COMPANY TO WATCH · De Novo Pharmaceuticals identifies novel compounds right before your eyes.
By Kevin Davies
November 19, 2004 | "When it comes to screening targets, very little fundamentally has changed in 20 years," explains Bart Wuurman, CEO of De Novo Pharmaceuticals. "You buy a library of compounds, you screen it against a target, and see what sticks. If something sticks, then it's left to the magic of the medicinal chemists to turn that into a lead compound for further development."
Wuurman wants to bring some intelligence to the problem of drug discovery. "Rather than look for a needle in a haystack, we create the needle," says Wuurman, formerly the business development director at Antisoma.
De Novo was co-founded in 1999 by chief scientific officer Philip Dean, after about 30 years in the University of Cambridge's Department of Pharmacology, where he ran the Drug Design Group. Dean motions to a nearby laptop that vividly illustrates the in silico "de novo design" process of SkelGen, the company's design technology platform.
Onscreen, a procession of 1,700 chemical fragments, designed in conjunction with Hoffman-LaRoche, is randomly built into compounds that nestle into the binding site of a blood-clotting factor. The fit of each molecule is calculated based on various criteria, including hydrogen bonding, hydrophobic and steric effects, and so on. "If you have a crystal structure, these algorithms design [ligand] structures very rapidly," Dean says.
VIRTUAL FIT: An ATP-like molecule binds in silico to DNA gyrase.
Indeed. The total fragment library can be assessed in a matter of minutes to about half an hour, depending on the structure of the target site. Using an 80-node SunSPARC farm on a grid system, the algorithms then design and produce a ranking of the best molecules, of which De Novo says 90 percent are both synthesizable and patentable.
Overnight, Dean says, SkelGen can design hundreds of candidate structures. "If a molecule doesn't meet the criteria for optimality, it's automatically thrown out. We're only creaming off the best structures. Within that best set of structures, we re-order them for feasibility of chemical synthesis."
But a target crystal structure is not essential. Ligand-binding pockets can also be modeled in cases where actively binding molecules have been identified but a crystal structure is lacking. In this process, called "ligand-based design," De Novo generates druglike molecules with more desirable properties and different scaffolds.
"We can take high-throughput screening (HTS) hits," Dean says, "and build a model of the [target binding] site by overlaying the structures in terms of their molecular similarity ... De Novo evaluates each model by virtual HTS against a test set of ligands to determine the accuracy. The best model is fed into SkelGen as though it was a site. SkelGen then creates novel structures" that fit this pseudo-target protein (see Lloyd, D.G. et al. J. Med Chem 47, 493-96; 2004).
Another application is scaffold hopping, as Dean explains: "If you have an actual compound, and there are some problems — for example, toxicity or ADME properties — then we can redesign the scaffold. Those structures are equivalent, although they're different structures."
|DE NOVO DETAILS
Location: Cambridge, England
CEO: Bart Wuurman
No. of Employees: 24
Funding: $30 million
(Merlin Life Sciences, Avlar Bioventures, Life Sciences Partners, Roche)
De Novo's first Big Pharma partner was Hoffman-La Roche. Wuurman says Roche was "amazed" that De Novo could not only reproduce inhibitor structures for four well-known pharmaceutical targets but also identify novel lead molecules (see Stahl, M. et al. J. Comp. Aid. Mol. Design 16, 459; 2002). Roche subsequently invested about $3 million in De Novo, receiving a license in return.
De Novo signed a similar deal with Eli Lilly at the beginning of 2004, and Wuurman says he's in discussion with a couple of other big pharmas about technology licenses and with biotechs about drug discovery partnership deals on specific drug targets.
Meanwhile, De Novo, in partnership with Peakdale Molecular Ltd., has begun to create and synthesize a new range of specialized compound libraries. The first is a set of more than 1,000 diverse compounds that target G-protein-coupled receptors, selected from more than 15 million virtual structures.