An AI-Built Lung Drug May Have Also Reversed 3 to 4 Years of Aging | Edition 341

Last year, Insilico Medicine ran a mid-stage clinical trial testing a new pill for idiopathic pulmonary fibrosis (IPF) — a fatal lung disease with no cure and a median survival of three to four years after diagnosis. The trial worked well enough to move the drug, called rentosertib, into Phase III testing.

Last month, the same 42 patients’ blood samples got a second look, and that’s why “AI drug reverses aging” is the story making the rounds instead of “AI drug treats rare lung disease.” Six independently built AI models, none of which shared data or code with each other, all agreed that patients on rentosertib had a lower predicted biological age after treatment than patients on placebo.

Here’s what’s actually new, what the numbers really say, and what a Nobel laureate who reviewed the data thinks you should — and shouldn’t — take from it.

The part that’s genuinely a first

Rentosertib isn’t AI-flavored marketing. Insilico used one AI system to scan biology and disease data and identify a target — a protein called TNIK, which scored highly across six separate “hallmarks of aging” tests and was already tied to fibrosis biology. A second AI system, Chemistry42, then generated the actual molecular structure to hit that target. No human chemist started from an existing drug and modified it. That combination — an AI-found target and an AI-designed molecule, together, reaching Phase III — has never happened before. It took about 18 months from target identification to a nominated drug candidate; traditional drug discovery routinely takes years for that step alone.

What the new data actually measured

The original Phase IIa trial (published in Nature Medicine, June 2025) tested four groups over 12 weeks: 30 mg once daily, 30 mg twice daily, 60 mg once daily, and placebo. It hit its safety goal and showed a dose-dependent trend on forced vital capacity (FVC) — a standard lung-function measure that also declines naturally with age. In the 60 mg once-daily group, FVC improved by an average of 98.4 mL, versus a 20.3 mL decline on placebo.

That same trial also collected blood samples at four time points, measuring 2,841 proteins per sample. Those samples sat in a public database until a team from Harvard, Stanford, the Broad Institute, Peking University and others ran them through six separate “proteomic aging clocks” — AI models trained to estimate a person’s biological age from blood protein patterns, each built independently, on different training data, by different labs.

All six agreed: patients on rentosertib trended toward a younger predicted biological age than patients on placebo. The strongest signal showed up in a different dose group than the one with the best lung results — the 30 mg twice-daily arm, at week 4 — where several of the six clocks estimated roughly three to four years of reversal, and one estimated close to six. The researchers point out that the dose that helped lungs the most wasn’t the dose that moved the aging clocks the most — evidence, they argue, that the aging effect isn’t just a side effect of healthier lungs.

What it doesn’t prove

Michael Levitt, the 2013 Nobel laureate in Chemistry who reviewed the study, put it plainly: the trial can’t yet tell the difference between “this person is aging more slowly” and “this person’s lungs got better, and that shows up in their blood.” He said the real test is the same drug in healthy volunteers with no lung disease at all — a study that hasn’t happened yet.

There’s also the ordinary caution that applies to almost any early biotech headline: 42 people is a small sample; every one of them already had a serious disease, with a mean age of 67; the “aging clocks” measure blood proteins correlated with age, not lifespan or organ health directly; and the paper’s lead author is Insilico’s own co-founder and co-CEO, Alex Zhavoronkov. None of that makes the finding fake. It does mean “reverses aging” is doing a lot of work in a headline built on one small, company-run study.

Why it matters anyway

The real value here might not be rentosertib itself — it’s the template. Regulators have been looking for a way to fold “geroscience” biomarkers into ordinary disease trials, rather than requiring a separate, multi-decade study to prove a drug extends healthy life. This trial is a working example: aging biomarkers, collected as a side measurement inside a trial that was already happening for an unrelated disease, at effectively zero extra cost to the company. If that pattern holds up, it could shorten the timeline for finding — and proving — drugs that target aging itself, instead of waiting for old drugs like metformin or rapamycin to get repurposed after the fact.

It’s also a business story, not just a science one. Insilico posted about $106 million in revenue in the first half of 2026 — up 287% year over year — and turned its first profitable half-year since going public in Hong Kong last December. The company is now sitting on roughly $7.3 billion in announced deal value this year alone. The aging data didn’t cost them a new trial; it cost them re-running six existing AI models against data they already owned.

Rentosertib is now in Phase III trials for IPF in China — that’s the approval path that actually matters first. If it succeeds there, it becomes the first fully AI-designed drug to reach the market, on the strength of treating a lung disease, with an aging-biomarker story riding along as a genuinely interesting open question rather than a settled answer.