Note: LigandScout has since evolved beyond version 4.3 (current releases include 4.4 and 5.x), but 4.3 remains widely cited in academic literature for its robust pharmacophore performance.
and is widely used by researchers to identify key chemical features responsible for the biological activity of small molecules. Key Capabilities ligandscout 4.3
Perhaps the most disruptive addition in is the integration of DeepScoring 2.0 . Previous versions used a simple RMSD-based fit value. Version 4.3 employs a graph neural network (GNN) trained on the BindingNet v2 dataset (1.2 million annotated ligand-target pairs). Note: LigandScout has since evolved beyond version 4
A 2024 pre-print from the University of Barcelona utilized to identify novel ROCK1 inhibitors. Using a structure-based pharmacophore derived from PDB: 2ETR, the team screened a library of 2.5 million compounds in 14 hours on a single workstation (Intel i9, 128GB RAM, RTX 4080). The top 10 scoring compounds (by DeepScoring 2.0) were tested in vitro . Three had IC50 values below 500 nM, and one provided a novel scaffold (patent pending). The researchers explicitly noted that version 4.2 had previously failed to map the solvent-excluded volume correctly for this target, leading to false negatives. Previous versions used a simple RMSD-based fit value