Docking At UTMB is a web-resource that provides controlled access to molecular docking software running on a cluster of computers at UTMB. Approved users can login, upload a protein PDB file along with an active-site specification, and dock that protein against one of several small-molecule libraries.

For temporary access to perform single protein-ligand dockings, login as user "anonymous" (password "anonymous").
To request an individual account to dock against virtual libraries, please e-mail Dr. Stan Watowich at watowich@xray.utmb.edu.

Preamble: Computer docking and virtual screening are inexact, but potentially very valuable, tools. This site is intended to provide easy access to researchers wishing to perform small numbers of docking or virtual screening experiments, but who do not have the necessary computer resources and/or computational biology backgrounds. This interface can handle most protein-ligand docking experiments, but there will always be systems where this simple interface will fail. For those cases, researchers will need to develop the necessary expertise or collaborations to perform the experiment.

Suggested Use: Before beginning any virtual screening experiment against their target protein, researchers should (must) validate that the docking software is working correctly for their system. The target protein can frequently be found in the Protein Data Bank as a cocrystal structure with a bound ligand. The ligand should be extracted from the cocrystal structure and used in a single docking experiment against the target protein. If the docked structure predicted by this test does not reproduce the cocrystal structure, there is little point performing a virtual screen with the target protein. This test can identify problems such as incorrect His charging in the active site, incorrect search space, missing cofactors, etc. Manually correct these problem and repeat the docking test.

If the above validation proves successful, researchers can request an individual account (e-mail Dr. Stan Watowich) in order to screen their target protein against either a Maybridge "soluble" drug-like library of 5,163 small molecules, the MayBridge HitFinder library of 14,400 small molecules, or a Chembridge "soluble" drug-like library of 23,311 small molecules. This virtual screening experiment will return predictions of good binding ligands. These ligands can be purchased from Maybridge or Chembridge for wet-bench validation. Success rates (i.e., binders found from experimental testing vs number of computer predtions tested) are quite variable, but can range as high as 40% for some test systems. See the question "How good is the docking software" for enrichment details for a number of test systems.

If researchers have screened the Maybridge and/or Chembridge libraries and experimentally validated the Autodock Vina predictions, they can request access to screen our virtual library of ~600,000 commercially-available drug-like molecules.

Autodock Vina [1], written by O. Trott and A. Olson at the Scripps Institute is used to perform the actual docking. This open-source program has been made freely available by the authors.

AutodockTools, written by G. Morris and A. Olson at the Scripps Institute is used to convert proteins and ligands to the format required by AutoDock Vina. This open-source program has been made freely available by the authors.

[1]
AutoDock Vina: Improving the Speed and Accuracy of Docking with a New Scoring Function, Efficient Optimization and Multithreading
O. Trott and A. J. Olson
Journal of Computational Chemistry 31 (2010) 455-461

Before using Autodock Vina for this project it was tested in-house against the Directory of Useful Decoys (DUD) Database [1], a database designed specifically for rigorously testing docking algorithms. The testing protocol was taken from from an earlier paper [2] written at Wyeth Research that tested six other molecular docking programs against the DUD database. The results showed Autodock Vina to be a strong competitor against the other programs, and at the top of the pack in many cases.

[1]
Benchmarking Sets for Molecular Docking
N. Huang, B. K. Shoichet, and John Irwin
J. Med. Chem., 2006, 49(23), pp 6789-6801

[2]
Comparison of Several Molecular Docking Programs: Pose Prediction and Virtual Screening Accuracy
J. B. Cross, D. C. Thompson, B. K. Rai, J. C. Baber, K. Y. Fan, Y. Hu, and C. Humblet
J. Chem. Inf. Model., 2009, 49, 1455-1474

Job Name	A single word project name/descriptor.
Protein File	Hit the "Choose File" icon and upload a protein file from your computer.
Active Site	Enter an {x, y, z} to designate the center of the search volume. Typically, this coordinate is the center of the protein's active site.
Grid Size	Specify, in angstroms, the size of the box around the active site to search within. Too large a value can lead to inaccurate binding positions outside of the active site.
Database	Enter the ligand database to search or upload a single ligand file from your computer.

When used to dock a single ligand, Docking At UTMB will return:

Log	A log file describing the tasks performed and noting any warnings or errors.
DockedLigand.pdb	A file containing the five best (i.e. lowest docking scores) poses of the docked ligand in PDB format.
DockedLigand.log	A file containing the Vina docking scores for the five best poses. Poses with similar docking scores should be treated as equally likely.
Protein.pdbqt	A file containing the target protein in PDBQT format.

