Towards accurate high-throughput ligand affinity prediction by exploiting structural ensembles, docking metrics and ligand similarity

MotivationNowadays, virtual screening (VS) plays a major role in the process of drug development. Nonetheless, an accurate estimation of binding affinities, which is crucial at all stages, is not trivial and may require target-specific fine-tuning. Furthermore, drug design also requires improved predictions for putative secondary targets among which is Estrogen Receptor alpha (ER).\n\nResultsVS based on combinations of Structure-Based VS (SBVS) and Ligand-Based VS (LBVS) is gaining momentum to help characterizing secondary targets of xenobiotics (including drugs and pollutants). In this study, we propose an integrated approach using ligand docking based on multiple structural en-sembles to reflect the conformational flexibility of the receptor. Then, we investigate the impact of the two different types of features (structure-based docking descriptors and ligand-based molecular descriptors) for affinity predictions based on a random forest algorithm. We find that ligand-based features have limited predictive power (rP =0.69, R2=0.47), compared to structure-based features (rP =0.78, R2=0.60) while their combination maintains the overall accuracy (rP =0.77, R2=0.56). Extending the training dataset to include xenobiotics, leads to a novel high-throughput affinity prediction method for ER ligands (rP =0.85, R2=0.71). Methods robustness is tested on several ligand databases and performances are compared with existing rescoring procedures. The presented prediction tool is provided to the community as a dedicated satellite of the @TOME server.\n\nAvailabilityhttp://atome4.cbs.cnrs.fr/ATOME_V3/SERVER/EDMon_v3.html\n\nContactschneider@cbs.cnrs.fr, labesse@cbs.cnrs.fr