Virtual ligand screening followed by in vitro experiments is recognized as a promising approach in drug discovery research. The Genome Sciences Centre drug discovery platform is a high-throughput virtual screening facility employing state-of-the-art molecular docking software on a 400-CPU computer cluster. The GSC has the capability to virtually screen over 1 million compounds per day. This represents one of the largest international academic resources for in silico drug discovery research.

Technology

The drug discovery platform uses both free and commercial software including Dock, Autodock and ICM. We have licensed ICM, which has been assessed as one of the best commercial software for molecular docking, for 200 CPUs.

A three-dimensional structure of the drug target is the start point of virtual screening. When no experimentally (X-ray, NMR) solved structure is available, comparative homology modeling is the most reliable approach for structure prediction. In this approach, accurate native conformation prediction of structurally variable regions (SVRs), or loops, is a recognized bottleneck. We have developed one of the best algorithms for loop prediction based on the ICM package, which can predict loops up to 10 residues long with crystal structure accuracy (RMSD below or close to 1.0 Å).

Ligand binding pocket prediction

Prediction of putative small molecule binding sites within proteins that are drug targets is a key technology of large scale virtual screening. We have developed a new computational algorithm for the accurate identification of ligand binding sites. This ICM-based algorithm, called PocketFinder, has been validated on a large, systematically collected data set of 5616 known binding pockets from complexed structures of PDB. The result shows over 96% of binding sites are successfully predicted.

Pipeline

We have developed two pipelines: a docking pipeline for high-throughput virtual screening of large compound databases against specific protein targets, and an inverse docking pipeline for docking specific compounds against large protein target databases.

Database

We have created an in-house database with over 5 million compounds from various public databases (PubChem, Zinc, DrugBank, some literature) for our docking pipeline. We have also prepared a large in-house database of drug binding sites from therapeutic proteins and other human proteins for the inverse docking pipeline.

Projects

The drug discovery platform is currently involved in a number of collaborative projects, screening several drug targets for novel inhibitors.

W are also conducting a large-scale computational drug repositioning analysis based on known drugs and known disease targets. The goal of this research is to discover new indications for available drugs and/or obtain insight about the adverse effects of these drugs. We are also specifically applying this analysis to a smaller data set consisting of the human protein kinome.

We actively welcome academic collaborations involving the application of this platform.

Jianghong An

Staff Scientist
Back to top