Docking Ugi Products Against Tubulin Using Autodock Vina

Researchers: Andrew Lang, Lori Fielding, Jesse Patsolic, and Matthew Wilson

Objective

To dock CombiUgi library 7 (15.7 Mb, 117450 Ugi products) against tubulin using autodock vina (version 1.1.1). All of the compounds in this library have starting materials in abundance in the Bradley lab as of 07/22/2009.

Background

2D view of the 3 hydrogen bonding sites between paclitaxel and gamma-tubulin (from PDB ligand viewer )
paclibonds2.jpg
3D view of the 3 hydrogen bonding sites between paclitaxel and gamma-tubulin (from PDB ligand viewer )
pacliHbonds.jpg

Procedure

The crystal structure of tubulin with paclitaxel docked (1JFF) was downloaded from the Protein Data Bank and loaded into AutoDock Tools (1.5.4). All molecules were deleted except the receptor (beta-tubulin) which was prepared for docking as per the AutoDock Vina tutorial and saved as receptor.pdbqt. Similarly, the experimentally determined configuration of paclitaxel was extracted using AutoDock Tools and saved as exptaxol.pdb for comparison purposes.

The search space parameters were found using AutoDock Tools and saved as conf.txt, see figure 1.
receptor = receptor.pdbqt
 
center_x = -0.5
center_y = -16.5
center_z = 15
 
size_x = 14
size_y = 18
size_z = 14
searchspace.png
Figure 1. Search space parameters.

First Run

To dock paclitaxel (CSID 10368587) against tubulin (PDB 1JFF) to verify docking procedure and as a baseline for future results. The 3D structure of paclitaxel was generated from its SMILES (taken from ChemSpider) using molconvertor (part of Marvin 5.5.0.0 in the bin folder) and the following code:
molconvert -3 pdb taxol.smi -o taxol.pdb
The taxol.pdb file was then docked against the tubulin receptor using PaDel-ADV 1.6 as the GUI to AutoDock Vina 1.1.1.
[So you used only one conformer of paclitaxel? JCB][What's cool about AutoDock Vina is that is automatically detects all rotatable bonds and changes molecular conformation as it docks. AL]
---
Update 2011-05-18: Two more pdb files for taxol were created. The first (taxolallhydrogens.out_ligand_1.pdb), fully protonated, was created using molconvertor:
molconvert -3 pdb:H taxol.smi -o taxol.pdb
The second (taxolpolarhydrogens.out_ligand_1.pdb) was created using OpenBabel (using the GUI) and just the polar hydrogens were added. Comparison of the .pdbq files (taxolallhydrogens.out_ligand_1.pdbqt, taxolpolarhydrogens.out_ligand_1.pdbqt) shows no difference in protonation as AutoDock Vina 'merges' the non-polar hydrogens and leaves the polar hydrogens. This means that ligands can prepared by just using the much faster molconvert.

Docking was performed using all three pdb files (no hydrogens, polar hydrogens, all hydrogens - effectively the same, the pdbq file for all hydrogens only contain polar hydrogens) using PaDel-ADV with the best ligand (all hydrogens) shown below (Affinity: -8.8).
ligand1h.docked.png
Blue: docked - Yellow: experimental

Results of the First Run

It took 11 mins and 7 secs to dock paclitaxel against tubulin and resulted in a docked configuration with a binding affinity of -9.3, see figure 2. [Do you see the same 3 hydrogen bonding interactions as shown above for your run? JCB][I think so (visually). I'm not sure how to interpret 'affinity' and it doesn't really give any details, just a single number AL]
dockedblueisexp.png
Figure 2. paclitaxel docked in tubulin. Blue is the experimentally determined docked position, red is the docked position found using AutoDock Vina.

Second Run

To dock 10smiles.xlsx from the first half of CombiUgi library 7 against tubulin to see how long docking takes and to get a rough idea of how well the library docks. Ten 3D structures were generated from SMILES using Marvin (~3s) implementing the following code:
molconvert -3 pdb 10smiles.smi -m -o mol.pdb
The ten pdb files were then docked against tubulin using the same procedure as the first run.

Results of the Second Run

It took 22 mins 49 secs to dock all ten compounds - an average of 2 mins 16 secs per molecule. The best docking affinity was -9.2 (surprisingly close to that of paclitaxel above) for the molecule with the following SMILES: c1ccccc1C(=O)N(C1=CC(Cl)=CC=C1)C(c1ccccc1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O). See figures 3 and 4.
[Interesting - it doesn't look to me that you could get the same 3 hydrogen bonding interactions as shown above - bonding site 3 doesn't seem far enough from 1 and 2 in this Ugi product - what do you observe? I would imagine that most of the interactions are hydrophobic JCB][I think it is just bonding to sites 1 and 2 but it looks nice and snug in there. It is hard to tell if this is a fortuitous result or not until I can run a much larger sample. AL]
smi2depict.gif
Figure 3. 2D rendering of library 7 - molecule 5002

10smilesdocked.png
Figure 4. library 7 - molecule 5002 docked in tubulin using vina.

