Fig. 1
From: Cobdock: an accurate and practical machine learning-based consensus blind docking method
Schematic representation of CoBDock blind docking workflow. The docking methods, AutoDock Vina, PLANTS, GalaxyDock3 and ZDOCK, and binding site detection tools, P2Rank and Fpocket, are all executed by CoBDock in parallel. A three-dimensional 10 Å-resolution grid is drawn over the protein, and each predicted binding mode and pocket is assigned to the closest grid box. Boxes containing no binding modes or pockets are subsequently removed. Each remaining grid box is assigned an ML-computed “pocket score” that is used to rank them. The pocket closest to the top-ranked box is then selected as the true binding site. After binding site selection, molecular docking is executed at the binding site to produce the final binding mode for the ligand