library of atomic patterns on protein functional surfaces of ca 700 structures. Together with propensities of secondary structures and residue occurrence in active sites, we develop a method to identify functionally important surfaces on protein structures and to locate key residues. We discuss application of our methods to amylase, dioxygenase, deaminase, dehalogenase, and hydratase. A large scale cross-validated prediction study shows that our method is sensitive and specific. Our method can used to study enzyme function, drug design, and engineering novel biochemical function.