When used to dock against one of the provided ligand libraries, Docking At UTMB will return:

Log	A log file describing the tasks performed and noting any warnings or errors.
Report	A file listing the top 500 binders, sorted such that the best (lowest) score is first. The ligands are referenced by ZINC codes, and these can be extracted and pasted into the ZINC web site to return chemical structures and catalog numbers.
TopResults/	A folder containing the top 500 binders, each with their top five poses in PDB format. Additionally, a separate log file for each binder containing the Vina scores for it's five best poses.
Protein.pdbqt	A file containing the target protein in PDBQT format.

Maybridge "soluble" drug-like library (5,164 molecules):

• This is a collection of small drug-like compounds readily available from Maybridge.
  
  The Maybridge library used by Docking At UTMB was generated from the ZINC site by imposing the following filters:
  

  1.	Compounds had to be at least commercially availabe from the Maybridge chemical company.
  2.	Compounds had to be drug-like, with net charge 0, less than 9 rotatable bonds, less than 11 H-bond acceptors, and MW between 250 and 650 Da.
  3.	xlogP had to be less than 1.8. Ths restriction was imposed to bias the library towards more soluble compounds. These filters gave 5,164 compounds in MOL2 format.
  4.	MOL2 formatted files were converted to PDBQT formats using the prepare_ligand4.py program available in the AutoDockTools package. The charges provided from ZINC were retained. Some errors in charge state, torsions, bonding, atom types, etc. may exist in the final PDBQT formatted library. Please report any suspected errors to us.

Chembridge "soluble" drug-like library (23,311 molecules):

• This is a collection of small drug-like compounds readily available from Chembridge.
  
  The Chembridge library used by Docking At UTMB was generated from the ZINC site by imposing the following filters:
  

  1.	Compounds had to be at least commercially availabe from the Chembridge chemical company.
  2.	Compounds had to be drug-like, with net charge 0, less than 9 rotatable bonds, less than 11 H-bond acceptors, and MW between 250 and 650 Da.
  3.	xlogP had to be less than 1.0. Ths restriction was imposed to bias the library towards more soluble compounds. These filters gave 23,311 compounds in MOL2 format.
  4.	MOL2 formatted files were converted to PDBQT formats using the prepare_ligand4.py program available in the AutoDockTools package. The charges provided from ZINC were retained. Some errors in charge state, torsions, bonding, atom types, etc. may exist in the final PDBQT formatted library. Please report any suspected errors to us.

MayBridge HitFinder library (14,400 molecules):

• A virtual version of the MayBridge HitFinder library that is widely used in experimental High-Throughput Screening.This library contains 14,400 unique small molecules. This library was graciously generated for us by Dr. John Irwin, and was parameterized using the ZINC database protocols. MOL2 files from ZINC were converted to PDBQT formats as described above.
  

Custom Drug-Like ZINC Subset (597,000 molecules):

• A custom subset of the ZINC Database containing only "clean" drug-like compounds from the vendors ChemBridge, ChemDiv, Ryan Scientific, Maybridge, and Sigma-Aldrich. This very large library was graciously generated for us by Dr. John Irwin.

Before being utilized by Autodock Vina, any uploaded PDB protein is - by necessity - converted to the PDBQT format. This involves stripping all heteroatoms (including waters), re-protonating all atoms, and reassigning Gasteiger charges to all atoms. This conversion is done using tools (prepare_receptor4.py) provided in AutoDockTools. The PDBQT formatted protein is eventually returned as part of the results for your inspection.

If you wish to avoid any of these steps, provide your protein file already in PDBQT format. None of the above processing steps are performed on protein files already in the PDBQT format. Information on how to generate PDBQT files for proteins (including assigning hydrogens to His residues) can be found in the AutoDockTools manual and tutorial.

Single ligands can be uploaded in PDB, MOL2, or PDBQT format. Ligands uploaded in PDB or MOL2 format are automatically converted into PDBQT format. This involves re-protonating all atoms, reassigning Gasteiger charges to all atoms, and algorithmically determining which bonds are rotatable. Note that some functional groups, including F, do not seem to be processed by AutoDockTools.

If you wish to avoid the automatic conversion process (e.g., if your ligand generates an error message during conversion or functional groups were not converted correctly), provide your ligand file already in PDBQT format. None of the above processing steps are performed on ligand files already in the PDBQT format. AutoDockTools (see above) can be used to interactively convert ligands to PDBQT format.

Ligand files are limited to 200 total atoms and 15 total rotatable bonds.

Docking times will vary according to the library you select. Currently the Test library takes ~15 minutes, the Maybridge 5K library takes ~12 hours, the MayBridge HitFinder library takes ~2.5 days, the Chembridge 23K library takes ~3 days, and the ZINC library takes roughly 90 days.

We are moving this portal to the supercomputer at the University of Texas (TACC), and this should improve performance by a factor of ~500.