GlaxoSmithKline recruits PathAI to bring digital pathology AI to cancer, NASH drug development

After making its first foray into nonalcoholic steatohepatitis (NASH) drug development with a $1 billion RNAi therapy partnership last fall, GlaxoSmithKline is continuing to carve its own path in the race to develop the first FDA-approved NASH treatment.

Fittingly, its next step on its journey is to team up with PathAI, the Boston-based Fierce 15 honoree that’s developing artificial-intelligence-powered digital pathology tools. GSK and PathAI have signed a multiyear pact that will see them joining forces to improve the drugmaker's molecule discovery and clinical trial work not only in NASH, but across its oncology portfolio as well.

They’ll deploy GSK’s computational pathology algorithms on PathAI’s platform to support clinical trials and use PathAI’s own algorithms to advance the Big Pharma’s clinical development targeting NASH, PathAI CEO Andy Beck, M.D., Ph.D., told Fierce Medtech.

“Success will result in the development of novel diagnostics to support GSK’s assets,” said Beck, who added that the duo could ultimately extend their partnership beyond oncology and NASH, though they’ll continue to focus on disease areas that stand to benefit the most from AI and machine learning.

Beck noted that PathAI chose GSK as a partner based largely on its wealth of experience applying AI to drug discovery and clinical trials.

“The right partnerships allow us to lend our expertise in implementing AI-powered tools where they can be most valuable; GSK has the structure, pipeline and internal capabilities to allow us to co-develop models that can be leveraged in clinical trials and help deliver improved treatment options for patients,” the CEO said.

For its part, GSK saw in PathAI a partner that could help take its own work in AI, machine learning and computational pathology to the next level.

“We rapidly reached the point where we needed a partner who understood the potential of computational pathology the way we did and had several key components we lacked, including a validated environment to deploy into the clinic,” Kim Branson, GSK’s global head of AI and machine learning, told Fierce Medtech.

“PathAI was clearly able to take the models we already built and take them further. We had looked at their work in NASH and knew they were a world-class company,” Branson said. “PathAI had also solved some critical pain points in the development of computational pathology, and last, the annotation of pathology slides with their managed crowd of pathologists is world-class.”

PathAI’s digital tools use AI and deep learning technology to analyze tissue samples, looking for condition-specific biomarkers that can predict how a disease will progress and how an individual patient might react to various treatments.

The startup’s work with GSK will take advantage of several of those tools, including AIM-NASH, a model that’s already being used by many of GSK’s fellow Big Pharmas and, in fact, was developed in partnership with Gilead Sciences.

In a presentation (PDF) last summer, PathAI and Gilead explained how they trained the AI using nearly 6,000 annotated biopsies from NASH patients participating in several late-stage clinical trials. The resulting model produced biopsy analyses that outperformed a blinded central reader in how they lined up with the conclusions of a panel of expert hepatopathologists.

With that success, PathAI said at the time, AIM-NASH could therefore serve as a guide as researchers recruit NASH patients to clinical trials and identify the biomarkers that will determine the clinical endpoints for those trials, while also helping assess each participant’s response to the experimental therapies. Since then, the tool has been used by Novo Nordisk and Bristol Myers Squibb in their own studies of potential NASH treatments.

Just as PathAI has racked up a long list of pharmaceutical partners—a list that also includes Merck, an investor in its last two funding rounds, and Roche, which recently integrated PathAI’s tech into its own digital pathology platform—so, too, has GlaxoSmithKline made a habit of recruiting AI developers to boost its drug discovery and development work.

Its past tie-ups have partnered GSK with Nvidia to build the U.K.’s most powerful supercomputer for AI research purposes and with the University of Oxford, using advanced technologies like functional genomics and machine learning to target Alzheimer’s, Parkinson’s and other neurological diseases.