查看文章

Docking approaches are important tools in structure-based rational drug design. The ligandreceptor docking problem is a difficult optimization problem involving many degrees of freedom. The development of efficient algorithms and methodologies that find good solutions in less computational time would be of enormous benefit in the design of new drugs or small bioactive molecules. Genetic algorithms have been shown to be successful in docking problems [1]. In this work, a real-coded “steady-state” genetic algorithm was implemented with a grid-based docking methodology. The performance of the algorithm is tested in five HIV1 protease-ligand complexes with known threedimensional structures. In all five tested complexes the receptor structure is assumed to be rigid and the ligand molecule is considered rigid (when only the translational and rotational degrees of freedom are considered) or flexible (considering translational, rotational and conformational degrees of freedom). All ligands tested are highly flexible, having ten or more conformational degrees of freedom. The ligand-receptor scoring function used is the GROMOS96 [2, 3] classical force field implemented in the THOR [4] program. The protein active site is embedded in a 3D rectangular grid with a 0.25 angstroms discretization. On each point of the grid the electrostatic (considering a distance dependent sigmoidal dielectric function [5]) and the van der Waals terms (for each ligand atom type) of the interaction energy are pre-computed and stored, taking into account all the receptor/protein atoms, and are evaluated by tri-linear interpolation. The algorithm success is measured in …