Structure-based 3D-QSAR - merging the accuracy of structure-based alignments with the computational efficiency of ligand-based methods

One of the major challenges in computational approaches to drug design is t he accurate prediction of binding affinity of biomolecules. The strategies that can be applied for this purpose fall into two major categories-the ind irect ligand-based and the direct receptor-based approach. In this contribu tion, we used a combination of both approaches in order to improve the pred iction accuracy for drug molecules. The combined approach was tested on two sets of ligands for which the three-dimensional structure of the target re ceptor was known-estrogen receptor ligands and acetylcholinesterase inhibit ors. The binding modes of the ligands under study were determined using an automated docking program (AUTODOCK) and were compared with available X-ray structures of corresponding protein-ligand complexes. The ligand alignment s obtained from the docking simulations were subsequently taken as the basi s for a comparative field analysis applying the GRID/GOLPE program. Using t he interaction field derived with a water probe and applying the smart regi on definition variable selection, highly predictive models were obtained. T he comparison of our models with interaction energy-based models and with t raditional CoMFA models obtained using a ligand-based alignment indicates t hat the combination of structure-based and 3D-QSAR methods is able to impro ve the prediction ability of the underlying model. (C) 2000 Elsevier Scienc e B.V. All rights reserved.