
Bristol Myers' Richard Hargreaves pays $70M to launch a neurodegeneration alliance with a star player in the machine learning world
Bristol Myers Squibb is turning to one of the star upstarts in the machine learning world to go back to the drawing board and come up with the disease models needed to find drugs that can work against two of the toughest targets in the neuro world.
Daphne Koller’s well-funded insitro is getting $70 million in cash and near-term milestones to use their machine learning platform to create induced pluripotent stem cell-derived disease models for ALS and frontotemporal dementia.
Then they’ll use those insights to start building new drugs for those two ailments; a complex, ground-up approach that has already won a close alliance with Gilead.
Success would trigger up to $2 billion in milestones, running a gamut of research and commercial goals.

“We believe that machine learning and data generated by novel experimental platforms offer the opportunity to rethink how we discover and design novel medicines,” said Richard Hargreaves, the chief of the neuro group at Bristol Myers, who made the leap from Celgene.
Koller’s been making great strides with a new technology that has gained immense interest, but still has a long way to go to prove itself as the major players start figuring out how to integrate artificial intelligence and ML into their game plans for drug discovery and development.
As she told me in interviews for the Endpoints 11 awards, the field has been intensely hyped by a slew of new players. From her perspective, the tech holds lots of promise, but there’s a tremendous amount of groundwork that has to be done to glean the necessary data for it to work. And it starts with the right disease models.
I’ve looked for instance at cellular phenotypes and I’ve built a model that aligns those with human clinical outcome: Can I take a group of genetic backgrounds that I’ve never seen before and predict clinical outcomes for those patients? That is a confidence building measure. It says I built a disease model that is actually predictive of what we see in humans. By the way, animal models, they don’t even get asked that question in a lot of cases, which I find rather shocking.
On top of that, the ex-Stanford professor has been frank about understanding the tech better than drug development, a weakness she’s been working on by recruiting top talent from that other side of the R&D dimension. That strategy recently led to the arrival of Merck R&D chief Roger Perlmutter on the board.
Koller told me:
The early days were very challenging, I won’t lie. I was also actually new to the space…To build something from scratch you need a network, you need to be connected to people that you can recruit, you need to ask advice. When I came into this I had none of that. So it was an incredibly challenging thing to just start from a blank slate in a space where you were a newcomer and figure out even how to go about building a wet lab that I had never built before.
Now she’s working with Hargreaves, one of the top neuroscientists in the field, as Koller and her team continue to build up the company.