3D-QSAR-boosted SBVS

A script-driven workflow for the TDT Challenge 2 is proposed, which combines structure-based and ligand-based methodologies. Ligands and proteins were prepared using OpenBabel and UCSF Chimera. Ensemble docking of DSM inhibitors on five crystallographic DHODH structures was carried out with AutoDock VINA. Docking poses were pre-filtered by their VINA score, then selected based on their steric/electrostatic similarity to the experimental poses of the co-crystallized ligands, as assessed by Open3DALIGN. The same strategy was followed to predict the binding mode of compound 6. The selected poses of DSM inhibitors were then used to build and validate a 3D-QSAR model with Open3DQSAR. Pre-filtering and clustering of the eMolecules database via 2D descriptors and fingerprints was carried out with the RDKit. The most promising candidates were then ensemble-docked into the DHODH structures, and their pIC50s were predicted using the aforementioned 3D-QSAR model.