Full Run

To dock the entirety of library 7 against tubulin. With an average docking time of 2 mins 16 secs per molecule it would take over 6 months to run using a single computer. It may be possible to run the docking in under two days on a cluster currently being built at ORU.

The First 100 SMILES

The first 100 SMILES from library 7 were converted to .pdb format (fully protonated) - 100 pdbs - using the following code:
molconvert -3 pdb:H 100smiles.smi -m -o mol.pdb
These pdb files were docked against tubulin (same as above) on ORU's cluster computer in under three hours. The dockingresults.xlsx give the top ten ligands as:
Ligand
Energy
SMILES
mol42
-9.8
c2ccccc2C(=C\c1ccccc1)C(=O)N(C1=CC(Cl)=CC=C1)C(c1ccc(C(F)(F)(F))cc1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
mol28
-9.3
c2ccccc2C(=C\c1ccccc1)C(=O)N(C1=CC(Cl)=CC=C1)C(C(c1ccccc1)c2ccccc2)C(=O)NC1CCCCC1
mol32
-9.3
c2ccccc2C(=C\c1ccccc1)C(=O)N(C1=CC(Cl)=CC=C1)C(c2cc3c(c1c2cccc1)cccc3)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
mol7
-9.3
c2ccccc2C(=C\c1ccccc1)C(=O)N(C1=CC(Cl)=CC=C1)C(c1cc(O)ccc1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
mol12
-9.1
c2ccccc2C(=C\c1ccccc1)C(=O)N(C1=CC(Cl)=CC=C1)C(c1ccc(O)cc1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
mol17
-9.1
c2ccccc2C(=C\c1ccccc1)C(=O)N(C1=CC(Cl)=CC=C1)C(c1ccc(C)cc1C)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
mol92
-9.1
c2ccccc2C(=C\c1ccccc1)C(=O)N(C1=CC(Cl)=CC=C1)C(c1c(Cl)cccc1Cl)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
mol19
-9
c2ccccc2C(=C\c1ccccc1)C(=O)N(C1=CC(Cl)=CC=C1)C(c1ccc(C)cc1C)C(=O)NC(C)(C)CC(C)(C)C
mol44
-9
c2ccccc2C(=C\c1ccccc1)C(=O)N(C1=CC(Cl)=CC=C1)C(c1ccc(C(F)(F)(F))cc1)C(=O)NC(C)(C)CC(C)(C)C
mol52
-9
c2ccccc2C(=C\c1ccccc1)C(=O)N(C1=CC(Cl)=CC=C1)C(c1ccccc1O)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
The top ligand - 42 - is depicted below:
mol42.gif
[Unfortunately we have never been able to successfully run a Ugi reaction with TOSMIC despite many attempts - not sure why JCB]

Results of the Full Run

The first 1-58700 ligands of the full library were docked against tubulin in about 21 days using the ORU cluster computer. The first 58700 results: firstHalfENERGIESwithSMILES.csv. The top ten results output were

Ligand
Energy
SMILES
16562
-11.4
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(c1ccccc1)C(c2cc3c(c1c2cccc1)cccc3)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
24497
-11.1
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(Cc1ccccc1)C(c1ccc2OCOc2c1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
901
-11
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=CC(Cl)=CC=C1)C(c2cc3c(c1c2cccc1)cccc3)C(=O)NC(C)(C)C
8837
-11
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(Cc1oc(cc1)C)C(c1ccc2OCOc2c1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
24417
-10.9
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(Cc1ccccc1)C(C(c1ccccc1)C)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
8717
-10.8
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(Cc1oc(cc1)C)C(c1ccc(C)cc1C)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
24402
-10.8
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(Cc1ccccc1)C(c1ccc(C(F)(F)(F))cc1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
24492
-10.8
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(Cc1ccccc1)C(c1ccc(C)cc1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
904
-10.7
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=CC(Cl)=CC=C1)C(c2cc3c(c1c2cccc1)cccc3)C(=O)NC(C)(C)CC(C)(C)C
942
-10.7
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=CC(Cl)=CC=C1)C(c1ccccc1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)


