Extended pharmacophores in virtual screening for drug discovery  

Virtual screening is fast becoming a mainstay of the pharmaceutical industry as ‘intelligent’ or ‘focused’ discovery technologies supplant more traditional random wet-lab screening methods. Computational methods for virtual screening are many and varied but are usually described as being receptor- (docking) or ligand- (fingerprints / pharmacophore) based in nature. We present the concept of feature-rich or extended pharmacophore ligand-based site models, which can be produced through simultaneous superposing of multiple flexible ligands using our proprietary algorithm Quasi2™. Site models are then used in virtual high-throughput screening (vHTS) of corporate, commercially available or virtual compounds. The Quasi2 approach is not limited to traditional ligand-based design problems, but has equal application and success in challenges where the receptor structure is known. On an eighty-node compute farm the software is capable of screening up to 2 million compounds per day considering both ligand conformational flexibility and alternate protonation states. Enrichment rates are comparable to standard commercial docking or pharmacophore matching softwares. Screening results from both in silico validation exercises and from a number of internal discovery programs highlight the utility and efficacy of extended pharmacophore searching in vHTS.