New multibody statistical potential for protein models assessment.

Abstract

Pair-wise amino acid residue-residue contact potentials are widely used to describe the accuracy of 3D protein structure models. These contact potentials (or statistical potentials) are however approximations as they consider all pair of residues as non-interacting/independent entities. Increased efforts have been made to obtain higher order statistical potentials to address these shortcomings. Here, we propose a new multibody statistical potential focusing on local environments created by the close packing of amino acid residues inside a protein. We name these local environment descriptors as 'cliques' and its corresponding statistical potential 'CLIQUE'. CLIQUE potential takes into consideration the interdependence of interactions of residues in a given neighborhood. Its utility would be to accurately recognize fundamental elements of protein structure, such as motifs and folds. A globular, non-redundant, single domain set of 1442 protein structures was used to construct the CLIQUE potential. Means of using this potential to construct an appropriate scoring function to distinguish between native and mis-folded proteins were then explored.