hot-spot guided pharmacophore model building from the scratch

Structure-based pharmacophore models rely on the identification of key interactions between the ligand and its target. When the protein-ligand 3D structure is absent, this goal can be accomplished by identification of hotspots within the binding site. DRUGpy1 allows the user to find those regions based on FTMap server results, but it does no provides which features are relevant for binding. Aiming at circumventing this issue, we employed FragHotspot server (https://fragment-hotspot-maps.ccdc.cam.ac.uk/ ) to map 08 well known drug targets from DUD (http://dud.docking.org/ ), and identify the most relevant interactions within their druggable hotspots. Next, we employed Kmeans to reduce the number of features, highlighting those suitable for pharmacophore models generation. Visual inspection allows the user to select a subset of features that not only are compatible with the interactions expected for true ligands (not employed in model generation), but also afford enrichment factors higher than those found with GALAHAD (ligand-based method).