D-EXP019

Rational Drug Design for Tubulin Ligands
Researchers: Jean-Claude Bradley and Andrew SID Lang

Objective
We will use rational drug design (both human and bot) to find Ugi products that have the potential to bind to tubulin.

Background
Knowing the hydrogen bonding sites for a known ligand to tubulin (see below) it may be possible to design other ligands using rational drug design. Since the Bradley lab is already experienced in synthesizing Ugi products, we shall focus our design on Ugi products. 2D view of the 3 hydrogen bonding sites between paclitaxel and gamma-tubulin (from [|PDB ligand viewer] ) 3D view of the 3 hydrogen bonding sites between paclitaxel and gamma-tubulin (from [|PDB ligand viewer] )

First Run
The scaffold, see image below, was converted to using OpenBabel (GUI) with the option to add polar hydrogens. The ugi.pdb file was then used a the base ligand for AutoGrow 2, a program that runs through a series of docking rounds, growing and mutating the ligands at each round. AutoGrow was executed using the following command from the AutoGrow_2_0_4/bin directory code /usr/java/jre1.6.0_25/bin/java Main -run_mode Execute -parm_file default.prm code where default.prm is the parameter file: code //DIRECTORIES working root directory: /home/alang/AutoGrow_2_0_4/run_dir fragments directory: /home/alang/AutoGrow_2_0_4/fragment/large_fragment scripts directory: /home/alang/AutoGrow_2_0_4/scripts

//INPUT FILES initial ligand: /home/alang/D-Exp019/ugifirstrun.pdb receptor: /home/alang/D-Exp019/1JFF.pdb

//AUTODOCK PARAMETERS autodock grid center: -0.5 -16.5 15.0 autodock box size: 14 18 14

//EVOLUTION PARAMETERS number of carryovers: 10 //10 number of children: 20 //20 number of mutants: 20 //20 max number atoms: 150 //500 receptor location: -0.5 -16.5 15.0 receptor radius: 13.4 indices of hydrogens that are not linkers: -1 number of generations: 8 //8 code After 6 of 8 generations it was noticed that AutoGrow was only growing and mutating the original Ugi core structure about the single explicit hydrogen in the ugifirstrun.pdb file. This is not very useful from a combiugi viewpoint so the final two rounds were aborted. For interest, we present the two best (most negative affinity) ligands from round 6 - neither compound was found in ChemSpider: The following images are the 2D images rendered using depict, with both SMILES causing the following warning: code WARNING: Atom has unusual valence 4 (normal 5) (dy_rmbord) WARNING: ...C)(C)N(C(=O)C)C)C[NH2]c2[nH]c(nc2C(=NO)N... (dy_rmbord) WARNING:                    ^^^^^                      (dy_rmbord) code The following images are the 3D docked configurations created using AutoDock Tools:
 * 1) [[file:usefulchem/ligand1.out_ligand_1.pdb|ligand1.out_ligand_1.pdb]] (SMILES: C(=O)(N1C[NH2]c2[nH]c(nc2C(=NO)NN1c1nc(=O)c2c(nc(cn2)C(=O)O)[nH]1)Oc1c2nc(cnc2nc(N)n1)C(=O)O)C(C)(C)N(C(=O)C)C - Affinity: -10.9)
 * 2) [[file:usefulchem/ligand3.out_ligand_1.pdb|ligand3.out_ligand_1.pdb]] (SMILES: c1nc(c(n1Nc1nc2c(nc1C(=O)O)[nH]c(nc2=O)N1N(C(=O)C(C)(C)N(C(=O)C)C)C[NH2]c2[nH]c(nc2C(=NO)N1)Oc1c2nc(cnc2nc(N)n1)C(=O)O)N)C#N - Affinity: -10.8)

Second Run
The scaffold ugi.smi was protonated using molconvertor (part of Marvin 5.5.0.0 in the bin folder) and the following code: code molconvert -3 pdb:+H ugi.smi -o ugi.pdb code The file created had lines that ended in +0, e.g. code HETATM   1  C   UNK     0      -1.456  -5.067   0.269  0.00  0.00           C+0 code The file opened just fine in AutoDock Tools and Marvin but caused AutoGrow to fail (exit) when it tried to create the first set of mutants. All the end of line '+0's were removed from by using search and replace in a text editor. After they were removed, AutoGrow executed successfully using the same code as above, i.e.: code /usr/java/jre1.6.0_25/bin/java Main -run_mode Execute -parm_file default.prm code but with a slightly different default.prm file - I reduced the maximum number of atoms to 100 (from 150) and added instructions not to link to the polar hydrogen atom connected to the nitrogen (ironically the only hydrogen that was connected to in the first run): code //DIRECTORIES working root directory: /home/alang/AutoGrow_2_0_4/run_dir fragments directory: /home/alang/AutoGrow_2_0_4/fragment/large_fragment scripts directory: /home/alang/AutoGrow_2_0_4/scripts

