Abstract:
The interface type of the proteins are useful in determining the functioning mechanism of proteins. The interface type may either be obligatory, non-obligatory or crystal (non biological). Previous studies have tried to find the interface type by various sequence and structure based methods such as residue interface propensity, conservation, hydrophobicity, shape etc.. In this thesis, a methodology based on the fluctuations of residues by the Gaussian Network Model (GNM) was developed to predict the interface type for a given protein complex structure. The scoring function identifies the interface type by analysing the domains, the associating regions across the interface of the complex structure and the plausible binding sites of the chains of the complex structure in their isolated states. The reliability of this method was tested on two datasets; PPIPred, and CAPRI. Out of 111 proteins in the PPI-pred dataset, correct evaluation rate was 82 percent for obligatory proteins and 76.5 percent for non-obligatory proteins. In the CAPRI experiment, on the other hand, 6 out of 10 submitted models in a pool of predicted models of 1600, were successful. The latter suggests that the method is also successful in discriminating the biological and non-biological interfaces. A web server is built for the prediction of the type of the interface for any given protein complex structure (http://www.prc.boun.edu.tr/appserv/prc/interprot/).
