AI-Designed Drug Reverses Biological Aging in First Human Trial

Every few months, a story emerges that quietly rewrites an assumption everyone had stopped questioning.

One of those assumptions was that AI-designed drugs were a cost-saving gimmick -- faster, maybe, but not fundamentally different from what human chemists could discover. Another assumption was that biological aging, at least in humans, could not be meaningfully slowed by a single molecule.

Both assumptions just collided with a single data point.

Insilico Medicine, a Hong Kong-based AI-driven biotech company, reported that patients taking its AI-designed drug rentosertib showed blood protein patterns consistent with being biologically up to six years younger than the placebo group. The findings were published in Nature Biotechnology -- and the study used six independent proteomic aging clocks to reach its conclusion.

What the trial found

Rentosertib was originally developed for idiopathic pulmonary fibrosis (IPF), a progressive lung disease that scars tissue and impairs breathing. A Phase IIa trial last year with 42 patients showed improved lung function. But the researchers had also collected blood samples for protein analysis -- and that data, analyzed with AI aging clocks, produced the unexpected result.

The six aging clocks, built by separate teams at Harvard, Oxford, Peking University, and Insilico itself, all predicted a lower biological age for treated patients compared to placebo. The strongest effect appeared at week 4 in the 30 mg twice-daily group, with one clock showing as much as six years of age reduction.

"What convinces me is not the size of the effect but the agreement," cardiologist and aging researcher Eric Topol told The New York Times, "because these models share neither their features nor their training data."

Why six clocks matter more than one

Proteomic aging clocks are AI models trained on thousands of blood protein profiles to predict a person's biological age -- the age your body looks on a molecular level, as opposed to your chronological age. Different clocks use different protein panels, different training populations, and different algorithms. If they all converge on the same direction, the signal is harder to dismiss as coincidence.

The researchers also compared treated patients' blood protein patterns against more than 55,000 profiles from the UK Biobank, a database that tracks how specific proteins change with normal aging. According to Insilico, rentosertib reversed exactly those age-associated protein changes.

The caveats that matter

Cardiologist Eric Topol, while describing the results as encouraging, cautioned that "we do not yet have a definitive trial to make the final judgment." Harvard's Vadim Gladyshev pointed to the small sample size and the fact that biological clocks are not always reliable. He still called the study "the first study that shows, very clearly, that predicted biological age can be reduced."

A key question is whether the apparent age reversal is a direct effect of the drug or a side effect of treating the lung disease. The dose that helped lung function most (60 mg once daily) differed from the one that cut predicted biological age most (30 mg twice daily), suggesting at least partially independent mechanisms. But the trial was small, and the effect was measured in sick patients, not healthy people.

"From the data we have, it's not entirely clear whether the effect is from the direct anti-aging action of the drug or from the improvement in the disease," Gladyshev told The New York Times. What is needed, he said, is a trial in healthy people.

How AI designed the drug

Insilico's AI platform works in two stages. The first stage, PandaOmics, combs through health data and scientific literature to identify disease-relevant proteins. The second, Chemistry42, analyzes their molecular structure and generates candidate molecules. Together, the systems identified TNIK -- a protein kinase involved in both fibrosis and aging -- and designed rentosertib to inhibit it.

From target identification to drug candidate took about 18 months. The molecule is now in a Phase III trial for IPF, the final clinical stage before potential regulatory approval.

Insilico founder and CEO Alex Zhavoronkov has stated that the company has developed at least 28 drug candidates with generative AI as of March 2026, many of them now in clinical trials. Nvidia has invested in the publicly traded company.

What this means for the bigger picture

If this holds up in larger trials, it would be among the first evidence that an AI-designed drug can have a geroprotective effect -- slowing biological aging -- in humans. The field of AI drug discovery has been long on promise and short on clinical proof. A single Phase IIa result is not proof, but it is the kind of data point that shifts the conversation from "can AI find molecules?" to "can AI find molecules that do things human chemists wouldn't have thought to look for?"

The six aging clocks agreeing with each other, despite being built independently, is the most compelling part of the story. Not because it proves the drug works -- it doesn't -- but because it demonstrates that the signal is real enough to be detected by multiple methods. Whatever rentosertib is doing to blood proteins, it is not a statistical artifact of a single model.

The real test is still ahead: a larger trial, in healthy people, with direct measurement of aging rather than AI-predicted proxies. But for the first time, that test looks like it might be worth running.

Sources

• AI-designed drug appears to turn back the body's biological clock in early trial -- The Decoder
• Early Data Indicates an A.I.-Generated Drug Could Slow Aging -- The New York Times
• Integration of proteomic aging clocks in a Phase IIa clinical trial -- Nature Biotechnology
• Proteomic Aging Clocks Track Biological Age Reversal in Rentosertib Trial -- Unite.AI