//INPUT FILES initial ligand: /home/alang/D-Exp019/ugi.pdb receptor: /home/alang/D-Exp019/1JFF.pdb

//AUTODOCK PARAMETERS autodock grid center: -0.5 -16.5 15.0 autodock box size: 14 18 14

//EVOLUTION PARAMETERS number of carryovers: 10 //10 number of children: 20 //20 number of mutants: 20 //20 max number atoms: 100 //500 receptor location: -0.5 -16.5 15.0 receptor radius: 13.4 indices of hydrogens that are not linkers: 25 number of generations: 8 //8 code The following are the top ligands after all 8 rounds of AutoGrow None of the ligands above are in ChemSpider and I assume they would be difficult to synthesize. The ligands are also large. I would suggest using AutoGrow when the expected ligand is not too much bigger than the initial starting core (ligand). Using AutoGrow in this case seems to produce viable ligands but ligands that would probably be as difficult to synthesize as paclitaxel itself. **[Yes this is definitely a weakness of using an AutoGrow approach - even if you find a compound with a good theoretical fit it will almost certainly be difficult to synthesize. Since paclitaxel is such a large molecule I think we would have better luck creating a Ugi library with larger commercially available starting materials. JCB]**
 * ligand || affinity || SMILES ||
 * [[file:usefulchem/generation_8_ligand37.pdb|generation_8_ligand37.pdb]] || -11.9 || C(C(=O)N1CCC(=O)[C@H]([C@@H](O)c2ccco2)N2[C@@H]3C(=C(C[C@]1(C)C(=O)NCOC1=Nc4c(cccc4)N(C1)O)N2n1c(C)c(cn1)C#N)C(=S)NC(=S)N3)c1cc(=O)[nH]n2c1nnc2 ||
 * [[file:usefulchem/generation_8_ligand1.pdb|generation_8_ligand1.pdb]] || -11.6 || C(C(=O)N1CCC(=O)[C@H]([C@@H](O)c2ccco2)N2[C@@H]3C(=C(C[C@]1(C)C(=O)NCOC1=Nc4c(cccc4)N(C1)O)N2)C(=S)NC(=S)N3)c1cc(=O)[nH]n2c1nnc2 ||
 * [[file:usefulchem/generation_8_ligand2.pdb|generation_8_ligand2.pdb]] || -11.5 || C(C(=O)N(Cc1c(C)c2c(cc1)cn[nH]2)C(C)(C)C(=O)N[C@H](OC1=Nc2c(cccc2)N(C1)O)n1c(=O)ccc2c1C(=O)[C@@H](C=C2O)O)[C@@H]1C(=O)c2ccc(=O)[nH]c2C(=O)[C@@H]1O ||
 * [[file:usefulchem/generation_8_ligand28.pdb|generation_8_ligand28.pdb]] || -11.5 || C(C(=O)N1CCC(=O)[C@H]([C@@H](O)c2ccco2)N2[C@@H]3C4=C([C@H]([C@]1(C)C(=O)NCn1ncc5c(=N)sc(nc15)S)N1c5c(SC)nnc(c5N[C@H]1[S]=C4NC(=S)N3)S)N2)c1cc(=O)[nH]n2c1nnc2 ||
 * [[file:usefulchem/generation_8_ligand34.pdb|generation_8_ligand34.pdb]] || -11.5 || C(C(=O)N1CCC(=O)[C@H]([C@@H](O)c2ccco2)N2[C@@H]3C(=C(C[C@]1(C)C(=O)N[C@@H](n1ncc4c(=N)sc(nc14)S)c1cc(=O)n([nH]c1=O)c1[nH]nnn1)N2)C(=S)NC(=S)N3)c1cc(=O)[nH]n2c1nnc2 ||
 * [[file:usefulchem/generation_8_ligand29.pdb|generation_8_ligand29.pdb]] || -11.4 || C(C(=O)N1CCC(=O)[C@H]([C@@H](O)c2ccco2)N2[C@@H]3C(=C(C[C@]1(C)C(=O)NCn1ncc4c(=N)sc(nc14)S)N2)C(=S)NC(=S)N3)c1cc(=O)[nH]n2c1nnc2Sc1c2cn[nH]c2nnn1 ||
 * [[file:usefulchem/generation_8_ligand25.pdb|generation_8_ligand25.pdb]] || -11.3 || C(C(=O)N1CCC(=O)[C@H]([C@@H](O)c2ccco2)N2[C@@H]3C(=C(C[C@]1(C)C(=O)N[C@@H](n1ncc4c(=N)sc(nc14)S)c1c(C)[nH]c(n1)N(=O)=O)N2)C(=S)NC(=S)N3)c1cc(=O)[nH]n2c1nnc2 ||
 * [[file:usefulchem/generation_8_ligand3.pdb|generation_8_ligand3.pdb]] || -11.3 || C(C(=O)N1CCC(=O)[C@H]([C@@H](O)c2ccco2)N2[C@@H]3C(=C([C@@H]([C@]1(C)C(=O)NCOC1=Nc4c(cccc4)N(C1)O)Nc1nc(c4ncsc4n1)O)N2)C(=S)NC(=S)N3)c1cc(=O)[nH]n2c1nnc2 ||
 * [[file:usefulchem/generation_8_ligand5.pdb|generation_8_ligand5.pdb]] || -11.2 || C(C(=O)N(C)[C@](C)(CSc1nc2c(nc(n2n2c3c(c[nH]n3)c(=O)nc2S[C@@H]2c3ncnc(c3N=N2)O)O)c(n1)O)C(=O)NCOC1=Nc2c(cccc2)N(C1)O)[C@@H]1C(=O)c2ccc(=O)[nH]c2C(=O)[C@@H]1O ||
 * [[file:usefulchem/generation_8_ligand4.pdb|generation_8_ligand4.pdb]] || -11.2 || C(C(=O)N(CNC(=O)c1c[nH]nc1NN)C(C)(C)C(=O)N[C@H](OC1=Nc2c(cccc2)N(C1)O)n1c(=O)ccc2c1C(=O)[C@@H](C=C2O)O)[C@@H]1C(=O)c2ccc(=O)[nH]c2C(=O)[C@@H]1O ||

