MEDock: a Maximum-Entropy based docking web server for efficient prediction of ligand binding sites
The prediction of ligand binding site is an essential problem in the drug discovery process. The knowledge of binding sites greatly facilitates lead optimization, site-directed mutagenesis experiment design, and the finding of structure motifs that may exist in other proteins which eventually cause drug adverse effect. However, docking is still the rate limiting step for such predictions and more efficient algorithms are demanded.
The MEDock (Maximum-Entropy based Docking) web server is aimed at providing an efficient utility for prediction of ligand binding site. A major distinction in the design of MEDock is that its global search mechanism is based on a novel optimization algorithm that exploits the maximum entropy property of the Gaussian distribution.
In many benchmark cases, MEDock has been demonstrated to achieve better convergent results for very diversified scenarios of ligand binding interactions. These benchmarks were carried out in comparison with the Lamarckian genetic algorithm. They consistently showed that MEDock converged to the correct binding modes with significantly less numbers of energy evaluation. Given a threshold for number of energy evaluation in the docking simulation, MEDock also greatly elevates the rate of accurate prediction for all benchmark cases.
PDBQ is an extended PDB format that is used in many docking programs, e.g., AutoDock. The PDBQ format can be generated by many chemical software or web servers, among which Dundee's PRODRG server provides a nice visual interface to generate this file format from PDB file of a ligand. The PDBQ file for proteins can be derived from the PDB2PQR server and a simple awk script. Standalone programs for both PRODRG and PDB2PQR provide stand-alone are also available. Following are the PDBQ files of some benchmark cases as concrete examples: