IMPROVING PROTEIN STRUCTURE PREDICTION, REFINEMENT AND QUALITY ASSESSMENT TECHNIQUES

Abstract

Several novel techniques have been combined to improve protein structure prediction, structural refinement and quality assessment of protein models. We discuss in brief the development of four-body potentials that take into account dense packing and cooperativity of interactions of proteins, and its success. We have developed a method that uses whole protein information filtered through machine learning to score protein models based on their likeness to native structure. Here we consider electrostatic interactions and residue depth, and use these for structure prediction. These potentials were tested to be successful in CASP9 and CASP10. We have also developed a Quality Assessment technique, MQAPsingle, which is a quasi-single-model MQAP, by combining advantages of both “pure” single-model MQAPs and clustering MQAPs. This technique can be used in ranking and assessing the absolute global quality of single protein models. This model (Pawlowski-Kloczkowski) was ranked 3rd in Model Quality Assessment in CASP10. Consideration of protein flexibility and its fluctuation dynamics improves protein structure prediction and leads to better refinement of computational models of proteins. Here we also discuss how Anisotropic Network Model (ANM) of protein fluctuation dynamics and Go-like model of energy score can be used for novel protein structure refinement.