If all goes according to plan, insitro could finally start testing its first drug in clinical trials in 2026, eight years after Daphne Koller founded the AI-powered biotech. To get there, it’s teaming up with Eli Lilly to access a key technology to make its experimental genetic medicine a reality.
The South San Francisco-based startup has identified two genes with its machine learning tools that it believes are linked to metabolic diseases, including fatty liver diseases.
While insitro has designed gene-silencing siRNA molecules to target those genes, it still needs a chemical conjugate called GalNAc to shuttle those treatments into liver cells. Through the deal, insitro will have the option to license Lilly’s GalNAc molecules.
Insitro will also work with Lilly to develop an antibody against a secreted target. For all three programs, insitro is responsible for running clinical tests. There is no upfront money, but Lilly will get milestone and royalty payments if the drugs are successful.
It’s an arrangement that flips the normal order of partnerships between biotech and pharma, which usually sees the bigger company run expensive clinical trials and dole out payments to the smaller company upon positive clinical trial results and FDA approval.
“I think they recognize that even the biggest pharma companies in the world can’t do everything internally,” Koller told Endpoints News in an interview. “We were able to really come to a set of economic terms that are very beneficial to both sides.”
The siRNA drugs are insitro’s most advanced programs. Koller said she hopes to announce the development candidates later this year or early next, then file an IND for the first one within 12 months after that — a timeline that could lead to a clinical trial in 2026. The second siRNA drug will be a few months behind.
AI for new drug targets
Koller is one of the biggest names at the intersection of AI and biology, particularly after raising $643 million across three funding rounds since 2018. Despite that, she’s developing a reputation for warning about the dangers of overhyping AI and insitro has been quiet about the specifics of its research.
The company started to open up a bit more in January, when it published a trio of preprints showing how it used machine learning to predict genetic details about cancer cells just by analyzing stained biopsies; to develop a new screening method for ALS drug discovery; and to identify hundreds of genetic loci linked to metabolic dysfunction-associated steatotic liver disease. That common condition, also called MASLD, is one focus of the Lilly partnership.
The company’s fatty liver disease work is furthest along.
“These are novel targets with strong genetic support,” Koller said of the program. “The genetics of fatty liver disease have been largely uncharted territory, because there’s just not a lot of patients in which the disease has been officially diagnosed.”
One of the January preprints shows how insitro used machine learning to predict liver fat in UK Biobank participants by looking at data from bone density scans and molecules in the blood.
“It turns out that AI could read liver fat quite reliably from that,” Koller said. “We now have liver fat measurements — even if not 100% accurate — not for a few thousand people, but for a few hundred thousand people,” she added.
That helped insitro uncover new genetic links that would have been difficult or impossible to find with a smaller dataset, Koller said. But since the company’s AI advantage is largely centered around finding a new drug target before someone else, she isn’t saying what those genes are just yet. Getting scooped is a big concern.
“As soon as there is a good drug against a biology, you have 20 other companies that are making drugs against that same thing,” Koller said.