AI-designed lung drug reverses aging clock signals, but not aging itself

An experimental lung drug discovered and designed using artificial intelligence lowered several molecular estimates of biological age in a small group of patients, offering an intriguing signal for longevity research but no evidence yet that the medicine makes people younger or extends their lives.

Rentosertib, developed by Insilico Medicine to treat idiopathic pulmonary fibrosis, or IPF, was assessed using six independently developed proteomic aging clocks, or machine learning models that analyze proteins in the blood to estimate biological age.

All six clocks showed lower predicted biological age among patients treated with the drug, according to a study published Monday, 7 September, in Nature Biotechnology. The strongest and most consistent effects appeared after four weeks of treatment.

But the researchers cautioned against interpreting the results as proof that rentosertib slows or reverses human aging.

The clocks measure patterns in proteins associated with aging. They cannot determine on their own whether the changes resulted from an effect on aging itself or simply reflected changes caused by treatment of IPF, a serious age-related disease that scars the lungs.

That distinction is key to the findings. The new study analyzed serum samples from 42 patients who had participated in an earlier Phase IIa trial of rentosertib. Forty-three participants consented to the additional proteomic analysis, but one was excluded because a final measurement was missing.

Researchers applied six proteomic aging clocks to blood samples collected before treatment and at two, four and 12 weeks. Across the treatment groups, all six produced shifts toward a younger predicted biological age, while the placebo group showed little change or slight increases.

Of 54 comparisons made for each treatment regimen across the six clocks and three measurement points, 21 were statistically significant. The signals were concentrated at week four, when 11 of 18 comparisons showed significantly lower biological-age estimates in treated patients.

The 30-milligram twice-daily regimen produced the most consistent aging-clock signal. By contrast, the 60-milligram once-daily dose produced the biggest improvement in lung function in the original trial but a less consistent aging-clock response.

That difference is potentially important because it suggests the clock changes may not simply be tracking improvements in lung function. The researchers also compared treatment-related protein changes with age-associated patterns in more than 55,000 UK Biobank participants and found additional signals consistent with an effect on aging-related biology.

None of that establishes that rentosertib is an anti-aging drug. The authors said the small sample, 12-week duration and reliance on computational measures made it impossible to clearly separate the drug’s anti-fibrotic effects from any effect on aging. Confirming a broader geroprotective effect would require studies in people without IPF and ultimately evidence tied to meaningful health outcomes.

Insilico founder and co-chief executive Alex Zhavoronkov, one of the study’s authors, has similarly cautioned against treating the findings as evidence that the drug reverses aging. In an interview with The Wall Street Journal, Zhavoronkov cautioned that no drug has yet been clinically proven to extend human life and acknowledged the uncertainty surrounding rentosertib’s implications for aging.

Rentosertib was developed primarily as a treatment for IPF, a progressive disease in which lung tissue becomes scarred and breathing becomes increasingly difficult.

Its development has attracted attention because AI was used at unusually early stages of the drug-discovery process. Insilico’s PandaOmics platform helped identify TNIK, or TRAF2 and NCK interacting kinase, as a potential target for fibrosis. Its Chemistry42 generative chemistry platform was then used to generate and optimize molecules targeting TNIK.

That work was described in a 2024 Nature Biotechnology paper that followed the drug from AI-assisted target identification through preclinical development and early human testing.

The subsequent randomized Phase IIa trial enrolled 71 patients with IPF in China and assigned them to three rentosertib dosing groups or placebo for 12 weeks. Its primary objective was to assess safety and tolerability rather than prove efficacy.

Lung function was nevertheless measured as a secondary endpoint. Patients receiving the highest dose, 60 milligrams once daily, recorded an average 98.4-milliliter increase in forced vital capacity after 12 weeks, compared with a 20.3-milliliter decline in the placebo group.

The trial was small and short, and its authors said the lung-function findings required confirmation in larger and longer studies.

Insilico said in July that it was initiating a Phase III trial expected to enroll 320 IPF patients and evaluate once-daily rentosertib over 52 weeks.

That study will provide a much more consequential test of whether the drug is safe and effective against the disease for which it was developed.

The new aging analysis raises a different possibility.

If researchers can eventually show that drugs developed for age-related diseases also modify biological processes involved in aging more broadly, clinical trials for established diseases could become a way to investigate potential longevity treatments without trying to prove an effect on aging from the outset.

The Nature Biotechnology authors proposed precisely that approach, with aging biomarkers incorporated into conventional disease trials as exploratory measures.

For now, however, rentosertib has produced a biomarker signal, not a longevity breakthrough.

Its immediate test remains considerably more conventional and considerably harder to fake with an algorithm: whether an AI-discovered medicine can deliver a meaningful clinical benefit to patients with a serious lung disease.