Reverse Fingerprints (III): Ligand-Based Pharmacophores and Virtual Binding Pockets
Incorporating 3D information into small-molecule drug design can be challenging when structural data for the biological target is absent. Given a series of ligands and associated activity data, we show how reverse fingerprints can identify important molecular motifs and generate a consensus pharmacophore query that includes atom-centered and projected features. Ligand shape and excluded volumes features can be derived from the ligand ensemble, and physical properties such as charge and lipophilicity can be mapped from the ligand ensemble onto the ligand shape surface and onto a "pocket surface" created at the ligand shape/excluded volume interface. The resulting "virtual pocket" is a hypothesis of how the ligands appear in the bound state, and what the binding pocket looks like around the bound ligands. The virtual pocket can be used for retrospective analysis or converted to a 3D pharmacophore query that can be used to search databases for new active molecules in a vHTS context.
Airdate: January 21, 2021
