In 71 IPF patients over 12 weeks, an AI-designed drug produced a 98.4 mL lung function improvement — real signal or noise? The Phase III will decide, but Insilico just signed a $2.75B Lilly deal before a single Phase III patient was enrolled.
image from grok
Insilico Medicine achieved a landmark this week with its AI-designed drug rentosertib showing statistically significant efficacy in a Phase IIa trial for idiopathic pulmonary fibrosis—the first AI-discovered and AI-designed small molecule to do so. The company simultaneously closed a $2.75 billion licensing deal with Eli Lilly and a $293 million Hong Kong IPO, validating investor interest. However, CEO Alex Zhavoronkov cautions that the Phase IIa data (71 patients, 12 weeks) represents early evidence, not proof of efficacy, with the critical gap to a registrational Phase III trial (typically 52+ weeks, hundreds of patients) representing where most drug candidates fail.
The headline from Insilico Medicine's latest milestone was supposed to be about AI finally designing a drug that works in humans. The more honest headline is that Insilico just completed one of the most consequential weeks in AI-drug history and still has everything to prove.
On March 29, Eli Lilly signed a licensing deal with Insilico worth up to $2.75 billion, $115 million upfront, the rest tied to development and commercial milestones, for access to Insilico's preclinical pipeline of AI-designed oral drugs. Three days earlier, the company had closed a $293 million IPO on the Hong Kong Stock Exchange. The week before that, it published Phase IIa results in Nature Medicine showing that rentosertib, the company's AI-designed molecule for idiopathic pulmonary fibrosis, produced a mean 98.4 mL improvement in lung function over 12 weeks compared with a decline in placebo patients. The science is real. The timeline to a real medicine is not yet known.
Rentosertib, formerly known as ISM001-055, is the first AI-discovered and AI-designed small molecule to advance through a peer-reviewed Phase II trial and show a dose-dependent efficacy signal. Insilico used its Pharma.AI platform to identify TNIK (Traf2- and NCK-interacting kinase) as a novel target in IPF and then generated the molecule itself, rather than starting from a known target and optimizing a compound around it. That is unusual. Most AI drug programs work on the chemistry side, finding better molecules for targets that already exist. Insilico claims it went further upstream, using generative models to nominate both the target and the drug candidate.
The Phase IIa data, published in Nature Medicine in June 2025, enrolled 71 IPF patients across 22 sites in China and randomized them to placebo or one of three rentosertib dosing arms for 12 weeks. The primary endpoint was safety, which the trial met. Adverse event rates were similar across arms and most events were mild or moderate. The secondary efficacy signal was the headline number: patients on the highest dose, 60 mg once daily, showed a mean forced vital capacity (FVC) improvement of 98.4 mL compared with a mean decline of 20.3 mL in the placebo arm. FVC is the gold-standard measure of lung function in IPF trials. A 118.7 mL difference between active and placebo over 12 weeks is real. It is also small.
Idiopathic pulmonary fibrosis is a chronic scarring lung disease with no cure and a median survival of three to four years. Current approved antifibrotic drugs can slow progression but do not reverse it. A therapy that actually improved lung function in IPF would be transformative. The problem is the gap between Phase IIa and Phase III. Phase IIa was 71 patients over 12 weeks. A registrational Phase III trial in IPF would typically run for 52 weeks or longer with hundreds of patients. Small, short trials are where biological noise looks like signal. The regulatory path for rentosertib is the subject of active discussions with health authorities, but no timeline for a Phase III start has been disclosed.
The deal with Lilly adds financial runway but also raises questions about who controls the next chapter. Lilly gets an exclusive license to preclinical oral drug candidates in selected disease areas. Reuters reported that the deal includes a GLP-1 receptor agonist for diabetes from Insilico's pipeline, though the companies have not formally disclosed the specific programs. Rentosertib itself remains Insilico's asset. The partnership builds on a prior collaboration the two companies struck in 2023. For Insilico, the Lilly money buys time. For Lilly, it buys optionality on an AI engine that has now demonstrated at least one Phase II signal.
Alex Zhavoronkov, Insilico's founder and CEO, has spent years arguing that the pharmaceutical industry was moving too slowly and that AI could compress timelines dramatically. He is not wrong that the industry needed new tools. He is also, by his own admission, a man who has been saying so for a long time without a registrational dataset to show for it. Rentosertib is that dataset's opening argument. The Phase III will be the closing one.
What to watch: whether Insilico can initiate a Phase III trial this year, what the trial design looks like, and whether the 98.4 mL signal survives contact with a larger, longer, more rigorous study. IPF is a graveyard for promising Phase II results. So is AI drug design, for that matter. The next twelve months will tell us whether rentosertib is the proof that generative AI can do this, or whether it was the most expensive slide deck Insilico ever produced.
Story entered the newsroom
Assigned to reporter
Research completed — 0 sources registered
Draft (663 words)
Reporter revised draft (771 words)
Reporter revised draft (777 words)
Reporter revised draft (756 words)
Reporter revised draft (762 words)
Approved for publication
Headline selected: Insilico's AI Drug Works. CEO Warns: That's Not the Same as a Cure
Published (757 words)
Get the best frontier systems analysis delivered weekly. No spam, no fluff.
Biotech & Life Sciences · 18h 26m ago · 4 min read
Biotech & Life Sciences · 1d ago · 4 min read
