Estimation of Residue-Residue Coevolution using Direct Coupling Analysis Identifies Many Native Contacts Across a Large Number of Domain Families

Abstract

It has long been suggested that correlations among amino acid compositions at different sequence positions can be exploited to infer spatial contacts within the three dimensional structures of proteins. Here we develop a computationally efficient implementation of Direct Coupling Analysis, termed mean field DCA (mfDCA), which has the ability to disentangle direct and indirect correlations and allows us to evaluate the accuracy of contact estimation for a large number of protein domains, based purely on sequence information. mfDCA produces a large number of correctly predicted contacts, reconstructing the global structure of contact maps for many of the protein domains examined. In addition to this, our results capture clear signals beyond intra-domain residue contacts, for instance, alternative protein conformations, ligand-mediated residue couplings, and inter-domain interactions in protein multimers. Our findings suggest that inferred contacts by mfDCA can be utilized as a reliable guide in computational estimates of alternative protein conformations, protein complex formation, and even de novo domain structure prediction. This can be accomplished provided the existence of a large number of homologous sequences which are becoming available in large quantities due to recent advances in next generation genome sequencing.