Hinge prediction using elastic network models

Abstract

The motion and interaction of rigid parts in a protein connected by hingers are important for the function of the protein. This makes identification of the hinge region particularly important. In this work a method and web server, HingeProt, which makes use of this method are developed and presented. The method employs elastic network (EN) models for predicting the rigid parts and the flexible regions connecting them in the native topology of protein chain. The elastic network models used in the method are Gaussian and Anisotropic network models (GNM and ANM, respectively). GNM calculates the mean-square fluctuations and the correlation between the fluctuations ofresidues in the most dominant (slowest two) modes. These correlations suggest hinge regions and the cooperation between them, whereas ANM provides the direction of the fluctuations of residues in the corresponding modes. The method requires a 3D structure of a single protein chain or multiple chains. Using this static conformation, a list of the rigid parts, a list of connecting hinge residues, short flexible loops and the predicted fluctuation of the residues, in the slowest two modes are calculated. The efficiency and validity of the method are tested by a method devised on two different data sets. The method is highly efficient and successful in providing the resuls within seconds for small size proteins and within several minutes for proteins consisting af a few thousands of residues. It can be used toward applications including protein function, flevible protein-protein and protein-ligand docking, flexible docking of protein structures. The server can be reached through http://www.prc.boun.edu.tr