Drug discovery startup Insitro burst onto the scene last year with ambitions of using machine learning techniques to find new therapies. Now the company has its first pharmaceutical partner. Gilead Sciences will work with Insitro to find medicines to treat a liver disease, nonalcoholic steatohepatitis (NASH), that is fast becoming a global epidemic fueled by poor diet and lack of exercise.
Insitro announced Tuesday that Gilead (NASDAQ: [[ticker:GILD]]) will pay the South San Francisco-based startup $15 million to kick off a three-year research collaboration. The deal calls for Insitro to create disease models for NASH in hopes of finding targets for drugs meant to slow or reverse its progression.
The 30-person company—which also revealed Tuesday that its Series A, the total amount of which was not disclosed when it closed last year, clocked in at more than $100 million—says it will combine elements of data science, human genetics, and functional genomics to advance toward this goal. The funding came from an assortment of high-profile biotech and hard tech investors.
When Insitro came out of stealth last May, CEO and founder Daphne Koller (pictured), a Stanford University professor and longtime machine learning researcher, said the company hoped to take advantage of advances in ML, including improved algorithms, but also the increasing availability of massive sets of data to help unearth new drugs. It isn’t alone; several startups have similar ambitions.
“Our hope at Insitro is that big data and machine learning, applied to the critical need in drug discovery, can help make the process faster, cheaper, and (most importantly) more successful,” she wrote at the time.
Koller was previously chief computing officer at Calico, the anti-aging research outfit at Alphabet (NASDAQ: [[ticker:GOOGL]]), Google’s parent company. Prior to that, she co-founded Coursera, a provider of online learning via massive open online courses.
Koller says what sets Insitro apart from the other startups promising to use advanced computing tools to improve drug discovery is its ability to create the enormous data sets necessary to feed its software’s algorithmic appetites. It can create those data sets thanks to its massive financial support.
“The large majority of companies out there start out from data sets that already exist, that were, invariably, not created for the purpose of machine learning. They’re usually the byproducts of whatever activities happen to be ongoing, and then people kind of try and bring those together, massage them into some reasonable form and glean something from those,” she said. “What we’re doing is taking the exact opposite trajectory, which is to say, let’s figure out what are the really important problems that we would like to solve in the drug development process, which are the ones for which machine learning could, in fact, provide good predictions … and which are the ones for which we are then able to create large data sets that will drive machine learning.”
The future nerve center of the company’s efforts is currently under construction: Insitro is building out a high-throughput “bio-data” factory, which Koller says will use sophisticated robotics to speed up the target discovery process through automation.
Another startup that also aims to use AI to improve drug R&D, San Diego’s Erasca, recently announced it extended its Series A to $64 million. At least two of its newest investors, Chicago-based Arch Venture Partners, Silicon Valley’s Andreessen Horowitz (a16z), are also investors in Insitro.
Last year Koller’s company said that its work was being funded by a16z, Arch, Foresite Capital, GV (formerly Google Ventures), and Third Rock Ventures. But others are also backing it, the company said Tuesday, including Alexandria Venture Investments, Bezos Expeditions (Amazon chief Jeff Bezos’s personal VC fund), Mubadala Investment Company, Two Sigma Ventures, and Verily, another health-focused Alphabet subsidiary.
Under the deal announced Tuesday, Foster City, CA-based Gilead has the right to move forward up to five targets identified by Insitro, which could earn up to $35 million more in “near-term payments” if it hits certain operational milestones. It could also get up to $200 million per target Gilead selects, based on each drug’s progress. If any of the NASH drugs discovered by Insitro reach the market, Gilead would pay its partner royalties from sales.
The deal also gives Insitro the option to sign on to develop one or more of these potential drugs with Gilead in the US. If it does so, it would get a cut of drug sales in China, plus milestone payments and royalties from sales in other countries outside of the US, according to the companies.
The deal with Gilead adds Insitro—which Koller says will more than double its current headcount by year’s end—to the crowded field of companies that are vying to come up with new ways to treat NASH. The only existing treatment for patients whose disease has advanced to the stage where the organ no longer functions is a liver transplant. The first could come from Intercept Pharmaceuticals (NASDAQ: [[ticker:ICPT]]), which said at a medical meeting last week it is on track to ask US and European regulators to approve its NASH drug, a once-a-day pill called obeticholic acid, this year.
In addition to Intercept and Gilead, which is already testing three experimental NASH drugs, Allergan (NYSE: [[ticker:AGN]]), 89Bio, San Diego’s Viking Therapeutics (NASDAQ: [[ticker:VKTX]]), Madrigal Pharmaceuticals (NASDAQ: [[ticker:MDGL]]), Akero Therapeutics, and others are also racing to develop treatments.
Koller tempered expectations about what ML-based techniques can do today for drug discovery and development, emphasizing that Insitro’s technology won’t be truly validated until a compound becomes a drug that provides a therapeutic benefit to people. There is a lot of hype when it comes to the potential impact of data science techniques on biopharma R&D, she says.
“When we have benefit in actual patients, we will be happy,” she said.
