FCANAL: Structure based protein function prediction method. Application to enzymes and binding proteins

Abstract

Structural genomics projects are beginning to produce protein structures with unknown functions, therefore, high-throughput methods for predicting functions are necessary. Although sequence based function prediction methods have been applied extensively, structure based prediction is believed to provide higher specificity and sensitivity because functions are closely related to three-dimensional structures of functional sites which are more strongly conserved during evolution than sequence. We have developed FCANAL, a method to predict functions using the score matrix obtained from the distances between Cα atoms and frequencies of appearance [1]. The previous report used key residues predicted from sequence comparison (motifs). In this report, we expanded the method to enzymes and binding proteins with key residues predicted on the basis of three-dimensional structures. Using FCANAL, we constructed score matrices for 31 enzymes. When we applied them to all the structure entries deposited at the Protein Data Bank, FCANAL could detect functional sites with high accuracy. This suggests that the FCANAL will help identify the functions of newly determined structures and pinpoint their functionally important regions. Copyright 2002 Chem-Bio Informatics Society.