A New Generation of Statistical Potentials for Proteins

Abstract

We propose a novel and flexible derivation scheme of statistical, database-derived, potentials, which allows one to take simultaneously into account specific correlations between several sequence and structure descriptors. This scheme leads to the decomposition of the total folding free energy of a protein into a sum of lower order terms, thereby giving the possibility to analyze independently each contribution and clarify its significance and importance, to avoid overcounting certain contributions, and to deal more efficiently with the limited size of the database. In addition, this derivation scheme appears as quite general, for many previously developed potentials can be expressed as particular cases of our formalism. We use this formalism as a framework to generate different residue-based energy functions, whose performances are assessed on the basis of their ability to discriminate genuine proteins from decoy models. The optimal potential is generated as a combination of several coupling terms, measuring correlations between residue types, backbone torsion angles, solvent accessibilities, relative positions along the sequence, and interresidue distances. This potential outperforms all tested residue-based potentials, and even several atom-based potentials. Its incorporation in algorithms aiming at predicting protein structure and stability should therefore substantially improve their performances.

Statistical Contact Potentials in Protein Coarse-Grained Modeling: From Pair to Multi-body Potentials

“…The knowledge-based potential functions can usually be divided into two types: atomic-level potentials (Samudrala and Moult 1998;Gatchell et al 2000;Lu and Skolnick 2001;Zhou and Zhou 2002;McConkey et al 2003;Hubner et al 2005;Qiu and Elber 2005;Shen and Sali 2006) and coarse-grained potentials (Tanaka and Scheraga 1976;Miyazawa and Jernigan 1985;Hendlich et al 1990;Sippl 1990;Hinds and Levitt 1992;Miyazawa and Jernigan 1996;Eisenberg et al 1997;Simons et al 1999;Zhang et al 2006;Dehouck et al 2006;Dong et al 2006). The latter have been demonstrated to be highly effective for reducing the computational cost in modeling native protein structures, although they are sometimes considered not to be sufficiently rigorous to reflect the entire landscape of a potential energy surface (Thomas and Dill 1996b;Skolnick 2006).…”

Statistical energy potential: reduced representation of Dehouck–Gilis–Rooman function by selecting against decoy datasets

“…This SEEF is of high accuracy in discriminating the native proteins from their decoys. It has outperformed all tested residue-based potentials, and even performed better than a couple of atom-based potentials (Dehouck et al 2006). One of the most outstanding merits of Dehouck-Gilis-Rooman potentials is that it is formed mainly based on the interdependence of correlations among several different sequence and structure descriptors.…”

“…Dehouck-Gilis-Rooman group has recently developed a novel SEEF in which the total folding free energy could be decomposed into a sum of lower order terms, i.e., the protein potentials could be decomposed into different coupling terms (42 terms in total), with each term being a function of a combination of sequence and structure descriptors, i.e., amino acid types (s i ), backbone conformations (t i ), solvent accessibility (a i ) and inter-residue distance (d i ) (Dehouck et al 2006). This SEEF is of high accuracy in discriminating the native proteins from their decoys.…”

Protein design with a comprehensive statistical energy function and boosted by experimental selection for foldability

“…Potentially, an SEF may pick up factors in protein sequence-structure relationships that are not yet treated properly by current physics-based models. Although SEFs for protein structure prediction have been well developed 20,21 and most current protein design approaches contain certain statistical terms 4,13 , a comprehensive or full-scale SEF that by itself achieves automated protein design has not been established to compete with state of the art physics-based models 22 . In most previous SEFs, probability distributions were estimated based on a prior discretization of structural properties, for example, the solvent accessibility partitioned into a few discrete categories, or a distance divided into bins.…”