Abstract: MATH/CHEM/COMP 2002, Dubrovnik, June 24-29, 2002

empirical potentials for protein folding based on delauney tesselation

Stephan H. F. Bernhart

Institute for Theoretical Chemistry and Molecular Structural Biology, University of Vienna, Währingerstrasse 17, A-1090 Vienna, Austria

The prediction of protein 3D structures from primary amino acid sequences and vice versa is one of the biggest challenges in modern molecular biology. The construction of empirical potentials from 3D-structure databases, pioneered by M. Sippl, has been shown to be suited for this purpose. We are using the statistical geometry approach introduced by A. Tropsha, partitioning polypeptide chains by Delaunay Tesselation. Dividing the quadruple potentials thus generated, we are able to extract more detailed information out of the amino acid pair interactions while retaining the contribution of triple and quadruple terms. Furthermore, we introduce an arbitrary watershell, simulating the solvent-solvent interaction of the solvent accessible area of globular proteins. Using this new approach, we were able to significantly improve the performance of the Tropsha based knowledge based potentials.