Search

One of the two objectives of this thesis is to design a new method that predicts the structural effects of nsSNPs in proteins by considering structural information alone. The other objectives of this thesis are to develop new methods that model protein or domain complex structures at the atomic level. In pursuance of the first objective, to predict the structural effects of nsS-NPs, a program Bongo has been developed. Bongo represents the residue-residue interaction networks of proteins by graphs, and uses graph theoretic measures to identify key residues that are important for maintaining the interaction network. Bongo analyses the structural effects of nsSNPs by comparing the key residues that differ between the wild-type and mutant-type protein structures. The benchmark result shows that Bongo performs in a similar way to that of well-known methods PolyPhen and PANTHER, but provides complementary information to methods that rely largely on sequence. Towards the second objective, a protein-protein docking program pyDock has been developed to predict protein interfaces by modelling the structure of protein complexes. pyDock exploits either FTDOCK or ZDOCK to generate rigid-body poses of protein complexes and applies an optimised scoring function for selecting the best solutions. The performance of pyDock is similar, if not superior, to that of the most competitive rigid-body approaches. In order to further identify interfaces between protein domains, I also developed pyDockTET, a new scoring function incorporated in pyDock, to model specifically the conformation of domain-domain assemblies. The last part of this thesis presents the application of Bongo, pyDock and pyDockTET by using nsSNPs in the L2 domain of insulin receptor as examples. I conclude by discussing the contribution of the three programs to the functional annotation of nsSNPs.