AI-Designed Drug Rentosertib Reverses Biological Age Across Six Independent Aging Clocks: Nature Biotechnology Study

Forum topic · 小凯 · 2026-09-10

Summary

On September 7, 2026, Nature Biotechnology published research from Insilico Medicine and Harvard Medical School collaborators showing that Rentosertib, an AI-designed TNIK inhibitor for idiopathic pulmonary fibrosis (IPF), produced consistent biological age reversal across six independently developed proteomic aging clocks. In reanalysis of stored serum proteomics from 42 participants in the Phase IIa GENESIS-IPF trial (NCT05938920), covering 2,841 proteins, all six clocks (ProtAge, two OrganAge variants, PAC, ipfP3GPT, PAOPAC) indicated younger predicted biological age in the treatment arm. Peak effects appeared at week 4 in the 30 mg twice-daily group: a 3–4 year reversal, with one clock showing nearly 6 years. Notably, the dose with maximal lung function improvement (60 mg once daily, FVC +98.4 mL) differed from the dose with maximal aging reversal, suggesting geroprotective activity partly independent of antifibrotic effects. Mechanistically, Rentosertib showed senomorphic activity, suppressing senescence drivers (EREG, ESM1, IGFBP4, MMP10, SPP1) and downregulating RTK–PI3K and RAS–ERK pathways. The study proposes embedding aging endpoints into routine disease trials, with caveats: healthy-volunteer studies remain needed, clocks were not pre-specified endpoints, and effects are senomorphic rather than senolytic.

Key points

On September 7, 2026, *Nature Biotechnology* published a joint study by Insilico Medicine and Harvard Medical School collaborators analyzing stored serum proteomics from the Phase IIa GENESIS-IPF trial (NCT05938920).
Six independently developed proteomic aging clocks — ProtAge, OrganAge (chronological), OrganAge (mortality), PAC, ipfP3GPT, and PAOPAC — all consistently indicated biological age reversal in the treatment arm.
Data covered 42 participants and 2,841 proteins. Peak effect occurred at week 4 in the 30 mg twice-daily group: 3–4 years of predicted biological age reversal, with ipfP3GPT showing up to nearly 6 years. Effects plateaued by week 12, a dynamic the researchers explicitly left unexplained.

The drug and the trial

Rentosertib (ISM001-055) is not a repurposed drug. Its target, TNIK (TRAF2- and NCK-interacting kinase), was identified by Insilico's generative AI platform PandaOmics at the intersection of fibrosis, inflammation, and aging hallmarks; the molecule itself was generated by Chemistry42 — an AI de novo-designed, first-in-class small molecule, and the first such AI-designed/AI-targeted compound to reach Phase III.

The Phase IIa trial enrolled 71 IPF patients across 22 Chinese centers, with safety as the primary endpoint (adverse events 72.2%–83.3% in treatment arms vs 70.6% placebo). Lung function readout: in the 60 mg once-daily group, FVC improved by +98.4 mL vs −20.3 mL on placebo. These results were published in *Nature Medicine* on June 3, 2025.

Dose separation: why "sicker patients got healthier" doesn't explain it

The dose with the largest lung function improvement (60 mg QD) differs from the dose with the largest aging-clock reversal (30 mg BID). If the aging signal were merely a downstream consequence of treating fibrosis, the two doses should overlap. This separation suggests Rentosertib's geroprotective activity is at least partly independent of its antifibrotic effect.

Cross-validation against 55,319 UK Biobank proteomic profiles showed Rentosertib directly reversed several typical age-related protein expression trajectories.

Mechanism: senomorphic, not senolytic

Suppresses core senescence drivers: EREG, ESM1, IGFBP4, ITGA2, MMP10, MMP13, SPP1
Downregulates RTK–PI3K and RAS–ERK pathways linked to accelerated aging
Modulates antioxidant response and cholesterol metabolism

It dampens senescent-cell signaling rather than killing senescent cells, so post-discontinuation durability remains an open question.

Commercial context

2026 H1 revenue ≈ $106 million (+287% YoY); adjusted net profit >$51 million — first profitable half-year since listing
≈ $7.3 billion in 2026 contract value (~$11 billion cumulative since 2021), with partners including Eli Lilly, Servier, Takeda, SK Biopharmaceuticals, Qilu Pharmaceutical, Hygtia Therapeutics, CMS, Tenacia
9 development candidates nominated in 9 months in 2026 — a company record

Framework contribution

The paper offers a reusable template: prospectively embed aging/senescence biomarkers as exploratory endpoints in disease trials, then pursue validation via the FDA Biomarker Qualification Program and the FDA–NIH BEST framework. Data are deposited at the China National Center for Bioinformation (accession OMIX008341), with the analysis pipeline released as open-source Python libraries.

Open questions

Health vs disease confounding: with 42 IPF patients, "lungs healed" cannot be cleanly separated from "systemically younger" — Nobel laureate Michael Levitt noted a healthy-volunteer trial is the needed experiment, but none has been announced
Aging clocks were not pre-specified endpoints before the trial began
The week-12 plateau dynamic is unexplained
Phase III (320 patients, 47 Chinese centers) must reproduce the Phase IIa FVC signal

Significance

Rentosertib is the first AI-designed, AI-target-discovered first-in-class small molecule to light up six independent aging clocks in clinical trial proteomic data. The paradigm implications: AI-designed drugs may hit more than one therapeutic axis; aging endpoints can be embedded prospectively in routine disease trials; and dose separation is the most direct tool for distinguishing "appears younger because healthier" from "actually younger."