Using a Mathematician's Brain (Andrew Lang)
Examining the hydrogen bonding sites (see the first two images) of paclitaxel I kept the first two hydrogen bond sites and made a combi-Ugi set of virtual compounds that will be docked against tubulin. These products should be easier to synthesize that the ligands found by AutoGrow, the paclitaxel carboxylic acid is available for purchase CSID: 2043006. The first two hydrogen bonding sites can be preserved if the following carboxylic acid is used in the Ugi reaction: c1ccccc1C(=O)NC(c2ccccc2)C(O)C(=O)O media type="custom" key="9621498" This will form a Ugi product with the following structure: c1ccccc1C(=O)NC(c2ccccc2)C(O)C(=O)N([R1])C([R2])C(=O)N[R3] Where the other components of the Ugi reaction are: amine: [R1]-NH2 aldehyde: [R2]-C=O isocyanide: [R3]-N#C
 * Building the Library**

Using the same set of amines, aldehydes, and isocyanides that were used to generate @UClib007 and that are readily available in the Bradley lab, we generated @UClib008 using the carboxylic acid side chain of paclitaxel. The generated were then converted to docking-ready (fully protonated) .pdb files using molconverter and the following code: code molconvert -3 pdb:H lib008.smi -m -o lib8ligand.pdb code Lib008 was docked against Tubulin on ORU's cluster (it took several days to run). The were analysed and the  were docked locally using PaDEL-ADV (3 hours run time). Docking scores were then averaged and the (pdb - zip) are presented below. A few interesting things to note here. First, it is surprising that the paclitaxel side chain is not always docked in the same orientation as it is in the full paclitaxel molecule. This could mean that the paclitaxel side chain is not as important as assumed in this analysis, though it could also mean that the docking algorithm is missing something important. Secondly, a big concern was that the Ugi products generated from a standard library (like library007) may not be big enough (paclitaxel is comparatively bigger than the Ugi products that have previous been made in the UsefulChem project). By examining the docking visualization it seems that that assumption may not be true and that library007 may contain potential ligands, e.g. 1482 looks big enough, though the acid used here is itself relatively large and may be worth including in the next version of lib007. Thirdly, TOSMIC is showing up again as it did in the second run of @D-EXP018. The sulfonyl group seems particularly well suited to sit between the two hydrogen bonding sites at the top.
 * Results**
 * ID || amine || aldehyde || isocyanide || product || AveScore ||
