Daphne Koller, A-list backers unite to create machine learning drug R&D startup

Investors have signed up in the belief that Daphne Koller can translate machine learning advances into biotech. (World Economic Forum/CC BY-SA 2.0)

Daphne Koller has unveiled her next goal: improving drug R&D. The ex-Calico AI expert plans to enlist data and machines to address shortcomings in drug discovery and development—and she has persuaded a stellar investment syndicate to bankroll the work. 

Koller disclosed the plan and the startup founded to enact it, insitro, in a blog post that outlines the broad concept without going deep into the specifics. What is known suggests insitro will serve as an interesting test case for the readiness of machine learning to improve drug R&D, a possibility that to date has been far more discussed than realized.

On paper, insitro is well placed to push machine learning-enabled drug R&D to its limits. Koller has been making waves in machine learning for more than two decades and has been applying her skills to biomedical research since the early 2000s, so the science side of the operation is in good hands.

INDUSTRY RESEARCH

Veeva 2019 Unified Clinical Operations Survey

Share your thoughts in this quick survey. The first 50 qualified respondents will receive a $5 Amazon gift card, and, in appreciation for your time, all qualified survey respondents will be entered in a drawing to win a $50 Amazon gift card.

And the investor syndicate reads like a who’s who of major biotech and tech VC shops.  ARCH Venture Partners, Foresite Capital, a16z, GV and Third Rock Ventures are all on board. 

The investors have signed up in the belief that Koller can translate machine learning advances into biotech. In recent years, breakthroughs at Alphabet’s DeepMind and elsewhere have reset expectations of what machine learning will do in the near term. But biology makes Go, the game DeepMind’s AI aced, look simple. 

In contrast to some others in the space, Koller is quick to acknowledge machine learning won’t provide a quick fix for all of drug development’s problems or magic ideas for breakthrough drugs out of data alone. But she does think the technology is mature enough and the datasets big enough for ML to start making a dent in some thorny challenges. 

“We plan to collect and use a range of very large data sets to train ML [machine learning] models that will help address key problems in the drug discovery and development process. The ML models that are developed will then help guide subsequent experiments, providing a tight, closed loop integration of in silico and in vitro methods.” Koller wrote. 

Koller established her reputation in machine learning at Stanford University, before going on to apply her skills to the broadening of access to education via online courses at Coursera. Calico came calling in 2016 and persuaded Koller to take up the post of chief computing officer. That put Koller in charge of building a team and methods capable of extracting insights into aging from biological data.