Researchers from Insilico Medicine and an international group of academic collaborators have applied six independently developed proteomic aging clocks to blood samples from a Phase IIa trial of the AI-designed drug rentosertib, finding that all six models consistently predicted lower biological age in treated patients. The study, published September 7, 2026 in Nature Biotechnology, is the first head-to-head clinical comparison of multiple proteomic aging clocks in a drug intervention, according to its authors.
The analysis drew on serum proteome data from a randomized, double-blind, placebo-controlled Phase IIa trial of rentosertib in idiopathic pulmonary fibrosis (IPF), conducted across sites in China in 2023 and 2024. Rentosertib is a small-molecule inhibitor of TNIK, a target Insilico identified with its PandaOmics AI platform as a gene involved in six hallmarks of aging, and the molecule itself was generated through the company’s Chemistry42 generative chemistry platform. Of the trial’s participants, 42 consented to longitudinal proteomic profiling at baseline, week 2, week 4, and week 12, covering 2,841 proteins measured with the Olink Explore 3072 panel. The cohort had a mean age of 67.1 years.
Six Clocks, One Consistent Direction
The six clocks — ProtAge, two OrganAge variants trained on chronological age and mortality risk, PAC, ipfP3GPT, and PAOPAC — were developed by different groups using different methods, from classical machine learning to deep learning, and trained to predict either chronological age or mortality risk. Despite these differences, every clock recorded a reduction in predicted biological age in the treatment arms relative to placebo over the 12-week period.
Comparing each treatment arm to placebo across the six clocks and three post-baseline timepoints produced 54 statistical comparisons, of which 21 reached significance, concentrated at week 4. The 30 mg twice-daily regimen produced the broadest cross-clock agreement, registering significant reductions across both chronological and mortality-trained clocks, with biological age reductions in the range of roughly 2.7 to 3.5 years at week 4 across the four chronological clocks in the 60 mg once-daily arm, and mortality-based organ clocks showing larger shifts in select arms. The effect plateaued by week 12, a pattern the authors said warrants further investigation.
Separating Aging Effects From Lung Improvement
A central challenge, the authors wrote, is that proteomic clocks alone cannot distinguish a genuine aging-modulatory effect from the downstream consequence of treating a fibrotic lung disease. The team addressed this indirectly through several complementary analyses.
First, the regimen producing the greatest improvement in forced vital capacity, the standard lung-function measure, was 60 mg once daily — yet that regimen showed a less consistent aging-clock response than 30 mg twice daily, and a regression analysis found lung-function change explained minimal variance in biological age change across all six clocks. Second, the researchers compared treatment-induced protein changes against age-associated protein trajectories in 55,319 UK Biobank participants: the 30 mg twice-daily regimen significantly reversed age-associated proteomic trajectories, while placebo patients’ proteomes drifted in the direction of normal aging. Third, gene-set enrichment analysis showed treated patients downregulated established senescence-associated protein signatures while the placebo group upregulated them: a pattern the authors describe as consistent with a senomorphic effect, in which a drug suppresses the harmful secretions of senescent cells without killing them. Seven proteins, including EREG, IGFBP4, MMP10, MMP13, and SPP1, were downregulated across every treated arm.
The authors acknowledged that full disentanglement of aging and disease effects is not achievable within an IPF cohort and would require validating the drug or its mechanism in healthy volunteers. They also noted limitations including the modest sample size, the 12-week observation window, and reliance on computational approaches without complementary omics modalities.
A Framework for Dual-Purpose Trials
Beyond the rentosertib findings, the paper lays out a stepwise framework for embedding geroscience endpoints into conventional disease trials: prospectively collecting aging and senescence biomarkers as exploratory endpoints, replicating effects in non-IPF age-related populations, and ultimately pursuing biomarker qualification under the FDA’s Biomarker Qualification Program and the FDA–NIH BEST framework. The authors contrast this path with the decades-long route by which approved drugs such as rapamycin and metformin were later evaluated as candidate geroprotectors.
All proteomic data from the study have been deposited with the China National Center for Bioinformation under accession OMIX008341, and the analysis pipeline has been released as an open-source Python library on GitHub. Rentosertib, which Insilico describes as the first drug candidate with both an AI-discovered target and an AI-designed molecule, has advanced to a Phase III trial in IPF. First author Alex Zhavoronkov, Insilico’s founder and co-CEO, is scheduled to present the results at the Nature conference “Redefining Healthcare in the Age of AI” at Sorbonne University in Paris on September 8, 2026, according to the company.