 * 1482 || NC1=C(Cl)C=CC=C1 || O=Cc2cc3c(c1c2cccc1)cccc3 || [C-]#[N+]C(S(c1ccc(cc1)C)(=O)=O) || O[C@H]([C@@H](NC(=O)c1ccccc1)c2ccccc2)C(=O)N(C1=C(Cl)C=CC=C1)C(c2cc3c(c1c2cccc1)cccc3)C(=O)NC(S(c1ccc(cc1)C)(=O)=O) || -10.8 ||
 * 1497 || NC1=C(Cl)C=CC=C1 || O=Cc1ccccc1C || [C-]#[N+]C(S(c1ccc(cc1)C)(=O)=O) || O[C@H]([C@@H](NC(=O)c1ccccc1)c2ccccc2)C(=O)N(C1=C(Cl)C=CC=C1)C(c1ccccc1C)C(=O)NC(S(c1ccc(cc1)C)(=O)=O) || -10.2 ||
 * 542 || NCc1ccccc1 || O=CC=CC || [C-]#[N+]C(S(c1ccc(cc1)C)(=O)=O) || O[C@H]([C@@H](NC(=O)c1ccccc1)c2ccccc2)C(=O)N(Cc1ccccc1)C(C=CC)C(=O)NC(S(c1ccc(cc1)C)(=O)=O) || -10.15 ||
 * 543 || NCc1ccccc1 || O=CC=CC || [C-]#[N+]C1CCCCC1 || O[C@H]([C@@H](NC(=O)c1ccccc1)c2ccccc2)C(=O)N(Cc1ccccc1)C(C=CC)C(=O)NC1CCCCC1 || -10.05 ||
 * 742 || NC1CCCCC1 || O=Cc1ccc(C)cc1C || [C-]#[N+]C(S(c1ccc(cc1)C)(=O)=O) || O[C@H]([C@@H](NC(=O)c1ccccc1)c2ccccc2)C(=O)N(C1CCCCC1)C(c1ccc(C)cc1C)C(=O)NC(S(c1ccc(cc1)C)(=O)=O) || -10 ||
 * 307 || Nc1ccccc1 || O=Cc1ccc(C)cc1C || [C-]#[N+]C(S(c1ccc(cc1)C)(=O)=O) || O[C@H]([C@@H](NC(=O)c1ccccc1)c2ccccc2)C(=O)N(c1ccccc1)C(c1ccc(C)cc1C)C(=O)NC(S(c1ccc(cc1)C)(=O)=O) || -9.95 ||
 * 107 || NC1=CC(Cl)=CC=C1 || O=CC=CC || [C-]#[N+]C(S(c1ccc(cc1)C)(=O)=O) || O[C@H]([C@@H](NC(=O)c1ccccc1)c2ccccc2)C(=O)N(C1=CC(Cl)=CC=C1)C(C=CC)C(=O)NC(S(c1ccc(cc1)C)(=O)=O) || -9.85 ||

Log
**2011-05-12** **2011-05-13**  **2011-05-14**  **2011-05-28**  **2011-05-29**
 * Downloaded and installed Ubuntu using the Ubuntu Windows Installer (AutoGrow 2.0 only runs on linux).
 * Downloaded and installed on Ubuntu:
 * 1) Python (v 2.7.1)
 * 2) MGLTools (v 1.5.4)
 * 3) AutoDock Vina (v 1.1.1)
 * 4) AutoGrow (v 2.0.4)
 * By following the tutorial, set AutoGrow running overnight using the default settings.
 * Checked progress of AutoGrow and it was 1/4 complete (2 out of 8 runs complete) - left it running.
 * Checked progress of AutoGrow and it was 1/2 complete (4 out of 8 runs complete), seems to take a little longer to dock now, probably because the molecules are growing in size - left it running.
 * 2011-05-17**
 * With 6 of 8 rounds completed, docking was aborted when I noticed that AutoGrow was only attaching new functional groups to one end of the ugi.pbd file.
 * A new fully protonated ugi.pdb file was created using Marvin and AutoGrow was restarted and left to run overnight.
 * All 8 rounds completed in 11 days.
 * Began mathematician rational drug design
 * 2011-05-31 **
 * UClib008 generated and uploaded to wiki.
 * Docking-ready .pdb files for UClib008 were generated.
 * 2011-06-10 **
 * Results from docking of lib008 on cluster were analysed.