British startup BenevolentAI has become the first European firm to install a purpose-built version of the DGX-1, an Nvidia ($NVDA) supercomputer designed for deep learning. BenevolentAI will use the technology in its algorithm-driven drug discovery programs, the potential of which have reportedly netted the startup $100 million in funding and an Alzheimer’s pact worth up to $800 million.
London, U.K.-based BenevolentAI, the biotech wing of which used to be called Stratified Medical, has spent the past few years beavering away on algorithm-enabled drug discovery programs. But, until now, it lacked a supercomputer built specifically for deep learning, an offshoot of artificial intelligence that tries to mimic the pattern recognition and learning that takes place in the neocortex of human brains. Buying the DGX-1 gives BenevolentAI such a supercomputer.
“This new AI supercomputer will boost our processing power and accelerate the creation of new relationships amongst disparate information sources to yield faster scientific innovation,” Ken Mulvany, co-founder and director of BenevolentAI, said in a statement. “For the current application of our AI technology in the bioscience space, it will mean that new drug discoveries can be made faster and more efficiently than ever before.”
BenevolentAI, specifically its biotech-focused unit BenevolentBio, is using its computing power to sift through the scientific literature and medical databases. The idea is to apply deep learning to target validation, lead discovery, drug optimization and other tasks in an attempt to cut the time it takes to formulate ideas about how best to tackle various diseases, although, as is typical with early-stage, tech-enabled biotechs, BenevolentAI is yet to share a detailed look at its approach.
The lack of details is in keeping with the low profile BenevolentAI has kept over the past few years. BenevolentAI claims to have a dozen active drug development programs across four disease areas--it has expressed an interest in neurodegeneration, orphan indications and rare cancers--and a willingness to acquire oral therapies from preclinical toxicology up to the start of Phase IIb. Details of its R&D activities, and the technology that underpins them, are thin on the ground, though.
While such secrecy can provoke skeptical reactions, regardless of whether it is understandable from a business perspective, the information that has leaked out about BenevolentAI suggests it is worth taking seriously.
Mulvany set up BenevolentAI after leading Proximagen to a takeover by Upsher-Smith Laboratories for $347 million upfront and up to $206 million more in milestones. And, while BenevolentAI has said little about its fundraising activities, it claims to have raised approximately $100 million and lists Neil Woodford’s fund, Lundbeck, Upsher-Smith Laboratories and Lansdowne Partners among its backers. Ex-Lundbeck CEO Ulf Wiinberg, who lost his job over an undeclared gift of Stratified Medical shares from Mulvany, also has a stake in the company.
Fueled by the funding, BenevolentBio has been on a hiring spree. Ex-GlaxoSmithKline ($GSK) SVP Jackie Hunter came in as CEO, former Pfizer Neusentis COO became director of drug discovery and John Overington, who has stints at Pfizer ($PFE) and a Galapagos ($GLPG) subsidiary on his résumé, was named as VP of biomedical informatics.
The new hires joined a company that had already landed a deal with an unidentified U.S. drugmaker for two Alzheimer’s targets. A report in the Financial Times valued the deal at up to $800 million, but it is likely to be some time before BenevolentAI sees a sizable slice of that sum, if indeed it ever does. In 2015, the company reported £585,000 ($764,000) in revenue related to the deal.
Earnings from the deal should tick up if the programs continue to advance, but BenevolentAI is likely to face rising costs, too. While the company makes no mention of pipeline programs on its website, it is currently recruiting a clinical project manager to oversee Phase I and II clinical trials.