High resolution fast quantitative docking using fourier domain correlation techniques

Abstract

A ‘docking’ method based on finite grid forcefield sampling is proposed for fast evaluation of interaction energies between macromolecules and ligands. Forcefield used to calculate interaction energies utilizes a potential energy function composed of a 1/r-dependent electrostatic term and a (6-12) Lennard-Jones term for van der Waals interactions. Fast evaluation makes use of the convolution theorem allowing a point-by-point N-dimensional correlation in direct space to be replaced by a simple multiplication in spatial frequency space. Predictive accuracy was assessed by using seven protein-ligand complexes available from the Brookhaven Data Bank and determined crystallographically to high resolution. Successful prediction of ligand position and determination of ligand–protein interaction enthalpy was dependent on forcefield sampling grid size. Minimum interaction enthalpy calculated for four protein–ligand complexes coincided with crystallographic structures that used sampling grid sizes of 0.25 Å resolution and was independent of ligand starting position and orientation. Successful docking was obtained for the remaining complexes at same grid resolution provided ligand starting positions were not randomized. Sensitivity of the docking algorithm to starting orientation was a consequence of tight fit of respective ligand structures with their protein target sites for these three cases and can be circumvented by using finer rotational sampling grids for the ligand. Boltzmann statistics derived from calculated interaction energies successfully extracted the observed ribonuclease A cytidylic acid complex from a manifold of similar interaction energies. The proposed method was able to reproduce the observed crystallographic complex by using a dynamical description of ligand. © 1997 Wiley-Liss Inc.