CONFOLD new version : contact-guided ab initio protein folding with new features

Abstract

CONFOLD is an ab initio protein folding method that can build three-dimensional models using predicted contacts and secondary structures. Under this method, we can translate contact distance map and secondary structure into the distance, dihedral angle, and hydrogen bond restraints according to a set of new conversion rules, and then using this information as input to build structure models. To improve this method, we added some new features to CONFOLD, such as disulfide bond information, Beta contact prediction, and contacts distance multi-threshold. CONFOLD New Version allows using disulfide bond information and Beta strands prediction as input so that the Crystallography and NMR System can get the information directly, improving the accuracy and efficiency in some specific cases. And it can exclude some low probability residues contact information by setting multi-thresholds. I tested this method based on CASP 12 datasets, and results show that it can improve the efficiency of the program while keeping the TM-score.