Fig. 3
From: Applying atomistic neural networks to bias conformer ensembles towards bioactive-like conformations
Rigid-ligand docking and pharmacophore searching workflow for one ligand, using the AtNN ranking method (blue numbers are predicted ARMSDbio values). Up to 250 conformers have been first generated using the CSD conformer generator, then an AtNN is used to predict the ARMSDbio, that will be used to rank conformations and select only a fraction of top ranked conformers as rigid-docking seed or to be tested using a pharmacophore query. The AtNN ranking approach was compared to baselines