Six proteomic aging clocks applied to blood samples from a clinical trial of the AI-designed drug rentosertib consistently predicted lower biological age in treated participants, Nature Biotechnology reported. Researchers evaluated serum proteomes from a 12-week, randomized phase 2a trial of the candidate anti-fibrotic compound in patients with idiopathic pulmonary fibrosis.
Rentosertib, formerly known as INS018_055, is an artificial intelligence-designed inhibitor of TRAF2- and NCK-interacting kinase. The underlying phase 2a trial screened 128 patients across 21 locations in China and enrolled 71 participants, assigning them to receive 30 milligrams once daily, 30 milligrams twice daily, 60 milligrams once daily or a placebo. Researchers screened serum samples collected at baseline, week 2, week 4 and week 12 using the Olink Explore 3072 platform, yielding a complete longitudinal dataset for 42 participants with a mean age of 67.1 years.
Regimen comparisons and clock performance
Investigators tested six distinct models: ProtAge, PAC, ipfP3GPT, PAOPAC, and two variants of OrganAge. Four clocks trained on chronological age tracked calendar years with Spearman correlation values between 0.70 and 0.84, while two mortality-trained clocks showed lower correlations between 0.16 and 0.23. Despite those structural differences, all six models recorded reductions in biological age relative to baseline across the active treatment groups, whereas the placebo arm showed minimal change or slight increases.
Statistical testing across 54 arm-by-timepoint comparisons showed that 21 reached significance, clustering heavily at week 4. The 30-milligram twice-daily regimen generated nine significant reductions, outperforming the once-daily 60-milligram and 30-milligram doses, which yielded seven and five significant reductions, respectively. Baseline body mass index did not modulate the biological age shifts across the evaluated cohorts.
Protein shifts and disease dissociation
Linear mixed-effects modeling of 2,841 measured proteins revealed that rentosertib altered the trajectories of 326 proteins across treatment arms, compared to two altered proteins in the placebo group. Downregulated proteins included extracellular matrix drivers such as COL1A1, MMP10 and FAP, alongside the growth factor PDGFB. Upregulated proteins included metabolic and stress-response markers such as NAMPT, SOD2 and ALDH1A1.
Feature analysis of biological age shifts identified 35 proteins that each contributed more than 3 percent of the observed age change. The fibrosis regulator LTBP2 emerged as the single most influential feature across all six clocks, accompanied by markers tied to inflammaging, cellular senescence and neurodegeneration, including AGER, CXCL9 and NEFL.
Changes in lung forced vital capacity explained minimal variance in biological age shift across the six clocks, with a median R-squared value of 0.06. That dissociation, paired with the finding that the highest daily single dose produced the largest respiratory gains while the twice-daily split dose produced broader clock consensus, indicated that clock movements did not merely mirror reduced respiratory impairment.
