| ECCB 2002 Poster sorted by: Author | Number Next | Previous poster (in order of the view you have selected) |
Title: Development and Evaluation of a Knowledge-Based-Potential for the Scoring of Protein-Protein-Docking Results | P47 |
Grimm, V.; Schomburg, D. vera.grimm@uni-koeln.de Department of Biochemistry, University of Cologne |
| Rigid body methods are very efficient for docking co-crystallized (bound) protein conformations using mainly geometric complementarity. When docking conformations crystallized individually (unbound) several thousands of false positives with high scores can be generated. Therefore an atomic knowledge-based-potential has been developed which discriminates near-native conformations from non-natives. The potential is based on the inverse Boltzmann equation [1] and consists of two additive parts, namely a distance-dependent pair-potential of the interface atoms considering the specific atomic environment of 40 different atom types [2] and a one-body-surface-potential to correct missing entropic forces, which are believed to be substantial for the complex formation [3]. A trapeze function has been implemented to smooth the discrete distribution function. Hydrogen bonds and contacts between functional groups are taken into account as weighting factors. The repulsive part of a Lennard-Jones-potential penalizes steric overlaps. The potential parameters are optimized with a training set consisting mainly of the combase [4], displaying the most complete collection of protein-protein-complexes applied in statistical approaches to date. In eleven of fifteen unbound docking simulations with the Fourier correlation program ckordo[5] the potential ranked a near-native conformation within the top 1000. In six of these cases the near-native was found with a rank even better than 200. The geometric ranking scored the best near-native conformation only in three of fifteen cases within the best 1000. |
| [1] M. J. Sippl Knowledge-based potentials for proteins, Curr Opin Struct Biol, 5,229-35 (1995). [2] N F. Melo and E. Feytmans, Novel knowledge-based mean force potential at atomic level.J Mol Biol, 267, 207-22 (1997). [3] H. Gohlke and M. Hendlich and G. Klebe, Knowledge-based scoring function to predict protein-ligand interactions. J Mol Biol, 295, 337-56 (2000). [4] A. Sali. ftp://guitar.rockefeller.edu/pub/ilya/ [5] O. Zimmermann. personal communication |