Abstract:
Docking algorithms are widely used in determination of protein complex structures. In order to determine which structure is the most similar to native one, ranking algorithms are used in docking studies. Each year both docking and ranking algorithms are tested in a community-wide experiment called CAPRI (Critical Assessment for Prediction of Interactions). With the purpose of creating a ranking scheme, native protein complex structures were characterized according to their dynamic entities using Gaussian and Anistropic Network Models (GNM and ANM, respectively). Dynamic properties such as mean square residue uctuations (MSF), positions of hinge residues, and the extent of the correlation of the direction of structural motion in unbound and bound states were investigated. Results display that there is a drop in MSF of interface residues and increase in MSF of non-interface residues upon binding. Also, there is a shift in ANM modes upon binding, which tells that dynamic characteristics of a binding partner in unbound state is contained in dynamic characteristics a of complex. In addition, existence of hinge residues at interface is signi cant for binding. Degree of collectivity, a property which is calculated by ANM modes, increases upon binding. Using these properties, two distinct software were developed. CLAMshell, a dynamics-based clustering tool which is able to cluster native-like structures based on the similarity of the dynamics of protein chains in unbound and bound states; HiPlane, a tool to draw hinge planes and obtain plane normals for dividing the structure into dynamic domains described by global modes for characterizing orientations of binding partners. With the help of the software that were developed, dynamic properties of predicted models from former CAPRI targets and native structures were also investigated in order to build a ranking scheme.
