Computational methods are being applied across the drug-development process from the earliest stages of drug design. “We're seeing a lot of companies who don't want to spend time and money on screening existing compound libraries,” says Bart Wuurman, chief executive officer at De Novo Pharmaceuticals, a leading in silico drug-design company in Cambridge, UK. “The traditional approach is looking for a needle in a haystack — what we do is create a needle.”

SkelGen-modelled ligand. Credit: DE NOVO PHARMACEUTICALS

Founded in 1999 as a spin-out from the University of Cambridge's drug-design group, De Novo last year refocused itself from in-house drug discovery to meeting the growing demand for its proprietary computational platforms. De Novo's core platform is a suite of structure-generating programs called SkelGen. The core algorithm generates new molecular structures by the random assembly of small molecular fragments. This stochastic approach rapidly searches the whole of the chemical space of potential drug compounds, explains chief scientific officer Philip Dean. The search is constrained by any information available about the structure of the protein target from X-ray crystallography. In cases where little is known about the target, information about the structure of a known ligand that affects the target's activity can be exploited. The firm's Quasi2 program uses data on the known active ligands to create a negative image of the target site, which can then be input into the SkelGen algorithm to find new patentable compounds with a similar action.

“The unique technology is the ability to create novel chemotypes for any set of constraints,” Dean says. “We think there's no one else out there who can do de novo design from just an active compound.” The firm's researchers have also developed a method for accommodating the flexibility of the protein target, which structure-based molecular design has previously failed to do.

De Novo's platform takes into account the need to convert in silico designs into actual chemicals. “We have an algorithm that post-processes the SkelGen structures to prioritize those structures that are readily synthesizable in the smallest number of steps,” Dean says. “Not only are the compounds scored for the fit with the site but also for ease of synthesis.” In collaborations with Roche, over 90% of output compounds have been relatively simple for chemists to synthesize. “Basically it's our intent to get the designed molecules into wet chemistry as quickly as possible,” Wuurman says.

Tim Chapman