Virtual high-throughput screening: How to boost the pharmacophore approach


Frédérique Barbosa, Molecular Modelling, Molecular Modelling, Cerep, 128, rue Danton, 92500 Rueil Malmaison, France
The virtual screening of large libraries is structure-based or ligand-based. Structure-based virtual screening (docking) is time consuming and requires a precise knowledge of the 3D structure of the target and of the various binding contributions. For ligand-based screening, the only determining step is the appropriate choice of descriptors and similarity metrics. A careful pharmacophoric description of the 3D conformers of the ligands is required to take into account the features responsible for binding. We use such an approach to retrieve new lead compounds from large and chemically diverse virtual libraries. We have built internally a virtual library of 108 chemically feasible compounds from >7000 building blocks (>1500 proprietary) using validated chemistries. We use a so-called “ghost database” mechanism that allows for the fast calculation of multiple conformers and pharmacophoric fingerprints for each individual structure in the database. Therefore extensive virtual screening with pharmacophores can be done within tractable CPU time. The retrieved compounds are further analyzed by predicting ADME-T properties with Cerep proprietary QSAR models based on BioPrint® (>2000 drug and drug-like compounds tested across >170 in vitro assays). This two stage approach accelerates the identification of promising chemical structures for early drug discovery.