文摘
In this study, we present GlamDock, a new docking tool for flexible ligand docking. GlamDock (version1.0) is based on a simple Monte Carlo with minimization procedure. The main features of the method arethe energy function, which is a continuously differentiable empirical potential, and the definition of thesearch space, which combines internal coordinates for the conformation of the ligand, with a mapping-based description of the rigid body translation and rotation. First, we validate GlamDock on a standardbenchmark, a set of 100 protein-ligand complexes, which allows comparative evaluation to existing dockingtools. The results on this benchmark show that GlamDock is at least comparable in efficiency and accuracyto the best existing docking tools. The main focus of this work is the validation on the scPDB database ofprotein-ligand complexes. The size of this data set allows a thorough analysis of the dependencies ofdocking accuracy on features of the protein-ligand system. In particular, it allows a two-dimensional analysisof the results, which identifies a number of interesting dependencies that are generally lost or evenmisinterpreted in the one-dimensional approach. The overall result that GlamDock correctly predicts thecomplex structure in practically half of the cases in the scPDB is important not only for screening ligandsagainst a particular protein but even more so for inverse screening, that is, the identification of the correcttargets for a particular ligand.