Developing (SPIDER) for Scoring Docked Protein-protein Interaction Complexes
Interfacial residues in native protein-protein complexes were represented by graphs, and then frequently interacting residues were extracted and used to score other docked protein-protein complexes. We demonstrated that this novel method (SPIDER) is more accurate than (ZDOCK) a commonly used scoring function, and it was ranked among the top 6 (out of 28) scoring functions in round 21 of CAPRI (Critical Assessment of PRedicted Interactions) blind test of protein-protein docking methods.
Development of Scoring Function for Scoring Protein Folds
We extracted frequently interacting patterns of internal residues in native proteins and use them to score computationally generated folds of other proteins. The novelty of the approach comes from the fact that, unlike current methods, the number of interacting residues in a pattern has no limitation. The method (SCORPIONS) is being evaluated using a database (Decoys’R’Us) of incorrect protein conformations for evaluating scoring functions.
Development of In-Silico Structure-based Protein Function Prediction Method
Advances in protein sequencing technologies have made the rate at which protein functions can be experimentally characterized much slower than the rate at which new sequences become available. Thus, the function of new protein sequences is mostly predicted by computational methods since it can often be done quickly. We have used subgraph mining as a tool to extract essential information and employ it to score protein-ligand, protein-protein complexes, and protein folds. This was important work and provided great findings. We use graph representation for protein residues, then apply subgraph mining to extract the frequent 3D-structural motifs of residues responsible for determining the function of proteins. The uniqueness of the approach comes from the fact that the number of residues in a geometric motif has no limitation, which is something that was never done before. This is expected to improve the accuracies of classifying proteins by their biological functions; we will evaluate the method by comparing its accuracy against one of the commonly used protein-function prediction methods. As a result, having the 3D crystal structure of a protein it should be possible to know its biological function, and thus, aid in solving biological problems and designing drugs.
