Multiple initial conformations re-scored by Convex-PL with a regression model (solvent modeled with points on a grid).
AutoDock Vina with in-house modifications (Convex-PL as a scoring function), RDKit 2018, SciPy, PyMOL 1.8.4, unreleased version of Convex-PL scoring function
1000 conformations for each molecule in RDKit with rings clustered at 0.15A
Ligand conformations were generated in RDKit with similarity-based distance constraints, cycles of BACE 1 - 19 were clusterized with hierarchical clustering. All clusters were selected for docking.
Vina exhaustiveness 10, Convex PL as a scoring function, number of poses 350-450, Convex-PL rescoring with typization system 14 and solvent model with points on a grid, 0.5A clusterization. Regression part of Convex-PL was trained with 4.8 cutoff of the scores and with regularization 3000.
We ran docking experiments with all available conformations of each ligand except BACE_20, and did the rescoring with regression-based version of Convex-PL.
No
Yes