The top ligand - 16562 - is depicted below:
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(c1ccccc1)C(c2cc3c(c1c2cccc1)cccc3)C(=O)NC(S(c1ccc(cc1)C)(=O)=O).gif

The second 58701-117450 of the full library were docked in the same way as the first set against tubulin in about 18 days using the ORU cluster computer. The results of the run: secondHalfEnergiesOUTwithSMILES.csv.
The top ten result output were

Ligand
Energy
SMILES
79172
-11.3
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(C1CCCCC1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
79217
-11.1
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(c1ccccc1C)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
79307
-11.1
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(c1ccc2OCOc2c1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
79177
-11
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(c1cc(O)ccc1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
110522
-11
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C(C)(C)C)C(c2cc3c(c1c2cccc1)cccc3)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
79247
-10.9
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(c1cc(ccc1Cl)[N+]([O-])=O)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
79173
-10.8
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(C1CCCCC1)C(=O)NC1CCCCC1
79257
-10.8
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(c1ccco1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
79222
-10.7
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(c1ccccc1O)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)
79213
-10.5
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(c1ccc(C(F)(F)(F))cc1)C(=O)NC1CCCCC1
79282
-10.5
c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(CCCCC)C(=O)NC(S(c1ccc(cc1)C)(=O)=O)

The top ligand -79172- is depicted below:

c4ccc2ccc1cccc3c1c2c4cc3CCCC(=O)N(C1=C(Cl)C=CC=C1)C(C1CCCCC1)C(=O)NC(S(c1ccc(cc1)C)(=O)=O).gif

Discussion


Conclusion


Log

2011-05-06
Downloading and installing software - Not sure what will be needed
  • Downloaded & Installed Python(version 3.2 - 17.2 MB)
  • Downloaded & Installed AutoDock Vina (version 1.1.1 - 0.5 MB)
  • PyRx failed to install ("can't find a usable init.tcl" error) so switching to a different GUI.
2011-05-07
  • Downloaded and Installed MGL Tools 1.5.4 with Python 2.5.2 (uninstalled Python 3.2 - MGL Tools installer automatically downloaded and installed Python 2.5.2).
  • Investigated CADD Pipeline, uninstalled MGL Tools 1.5.4 and downloaded and installed MGL Tools (version 1.5.6 - 39.2 MB) - didn't work, reinstalled MGL Tools 1.5.4.
  • Downloaded PaDEL-ADV (version 1.6 - 10 MB) - Java GUI
  • Downloaded tubulin pdb file from the RCSB PDB, separated the ligand from the receptor using AutoDock Tools, and then prepared the receptor by following this tutorial.
2011-05-08
  • Determined the search space parameters using AutoDock Tools.
2011-05-09
  • Downloaded and installed OpenBabel (version 2.3.0 - 8.69 MB). OpenBabel took ~60s to generate the 3D coordinates of paclitaxel.
  • Downloaded and installed Marvin (version 5.5.0.0 - 86.5 MB). Marvin generated the 3D coordinated in ~1s.
  • Docked paclitaxel and tubulin - end of first run.
  • Start of second run - chose 10 SMILES (Numbers: 1, 5002, 10003, 15004, 20005, 25006, 30007, 35008, 40009, 45010) as a sample to dock against tubulin.
  • Docked a sample of 10 SMILES from library 7 - end of second run.
2011-05-18
  • Reran three taxol pdb files against tubulin with no hydrogens, polar hydrogens, and all hydrogens to see if it made a difference.
2011-05-21
  • Converted the first 100 SMILES from library 7 to test on cluster at ORU Monday
2011-05-23
  • Tried to use a cluster of 10 computers to do docking with MOLA with Matthew Wilson and Alejandro Hernandez at ORU. There was a problem with the switch not allowing communication between the server and the nodes. Docking aborted until the switch issue can be resolved.
2011-05-31
  • ORU undergraduate cluster computer research group (Daryl Charron, Alejandro Hernandez, Maria Hernandez, Jesse Patsolic, and Matthew Wilson) docked the first 100 SMILES from library007 against tubulin using a 11-core cluster in under 3 hours.
2012-02-14
  • Matthew Wilson started docking the first 58700 ligands of the combiugiLib007 on the ORU cluster computer. The cluster consisted of from 16-47 nodes.
2012-03-14
  • Jesse Patsolic began parsing the output files to obtain the energies given by the docking run. Matthew Wilson started docking the second half of the combiugiLib007 on the ORU cluster computer. The cluster consisted of approximately 47 nodes.
2012-04-01
  • The second half of combiugiLib007 completed docking.
2012-04-08
  • Jesse Patsolic parsed the output files from the second run to obtain the energy values and uploaded the results to the wiki.