On September 7, 2026, Insilico Medicine said a Phase IIa study of its AI‑designed idiopathic pulmonary fibrosis (IPF) drug candidate, rentosertib, showed a consistent drop in predicted biological age across six independent proteomic clocks. The company framed the result—reported alongside a paper it says is published in Nature Biotechnology—as a first for a de novo molecule and target discovered by AI and assessed with aging biomarkers. For readers tracking geroscience, the Rentosertib biological age signal is the part worth pausing on.
Six clocks, one signal in IPF
Insilico’s release describes a 12‑week analysis in 42 IPF patients from a Phase IIa trial using plasma proteins measured on the Olink platform. According to the company’s statement on PR Newswire, six proteomic aging clocks—built by groups from Harvard, Oxford, Peking University, and Insilico—each pointed in the same direction: an apparent reduction in predicted biological age during treatment. The models include names familiar in the field, such as ProtAge, OrganAge, and PAC.
Why does concordance matter? Aging clocks often disagree on individual readouts because they’re trained on different cohorts, endpoints, and laboratory methods. Seeing alignment across multiple independent models reduces the risk that one algorithm’s bias is driving the story. It doesn’t prove disease modification. It does raise the bar on internal consistency.
Why a biological-age readout could change IPF and aging trials
IPF trials live or die on lung function, imaging, and survival over long windows. Embedding proteomic aging clocks adds a faster, orthogonal readout that could help separate signal from noise early in development. Insilico argues the study offers a blueprint for building geroscience endpoints into disease trials, and says the approach fits the U.S. Food and Drug Administration’s Biomarker/BEST playbook. The agency’s biomarker qualification framework doesn’t bless any single assay here, but it does outline how sponsors can define context of use and pursue formal qualification.
For IPF specifically, a biological‑age signal could hint at systemic effects beyond the lung. If an antifibrotic candidate shifts a broad proteomic profile toward a younger state, developers get an early hypothesis: are we slowing processes that track with multiple age‑related outcomes, not just lung scarring? That’s the promise Insilico sets up. The Rentosertib biological age finding won’t replace hard outcomes, yet it could guide dose, duration, and combination strategy in follow‑on studies.
Rentosertib biological age signal: what we still don’t know
There’s important caution built into the same announcement. The dataset covers 42 patients over 12 weeks—short and small for an age‑related disease. Proteomic clocks are surrogate markers; none are validated as registrational endpoints for IPF or for aging. According to Insilico’s press release, the longitudinal proteomics tied to the trial are deposited under accession OMIX008341 at China’s National Center for Bioinformation; the OMIX portal lists such datasets. Independent re‑analysis will matter as much as the headline result.
The company also describes rentosertib as a de novo drug aimed at a de novo target discovered with AI and aging research. That’s the pitch, but the mechanism, target validation depth, and how the proteomic changes relate to lung function or patient‑reported outcomes still need transparent linkage in peer‑reviewed data. If future readouts show that the clocks shift while FVC or imaging stand still, the field learns a different lesson about what these biomarkers capture.
The wider test: can proteomic aging clocks earn trust?
Protein‑based aging clocks promise practical advantages over DNA methylation assays: faster turnaround, direct readout of circulating biology, and potential organ‑specific views. They also face the same hurdles—batch effects, demographic biases, and overfitting to limited training cohorts. A six‑clock agreement is encouraging because it trims model idiosyncrasies. It doesn’t remove confounders like inflammation, comorbidities, or changes in concurrent medications that could nudge many proteins at once.
That’s why pre‑specified endpoints, transparent code, and external validation cohorts matter. If others can reproduce the direction and magnitude of change in an IPF population, then test whether similar shifts track with function or hospitalization, proteomic clocks gain standing as drug‑development tools. If not, they stay interesting but ancillary. Either way, this study pushes sponsors to document how they collect, normalize, and interpret longitudinal proteomics—details that FDA reviewers will expect if a team tries to move a biomarker toward qualified use.
What happens next if the clocks keep agreeing
Expect two near‑term moves if the data hold up. First, more IPF and pulmonary trials will add proteomic aging clocks as secondary or exploratory endpoints. The cost is modest compared with imaging and extended follow‑up, and the readouts could inform go/no‑go calls earlier. Second, companies will start cross‑comparing clock shifts to core disease outcomes across studies. If a pattern emerges—say, a certain drop in predicted age correlates with a specific FVC change over six months—that could seed a case for context‑of‑use statements under FDA’s biomarker programs.
Investors and clinicians should also look for durability. A transient dip in predicted age that snaps back after dosing stops would argue for mechanism‑linked, on‑treatment effects rather than lasting remodeling. If the Rentosertib biological age signal lines up with sustained clinical benefit in larger cohorts, it strengthens the idea that proteomic clocks are tracking something causal, not just correlational noise.
Insilico’s claim, laid out in its press statement on September 7, 2026, doesn’t settle the aging‑biomarker debate. It does create a testable path. If multiple independent models keep pointing the same way, and those shifts tie to patient outcomes, then proteomic clocks move from slide decks into real development decisions. That’s where the Rentosertib biological age story either becomes a tool—or a cautionary tale. For more on this, see nytimes.com.
Related reading: AI Update • Automation • Generative AI