Inferring Protein Structural Contacts from the Evolutionary Record -- the Advantage of Considering Amino-Acid Identities
Covariance between amino acid at different residues of a protein family has been used to infer structural contacts. Different from conventional covariance-based approaches, direct-coupling analysis (DCA) is designed to disentangle direct and indirect correlations, enabling it to predict contacts accurately. Here, we show that in addition to the co-occurrence pattern between residues, the amino-acid identities underlying the residue pairs also provide information for structural contact prediction. We propose a framework that takes both evolutionary record and amino-acid identities into account and achieves better prediction accuracy than DCA across many protein families. In addition, we extend this framework to a recently emerged contact definition that is more structurally informative than previous ones. The contacts predicted by our approach could serve as reliable guides for protein structure prediction.
Events are free and open to the public unless otherwise noted.