| Protein-protein interactions are fundamental to cellular functions and are associated with processes such as signal transduction, transcription and cell cycle. In this post-genomic era, reliable methods for predicting protein-protein interactions are urgently demanded to mine the vast amount of data produced by genome projects. Current in silico methods are either incapable of revealing the spatial relationship between interacting residues (sequence-based methods) or unsuitable to be used on a large scale (docking). In this thesis, three steps have been taken to develop a structure-based computational tool that can be applied for genomic-scale prediction of protein-protein interactions. First, residue-based and atom-based interfacial potentials have been developed from a non-redundant, high-quality dimer database to assess the strength of protein-protein interactions. The performance of the interfacial potentials is evaluated by using four jack-knife tests and by assessing the potentials' ability to select true protein-protein interfaces from false ones. Compared to potentials developed for monomeric protein structure prediction, the interfacial potentials perform much better at distinguishing protein-protein interactions.; Second, the PROSPECTOR threading algorithm has been extended to the prediction of quaternary structure by incorporating the interfacial potentials. This multimeric threading approach, MULTIPROSPECTOR, is comprised of two phases. In the first phase, traditional threading on a single chain is applied to generate a set of potential structures for the query sequences. Then, for those proteins whose template structures are part of a known complex, we rethread on both partners in the complex and now include an interfacial energy. The algorithm has been tested on a benchmark set comprised of 40 homodimers, 15 heterodimers, and 69 monomers. Of these, the method correctly recognized and assigned 36 homodimers, 15 heterodimers, and 64 monomers.; Third, we have applied MULTIPROSPECTOR to the genomic-scale identification of interacting partners in the yeast proteome. Each possible pairwise interaction among 6,298 encoded proteins is evaluated against a dimer database of 768 complex structures using a confidence estimate of the fold assignment and the magnitude of the statistical interfacial potentials. In total, 7,321 interactions between pairs of different proteins are predicted based on 304 complex structures. Quality estimation based on the coincidence of subcellular localizations and biological functions of the predicted interactors shows that our approach ranks third when compared to all other large-scale methods. Unlike most other in silico methods, ours is able to identify the residues that participate directly in the interaction.; In summary, MULTIPROSPECTOR is one of the first successful attempts to employ a structure-based threading method to study the protein-protein interactions on a genomic scale. It can be a useful tool in proteomics studies. |
