Proteomic aging clocks detect biological age reversal in IPF drug trial

Six proteomic aging clocks all pointed in the same direction in a small phase 2a trial: patients who received an AI-designed drug for idiopathic pulmonary fibrosis (IPF) showed a shift toward a younger biological age, while the placebo group did not. The analysis, published in Nature Biotechnology on 7 September 2026, is the first head-to-head comparison of multiple proteomic aging clocks in a clinical intervention.

The drug is rentosertib (formerly INS018_055), a TNIK inhibitor developed by Insilico Medicine using generative AI. The phase 2a trial (NCT05938920) randomized 71 patients across China to receive 30 mg once daily, 30 mg twice daily, 60 mg once daily, or placebo for 12 weeks. A subset of 42 patients, with a mean age of 67.1 years, consented to serum proteomic screening on the Olink Explore 3072 platform, which measured 2,841 proteins.

Researchers led by Insilico Medicine founder Alex Zhavoronkov applied six recently published proteomic aging clocks to the data: ProtAge, OrganAgechrono, OrganAgemortality, PAC, ipfP3GPT and PAOPAC. The four chronological-age clocks tracked actual age well (Spearman r of 0.70 to 0.84), while the two mortality-trained clocks correlated more weakly (r of 0.16 to 0.23), as expected from their training objective.

Ad

Proteomic aging clocks all agree

Every clock predicted lower biological age in the treated arms relative to baseline. Of 54 comparisons per arm across six clocks, three timepoints and three regimens, 21 reached statistical significance (Q value below 0.10), concentrated at week 4, where 11 of 18 comparisons registered lower biological age. The 30 mg twice-daily regimen produced the most consistent signal, with nine significant comparisons, followed by 60 mg once daily (seven) and 30 mg once daily (five).

Separating aging from fibrosis

The harder question is whether these changes reflect aging itself or the drug’s anti-fibrotic effect. The authors point to a dissociation. The 60 mg once-daily group, which showed the largest improvement in forced vital capacity (FVC) in the original trial report, had a less consistent biological-age response than the 30 mg twice-daily group. Change in FVC explained little of the change in biological age across all six clocks, with a median R2 of 0.06.

As a further check, the team compared treatment-induced protein changes against age-associated changes in 55,319 UK Biobank participants. Rentosertib-modulated proteins were enriched 1.74-fold for age-associated proteins. The 30 mg twice-daily regimen preferentially reversed age-associated trajectories (Spearman r of -0.30, P < 0.01), while the 60 mg once-daily regimen showed no such correlation (r of -0.097, P = 0.37).

Senescence and metabolism

Gene set enrichment analysis found that placebo patients accumulated senescence markers over the trial, while every treated arm downregulated the SenMayo signature. Seven proteins appeared in the leading edge of every treated arm, including EREG, IGFBP4, MMP10, MMP13 and SPP1. The 30 mg twice-daily regimen was also the only arm to modulate cholesterol metabolism, the pentose phosphate pathway and glutathione metabolism, which the authors describe as aging mechanisms rather than features of the anti-fibrotic response.

The single most influential feature across all six clocks was LTBP2, a regulator of fibrosis in the TGF-β signaling axis and the only important feature present in every clock.

A framework for dual-purpose trials

The study’s broader argument is that proteomic aging clocks can be embedded in disease-focused trials to evaluate geroprotective effects at the same time. The authors propose a stepwise path: collect aging and senescence biomarkers prospectively in disease trials, replicate effects in non-IPF populations, then pursue biomarker qualification or composite clinical endpoints.

The findings carry caveats. The sample is modest, the observation window is short, and the analyses are largely computational. The authors stress that disentangling anti-fibrotic from anti-aging effects will require validation in healthy volunteers, and that a proteomic platform measuring fewer than 3,000 proteins cannot match the near-million sites of modern epigenetic arrays.

Source: Zhavoronkov A, Galkin F, Chen S, et al. Integration of proteomic aging clocks in a phase 2a clinical trial supports simultaneous geroprotective assessment. Nature Biotechnology (2026).