Convex-PL/Autodock Vina 2
OpenBabel/RDKit/Convex-PL/Autodock Vina/PyMol
Convex-PL typization 14, clusterization with 1.5 A, Vina exhaustiveness 1000
Poses obtained with Autodock Vina with both Convex-PL and weighted terms scoring functions were re-scored with Convex-PL. Convex-PL was re-trained on a set with additional halogen and sulphur-contaning ligands. Conformations with sulphonamide, triflouro and some other functional groups positions significantly different from those in the stage 1b co-crystal poses were rejected if the RMSD value was greater than a threshold. Convex-PL score was corrected with two terms involving ligand size trained with SVR. We took an average over top-5 scores as a final score.
No
Convex-PL/Autodock Vina 2
OpenBabel/RDKit/Convex-PL/Autodock Vina/PyMol
Convex-PL typization 14, clusterization with 1.5 A, Vina exhaustiveness 1000
Ligand poses were created with OpenBabel --gen3d and RDKit ETKDG algorithms. Receptors were chosen based on ligand similarity with the first 24 Cat_S ligands.
-
Docking was done with two sampling methods and two scoring methods (Convex-PL and default Vina weighted terms). Clusterization was done with RDKit GetBestRMS modification. Conformations with sulphonamide, triflouro and some other functional groups positions significantly different from those in the stage 1b co-crystal poses were rejected if the RMSD value was greater than a threshold. Obtained poses were then re-scored with Convex-PL.
No
Yes