Glide Docking
Maestro Molecular Modeling Interface Version 11.5.011, MMshare Version 4.1.011, Release 2018-1, Platform Linux-x86_64; Glide
Target pH 7.0 +/- 2.0
Included tautomers and at most 32 stereoisomers for ligands
1) Ligand preparation
We ran our ligand SDF file through Maestro LigPrep for ligand preparation using default settings. Ligands were desalted and generated at pH 7.0 +/- 2.0 state using Epik. We included tautomers and at most 32 stereoisomers using specified chiralities. From 39 molecules, we generated a total of 169 molecule conformations.
2) Protein preparation
Using Maestro Protein Preparation Wizard, we used the default settings and added hydrogens to proteins 3IEJ and 5QBU. The settings assigned bond orders, created zero-order bonds to metals, created disulfide bonds, used the CCD database, deleted waters beyond 5.00 angstroms from het groups, and generated het states using Epik at ph 7.0 +/- 2.0. Additionally, the settings sampled water orientations, used PROPKA to predict pKas of residues, removed waters with less than 3 H-bonds to non-waters, and converged heavy atoms to 0.30 angstroms.
3) Receptor grid generation
For both 3IEJ and 5QBU, Chain A was the receptor site and Ligand A was the reference ligand in our receptor grids. We picked Ligand A to remove it from the protein and define our binding site, then kept the default settings. For the grid, the Van der Waals radius scaling factor was 1.0 with partial charge cutoff of 0.25, the center was set to the centroid of the selected Ligand A, and we chose to dock similarly sized ligands to Ligand A.
Used Glide standard precision
We performed ligand docking with the generated 3IEJ receptor grid, the 5QBU grid, and the previously prepared 169 ligand conformations using Glide SP. We submitted the top pose for each unique ligand for both 3IEJ and 5QBU. The docking scores are located in the SuppInfo directory.
No
Glide Free Energy Scoring
Maestro Molecular Modeling Interface Version 11.5.011, MMshare Version 4.1.011, Release 2018-1, Platform Linux-x86_64; Glide
Target pH 7.0 +/- 2.0
1) Ligand Conversion
We converted the provided SMILES strings to an SDF file using Online SMILES Generator (https://cactus.nci.nih.gov/translate/) and OpenBabel.
2) Ligand preparation
We ran our ligand SDF file through Maestro LigPrep for ligand preparation. The default settings were used. Ligands were desalted and generated at pH 7.0 +/- 2.0 state using Epik. We included tautomers and at most 32 stereoisomers using specified chiralities. From 39 molecules, we generated a total of 169 molecule conformations.
3) Protein preparation
We used 3IEJ, a crystal structure of cathepsin S in complex with a pyrazole based cathepsin S inhibitor to guide our free energy calculations. Using the Maestro Protein Preparation Wizard, we used the default settings and added hydrogens to the protein. The settings assigned bond orders, created zero-order bonds to metals, created disulfide bonds, used the CCD database, deleted waters beyond 5.00 angstroms from het groups, and generated het states using Epik at ph 7.0 +/- 2.0. Additionally, the settings sampled water orientations, used PROPKA to predict pKas of residues, removed waters with less than 3 H-bonds to non-waters, and converged heavy atoms to 0.30 angstroms.
4) Receptor grid generation
To define the receptor grid, we used the protein processed version of 3IEJ. Since the crystal structure of Cathepsin S was a dimer, Chain A was the receptor site and Ligand A was the defined ligand. We picked Ligand A to remove it from the protein and define our binding site and kept the default settings. For the grid, the Van der Waals radius scaling factor was 1.0 with partial charge cutoff of 0.25, the center was set to the centroid of the selected Ligand A, and we chose to dock similarly sized ligands to Ligand A. We repeated this process with the protein 5QBU.
5) Ligand docking
We performed ligand docking using the generated 3IEJ receptor grid, the 5QBU grid, and the previously prepared 169 ligands. The ligand docking used standard precision or Glide SP.
6) Viewed poses, ligand interaction diagrams
The docking job generated a docking score, glide GScore, and glide EModel. Docking scores were similar to Glide GScore but included Epik penalties. Glide GScores were the ligand binding free energy calculations. Glide EModel used GScore to calculate the best pose among the conformations of the same ligand. The free energy results submitted came from Glide GScore and were the average of the free energies from 3IEJ and 5QBU. The poses for the 170 ligands were viewed using Pose Viewer and ligand interaction diagrams were examined for the top conformations. For 3IEJ, the original ligand interaction diagram for Ligand A with cathepsin S included pi-pi stacking between the pyrazole ring and Phe 211 and Phe 70, pi-pi stacking between the benzene ring adjacent to the pyrazole ring and Phe 70, an H-bond between the ketone and Val 162, and a pi-cation bond between Phe 70 and the NH+ on the ligand. To check the best poses, we looked at the ligand interaction diagrams for similar interactions as the ones shown in Ligand A. For example, the rank 1 ligand for 3IEJ exhibited pi-pi stacking between the Phe 211 and its benzene ring, and created multiple H-bonds between NH2 and LEU 116/GLU 115 and SO2 with THR 72/water. We repeated this process with the protein 5QBU. Due to time constraints, we had a free energy sample size of 2.
References:
Ameriks MK, Axe FU, Bembenek SD, et al (2009) Pyrazole-based cathepsin S inhibitors with arylalkynes as P1 binding elements. Bioorg Med Chem Lett 19:6131-6134. doi: 10.1016/j.bmcl.2009.09.014
Ameriks MK, Bembenek SD, Burley SK, et al (2017) Crystal structure of human Cathepsin-S with bound ligand. doi: 10.2210/pdb5QBU/pdb
No