An experimental lung drug designed with artificial intelligence made patients’ blood look biologically younger on six separate aging tests, according to a study published on September 7, 2026 in the journal Nature Biotechnology1. The drug is rentosertib, made by Insilico Medicine to treat idiopathic pulmonary fibrosis (IPF), a disease that scars the lungs and makes breathing steadily harder2. In the group on a placebo (a dummy pill), biological age readings barely moved over 12 weeks or rose slightly3.
The study also shows something new in how drugs get tested. Six independently built “aging clocks” were run on blood from a single drug trial, and an industry publication called the approach a template other trial sponsors could copy4. The researchers stress that the results do not show the drug slows human aging3.
What happened
This was not a new trial. The researchers went back to blood samples from an earlier 12-week study of rentosertib1. That study ran at 21 hospitals in China and enrolled 71 people with IPF aged 40 and older3. It was randomized, double-blind and placebo-controlled, meaning patients were assigned by chance and neither they nor their doctors knew who got the real drug5. Patients took 30 milligrams once a day, 30 milligrams twice a day, 60 milligrams once a day, or a dummy pill3. Insilico published the main lung results in Nature Medicine in June 20256.
For the new analysis, the team used blood from the 42 patients who agreed to have their blood proteins studied for aging research. Their average age was 67.13. Blood was taken at the start and at weeks 2, 4 and 127. The team measured 2,841 proteins in each sample7.
They then ran those protein readings through six “proteomic aging clocks.” These are computer models that estimate how old a body looks from the levels and patterns of many blood proteins, separate from a person’s real age3. Different teams built the clocks using UK Biobank data, a large British health database, and one of the six was built by Insilico itself. Four were trained to estimate age and two to predict the risk of dying5.
The evidence so far
All six clocks gave lower age estimates in patients taking rentosertib, though not at every dose or every visit5. Out of 54 comparisons between drug and placebo groups, 21 were statistically significant (unlikely to be chance) after a standard correction for running many tests8. A further check put the number expected by chance alone at 0.151.
The clearest changes showed up at week four. In the 60-milligram once-daily group, the four age-based clocks gave readings 2.71 to 3.46 years younger2. Reports differ on what that is measured against: Lifespan.io describes it as roughly three years younger than at the start of treatment5, while Medical Daily describes it as younger than placebo1. The two clocks built to predict risk of death did not show a significant change at that dose at week four5. Some clocks and doses gave other figures, and the researchers did not name one overall number for how much rentosertib lowered biological age2. Insilico’s own announcement put the peak at about 3 to 4 years in the 30-milligram twice-daily group, and up to 6 years on one clock6.
The split dose, not the strongest dose, did best on aging
Patients taking 30 milligrams twice a day showed the most consistent drops across both types of clock5. The 60-milligram once-daily dose did best on the lung disease itself7. In that group, Insilico reported, lung capacity rose by an average of 98.4 milliliters, while it fell by 20.3 milliliters on placebo6. Medical Xpress noted that the twice-daily dose also kept drug levels in the blood steadier through the day7.
Beyond the clocks, the drug changed the levels of 326 proteins, compared with only two in the placebo group2. When the team compared these changes with aging patterns in 55,319 UK Biobank participants, proteins affected by the drug were 1.74 times more likely than expected to be linked to aging7. On the twice-daily dose, proteins that are usually higher in older people tended to fall, and the reverse. This pattern was not statistically significant on the once-daily dose5.
What it does not show yet
Every patient had a serious lung disease. If the drug eased their fibrosis, that alone could have made their blood look younger, and the study could not fully separate the two possibilities2. One fibrosis-linked protein, LTBP2, was a major driver of the younger readings and was used by all six clocks. That raises the question of whether the clocks agreed mainly because the drug affected scarring5.
The team tried to test this. Changes in the clock readings did not closely follow changes in lung function, which suggests better breathing alone may not explain them. Lung function does not capture every part of IPF, though5. Proteins tied to cellular senescence (worn-out cells that stop dividing) tended to rise on placebo and fall on the drug, but the researchers did not directly measure such cells in tissue5.
Other limits:
- Only 42 people, for 12 weeks2.
- The drop in age readings stopped growing after week four and levelled off. The researchers do not yet know why2.
- The drug has not been tested in healthy people to look for effects on aging3.
- It is not approved by any regulator for IPF and needs more trials3.
Vadim Gladyshev of Harvard Medical School called it the first study to show that predicted biological age can go down, but he pointed out that the sample was small and that aging clocks are not always reliable3. Eric Topol of the Scripps Research Institute said the drug looks encouraging but that the decisive trials needed for a final judgment have not been done3.
Who is behind it
Insilico Medicine is developing rentosertib, and several of the study’s authors work there2. The lead author, Alex Zhavoronkov, is the company’s founder and co-CEO2. Insilico ran the study with scientists from Harvard Medical School, Stanford University, the Broad Institute, RWTH Aachen University, Peking University and Westlake University6. Gladyshev, quoted above, is a coauthor5.
Insilico says its AI tools picked the drug’s target, a protein called TNIK, because it was linked to both aging and fibrosis, and then designed the molecule to block it6. The company’s announcement calls the work a “blueprint” for trials that measure aging and disease together6.
Zhavoronkov told Bloomberg that a lower biological age means a person is likely to die later, and also said the results are early and long-term effects in larger groups are unknown9. He told Fox News:
“It basically means that you are predicted to be younger using a validated set of aging clocks. We don’t know what that means yet.”10
How it fits the field
Many drugs made for one disease also act on processes linked to aging, so a single trial could check both, according to Lifespan.io5. Drugs studied for longevity so far, such as rapamycin and metformin, were first developed for something else8. Biohack recently covered another case of aging measures being used on existing drugs: drugmakers say Ozempic-style drugs slowed lab measures of aging.
The Clinical Trial Vanguard argued that the key point is agreement among six clocks built in different ways. That avoids staking a trial on one unproven measure, though it does not get any clock officially accepted by the FDA as a stand-in for health outcomes4. The publication also warned that running six clocks is costly and needs careful planning from the start of a trial4. Gladyshev told Lifespan.io:
“This study illustrates how biomarkers of aging can be incorporated into conventional clinical trials,”5
What to watch next
The findings need to be tested in larger and longer studies, including in people who do not have IPF2. Nobel laureate Michael Levitt said what convinced him was that the clocks agreed, not the size of the effect. He also said the trial cannot separate slower aging from a treated lung, and that a test in healthy volunteers is the experiment he wants to see next8. Biohack will follow whether rentosertib, or other drugs, are tested with aging clocks built into the trial from the start.
