Flexible-ligand docking protocol using RDkit generated low energy conformers and pocket analysis
Smina (2017.4.29)/ RDKit (Version 2018.03.2)/ PDB2PQR (Version 2.1.1)/ MGLTools (Version 1.5.6)/ AlphaSpace (Version 1.1.10)/ OpenBabel (Version 2.4.1)
Assumed pH 4.5
Gasteiger charges
1. Receptor selection for docking:
13 available cyclic-ligand bound BACE-1 structures in PDB database plus additional 20 given structures after stage 1b,
and then select the top 3 ligand-bound receptor structures by ligand similarity with the predicted ligands.
Ensemble docking using the selected 3 receptor structures.
2. Receptor preparation: PDB2PQR was used to add hydrogens to the receptor structures (—-ph_calc-method=propka —with-ph=4.5),
MGLTools was used to convert the input receptors files to pdbqt format and assign Gasteiger charges (All the waters were retained).
3. Ligand conformation selection for docking:
Ligand conformational libraries were generated using RDkit (maximum 1000 conformers per ligand, ETKDG method).
Ligand conformers within 5 kcal/mol from the lowest energy were selected for docking. (Energy calculation of ligand conformers using MMFF94)
Ligands were treated as flexible, receptor structures were treated as rigid.
The docking box site were determined by the known crystal ligands. (Known crystal ligand is used for autobox)
Exhaustiveness=32
Num_modes=20
Energy_range=7
Standard Smina scoring function
Docking runs were executed with the above specified parameters while default values were applied for the rest of the variables. And the best pose was selected in consideration of both Vina score and pocket occupancy from top 100 docked poses: (i) Pocket-based analysis of known cyclic ligands in PDB database by using AlphaSpace. (ii) On the basis of the knowledge of pocket occupancy of known cyclic ligands to select best pose from top 100 docked poses.
No
Yes
DeltaVinaFragXGB_B
chimera(Version 1.10.2)/xgboost(Version 0.80)/RDKit(Version 2018.03.2)/python(Version 3.6.3)/MSMS(Version 2.6.1)/MGLTools(Version 1.5.6)/OpenBabel(Version 2.4.1)/AutoDock Vina(Version 1.1.2)
explicit co-crystal water included
We used DeltaVinaFragXGB_B scoring function to rescore selected poses based on pose selection protocol Alignment/AlphaSpace/Smina. DeltaVinaFragXGB is a novel machine learning scoring function trained on PDBbind(v2016) and CSAR datasets via XGBoost and delta Vina parameterization. It incorporates features related to ligand conformation stability, explicit mediating water molecules and ligand fragments. DeltaVinaFragXGB_B was validated using a time-split dataset based on PDBbind and a in-house curated BACE dataset.The detailed approach will be published in upcoming paper.
No
Yes