Insilico's AI-designed drug appears to reverse aging markers in trial

Insilico Medicine, an AI-driven pharma company founded in Hong Kong in 2014, developed rentosertib as a treatment for idiopathic pulmonary fibrosis (IPF), a disease that scars lung tissue. A trial last year with 42 patients showed improved lung function, and researchers also collected blood samples for protein analysis along the way. A new analysis of that data, published in Nature Biotechnology, reports that the drug may additionally reverse markers of biological aging.

Researchers ran six independent AI aging clocks, built by separate teams at Harvard, Oxford, Beijing and Insilico itself, on the patients' blood protein data. All six predicted a lower biological age for treated patients than for those on placebo. The strongest effect across the clocks was a drop of three to four years by week 4, with one clock showing a reduction of as much as six years. The article notes this does not mean patients actually got measurably younger, only that their blood protein patterns shifted in ways the models read as a younger biological age.

Michael Levitt, a Nobel laureate in chemistry, said in an Insilico Medicine press release: "What convinces me is not the size of the effect but the agreement, because these models share neither their features nor their training data." The dose that helped lung function most, 60 mg once daily, differed from the dose that produced the largest drop in predicted biological age, 30 mg twice daily, which the report says suggests an effect at least partly independent of lung function. Researchers also compared the treated patients' blood proteins against more than 55,000 profiles from the UK Biobank, a database tracking how proteins typically change with age; according to the company, rentosertib reversed exactly the changes that database shows occurring with aging. The company describes these correlations as no proof but as showing "potential."

Outside experts remain cautious. Cardiologist Eric Topol, who is heavily invested in longevity research, told the New York Times: "This drug looks encouraging. But we do not yet have a definitive trial to make the final judgment." Vadim Gladyshev of Harvard Medical School pointed to the small sample size and the fact that biological clocks are not always reliable, while also telling the Times it is "the first study that shows, very clearly, that predicted biological age can be reduced." He said what is missing is a trial in healthy people, since the current results may only apply to patients who have the lung disease.

Insilico used two AI systems to develop the drug candidate: one searches health data and scientific literature for disease-relevant proteins, and the other analyzes their structure and generates matching molecules. That process led Insilico to the protein TNIK as a target for both aging and pulmonary fibrosis. Going from target protein to drug candidate took about 18 months, and rentosertib is now in a Phase III trial for IPF, the final clinical stage before potential approval. Pharma company Eli Lilly recently invested in the publicly traded Insilico Medicine to help bring AI-co-developed drugs to market. Founder and CEO Alex Zhavoronkov says Insilico has developed at least 28 drug candidates using generative AI as of March 2026, many now in clinical trials.

Key facts

Six independent AI aging clocks, built by teams at Harvard, Oxford, Beijing and Insilico, all read treated patients' blood protein patterns as biologically younger than the placebo group's, by up to six years.
The strongest effect across clocks was a drop of three to four years by week 4 of the 42-patient IPF trial.
The dose that most improved lung function (60 mg once daily) differed from the dose that most reduced predicted biological age (30 mg twice daily), a gap the researchers say points to an effect at least partly independent of lung function.
Treated patients' blood proteins were compared against more than 55,000 UK Biobank profiles, and the company says rentosertib reversed exactly the protein changes typically associated with aging.
Going from identifying the target protein TNIK to a drug candidate took Insilico's AI systems about 18 months; rentosertib is now in Phase III trials for IPF.

Why it matters

This is one of the first cases where an AI-designed drug, built for one disease, shows a separate signal on a marker widely used as a proxy for aging itself, biological age as read by AI models rather than a disease outcome. It adds evidence, still preliminary, that anti-fibrotic drug mechanisms could carry effects beyond their original target, and it showcases Insilico's two-AI drug discovery pipeline (target discovery plus molecule generation) as a working approach that has already reached Phase III trials.

Who it affects

Directly, patients with idiopathic pulmonary fibrosis, the population rentosertib was designed for and the only group tested so far. More broadly, it affects the longevity research field, which gains a data point from a real clinical trial rather than a lab or animal study, and pharma companies watching AI-driven drug discovery, including Eli Lilly, which has already invested in Insilico Medicine.

How to use it

There is no consumer product here: rentosertib is an investigational drug still in Phase III trials for IPF, not approved for any use and not tested in healthy people. Nothing in the source describes pricing, availability or a path for people without the lung disease to access it.

How solid is it

The analysis is published in the peer-reviewed journal Nature Biotechnology and draws on agreement across six independently built AI aging clocks that share neither features nor training data, which outside expert Michael Levitt calls the convincing part. But it rests on one 42-patient trial, and the source does not say whether the biological-age reductions were statistically significant, only that all six clocks pointed the same direction.

Risks and caveats

Experts quoted in the piece flag real limits: cardiologist Eric Topol says the drug looks encouraging but there is no definitive trial yet for a final judgment, and Harvard's Vadim Gladyshev notes the sample size is small, that biological clocks are not always reliable, and that no trial has been run in healthy people, so the results may apply only to patients who already have the lung disease. The article itself cautions that a lower predicted biological age does not mean patients got measurably younger, only that their blood protein patterns shifted in a direction the AI models read as younger; the UK Biobank comparison is described by the company itself as no proof, only potential.

“What convinces me is not the size of the effect but the agreement, because these models share neither their features nor their training data.”

— Michael Levitt, Nobel laureate in chemistry, in an Insilico Medicine press release