A Contact-Assisted Approach to Protein Structure Structure Prediction and Its Assessment in CASP10

摘要

Among different approaches to predict the 3D structure of a protein, one important idea is to predict a protein residue-residue contact map and then construct a full 3D structure from the contact-map. Instead of building a structure purely from contacts information, here we describe a contact-assisted structure prediction approach that uses only a few known contacts to improve the quality of already predicted models. Our approach for contact assisted structure prediction uses a novel method for selecting and refining protein structural models. With input test data as the predicted structures for 15 protein targets used in the contact-assisted prediction category in the 10th Critical Assessment of Techniques for Protein Structure Prediction (CASP10), we demonstrate that weighted contacts satisfaction score along with other established model quality assessment scores is a promising technique for selecting good structures and ultimately for better structure prediction.