Computational Approaches to Small Bioactive Molecule Discovery
Novel ligand discovery is a critical early step in drug discovery, and as a result much effort and millions of dollars are spent annually finding new small molecules that bind to proteins. Over the past 15-20 years, the dominant approach has been high throughput screening (HTS), in which a huge library of molecules is tested against each biological target using robotics. This approach has been both spectacularly expensive and surprisingly unsuccessful, at least as measured by the low number of approved new chemical entities.
Another approach to novel ligand discovery is molecular docking or virtual high throughput screening. Whereas it is perhaps the most practical approach to leverage the rapidly growing number of protein structures for ligand discovery, the technique retains important liabilities that make it challenging to deploy on a large scale. We have therefore created an expert system, DOCK Blaster, to investigate the feasibility of full automation1. The method requires a Protein DataBank (PDB) code, sometimes with a ligand structure, and from that alone can launch a full screen of large libraries. A critical feature is self-assessment, which estimates the anticipated reliability of the automated screening results by the ability to recapitulate known structure and binding data. From over 4000 proteins tested so far, over half yield results that look compelling enough to us that we would (and do) commit our own time and money to purchase and test molecules.
A second computational approach to ligand discovery harks back to classical medicinal chemistry of the past century, where the maxim was, "To discover a drug, start with a drug…". We have developed the Similarity Ensemble Approach, a statistical method of comparing any drug with ligands annotated for particular biological targets2. In this way we obtain predictions of biological activity, many of which are known, but some of which are surprising. With our collaborators, we have now had over 30 of these predictions tested, leading to over 25 new activities of existing drugs3.
Irwin, J.J. et al. Automated docking screens: a feasibility study. J Med Chem 52, 5712-20 (2009).
Keiser, M.J. et al. Relating protein pharmacology by ligand chemistry. Nat Biotechnol 25, 197-206 (2007).
Keiser, M.J. et al. Predicting new molecular targets for known drugs. Nature 462, 175-81 (2009).
John Irwin is Adjunct Associate Professor of Pharmaceutical Chemistry at the University of California San Francisco. John received his Ph.D. in organic chemistry at ETH Zurich. He has worked at the MRC Laboratory for Molecular Biology in Cambridge, England, at the European Bioinformatics Institute, and at a startup software company. For the past decade he has been working on computational ligand discovery, first at Northwestern University Medical School and since 2003 at UCSF. He creates and curates software tools and databases for ligand discovery including:
ZINC, a free database of commercially available compounds for virtual screening
DOCK Blaster, a free target based virtual screening web service
DUD, a dataset for benchmarking target-based virtual screening methods and
SEA, the Similarity Ensemble Approach for predicting biological activity of molecules.