Re-Docking by Analyzing the Profile of Protein-Protein Interaction

Abstract

Various proteins exert molecular functions through forming a protein complex. Elucidation of a complex structure is essential to understand its molecular functions. At present, the number of complex structures in PDB is not enough for us to understand the information coded on the protein-protein interaction network. In silico analyses, docking simulation has been applied to augment the lack of complex structure information. In general, software for docking simulation rotates a protein (ligand) around its interaction partner protein (receptor) to generate many complex structure candidates. Hereafter, these candidates are referred to as decoys. Similar complex structures to the native structure (near-native structures) are expected to exist among the decoys. However, in some protein pair cases, there is no near-native structure in a set of decoys because of the shortage of docking space, that is one of the important problems to be solved. This problem was addressed by expanding the docking space based on re-docking strategy [Uchikoga et al. 2013 PLOS ONE 8:e69365]. Re-docking is a second round docking step after the initial docking. Then, analyses of Interaction FingerPrints (IFPs) [Uchikoga and Hirokawa 2010 BMC Bioinformatics 11:236-245] have been proved to obtain near-native structures efficiently by focusing on the docking space corresponding to a decoy. Hence, we developed a software named Pftkool that is a tool for re-docking by calculating and classifying IFPs. Pftkool uses k-means algorithm to search receptor surface for target residues to perform re-docking. In general, the number of k of k-means algorithm should be specified by a user. However, our analyses show that k is automatically determined in the case of re-docking.