 |
|||
|
|||
        |
|||
   |
|||
 |
|
An independent user study ranks eHiTS v9.0 at the top A group of researchers from the University of Warsaw evaluated 7 docking programs from three aspects: pose prediction accuracy, score correlation with experimental binding energy and score-rmsd correlation. In all 3 categories eHiTS is among the best-2 contenders, while the other-best is always a different program of the seven. This makes eHiTS overall the most reliable and versatile tool. See detailsA comparative validation
study with eHiTS v 6.1 A comparative validation study was performed by Dr. Fedor Zhuravlev [ref 1], from the Technical University of Denmark following the methodology described by M. Kontoyianni et. al., [ref 2] for 69 PDB targets, belonging to 14 protein families. The PDB entry codes are listed below. Water was removed from the PDB files, protonation states and partial charges were assigned automatically by eHiTS during the run. The standard parameter set was used and we let the porgram generate up to 60 docking poses for each case. The RMS deviation of the heavy atom coordinates of each pose from the original X-ray ligand pose was calculated. The following table contains some statistics on the results for both the best-scoring and closest to Xray poses of each input case.
Note*: the average time was measured on a single processor Pentim IV, 2.4 GHz Linux PC.
In comparison to the other five docking tools that are analysed in the paper [ref 2]: DOCK, FlexX, LigandFit, GOLD and GLIDE, eHiTS seems to outperform them all on this vendor independent dataset. The above results were obtained using the standard default parameter set. eHiTS has a full suite of automated training utilities, which allow the parameters to be trained using a set of protein-ligand complex Xray structures (for family based validation training) or training using a set of actives and decoys (for enrichment training). When the parameters are tuned for a specific protein family, the results improve significantly. The eHiTS run typically takes 4-6 minutes on a single processor workstation when the standard parameter set is used (producing the above results). The program also offers the use of different accuracy levels easily set via command line flags or parameter files. The average run time ranges from 13 sec/ligand to 30 min/ligand depending on the setting. For further speed up, especially for multi ligand screening runs, one can use the database or the filter tool in eHiTS. The 69 PDB entry codes used in the comparative validation:
A large volume, drug-like validation studyA large volume validation study was performed on 1586 drug-like PDB codes. These codes were selected as a combination of the GOLD validation set, the PDB-bind 2004 set, augmented with some more codes from the PDB, filtered with the Lipinski's rule of five [ref 3] for the ligands. The PDB codes (see below) were processed with eHiTS as downloaded, i.e. no pre-processing of any kind (e.g. no pre-optimisation of the receptors or ligands) was done prior to the automated eHiTS run. This make eHiTS usage extremely simple since all the preparation steps are done automatically.eHiTS processed successfully all of the 1586 codes, i.e. did not fail to dock any of them. The average top-ranking solution's distance from the X-ray was: 1.1 A, while the average distance of the closest solution to the X-ray was: 0.73 A. Average docking time was 2.1 minutes on a Pentium IV PC. The docking jobs were executed in standard docking mode, i.e. accuracy level was set to 3. Results are highlighted in the chart below: 
The eHiTS docking validation setThe 1586 PDB entry codes used in the large-volume, drug-like docking validation can be downloaded from here: eHiTS_Validation_Set_DL1568 size: 131 MB. Notes: the directory name when unpacked is: ehits_validation_1568. Under that, there are 1568 subdirectories, named by the PDB code, each directory contains only 2 files, the protein and the ligand. The files are either mol2 (if available, e.g. from GOLD/PDB-bind set) or pdb format. The protein files have no water solvent and they are not the full file only a region around the active site (~7A margin) Reference:[1] Personal communication with Fedor Zhuravlev, Assist.Prof., Technical University of Denmark, Kemitorvet, DTU 201, DK-2800 Kgs. Lyngby; www.kemi.dtu.dk [2] Kontoyianni, M.; McClellan, L. M.; Sokol, G. S.: Evaluation of Docking Performance: Comparative Data on Docking Algorithms, J. Med. Chem., 2004; 47(3); 558-565. [3] Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney P. J.: Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv. Drug Deliv. Rev. 1997, 23, 3-25. |
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|
Copyright ©
2010 SimBioSys Inc., All rights reserved.
|