Decoy Discrimination Using Contact Potentials Based on Delaunay Tessellation of Hydrated Proteins

Three strategies are presented for developing a knowledge-based statistical contact potential from computationally hydrated proteins that may be useful in studies of protein folding and stability. The potentials employ a definition of nearest neighbor contact based on Delaunay tessellation that identifies quadruplets of neighboring residues. The potentials are derived from a large set of proteins with explicit waters of hydration to improve the representation of the protein environment. The decoy discrimination ability of the three potentials as well as an unhydrated potential are evaluated and compared using four sets of publicly available protein decoy structures. The results suggest that the potentials can be used to infer the compatibility of a sequence of residues with a specific protein structure.