Prediction of protein interactions by bioinformatics and physical chemistry approaches

Abstract

Proteins play key roles in many biological systems through protein interactions. Research of protein interactions can help to understand protein functions and develop new drugs. Protein interactions can be classified into homo-oligomer interactions, protein-peptide interactions, and protein-protein interactions. Protein interactions can be studied based on co-crystallized complex structure determined by X-ray crystallography or Nucleic Magnetic Resonance method, but experimentally determined structures cover only small part of the known protein- protein interactions. Therefore, there are many interests to develop computational methods for predicting protein interactions. Predicting protein interactions can be classified into methods based on bioinformatics and physical chemistry approaches. According to bioinformatics approaches, proteins with high sequence similarity convey similar interfaces and similar interactions. According to physical chemistry approaches, the funnel-like energy landscape is a general feature of protein interactions and protein interactions can be predicted by a global optimization method. In this thesis, I show bioinformatics and physical chemistry approaches for predicting homo-oligomer interactions, protein-peptide interactions, and protein- protein interactions. Both bioinformatics approaches and physical chemistry approaches played important roles to achieve improvement in predicting protein interactions.