What the paper reports
Insilico Medicine says six independently built proteomic aging clocks, applied to blood samples from a Phase IIa trial of its drug rentosertib, all indicated a lower predicted biological age in treated patients. The analysis was published in Nature Biotechnology on 7 September, according to the company’s announcement.
The trial (NCT05938920) was randomised, double-blind and placebo-controlled, and it was a trial of idiopathic pulmonary fibrosis, not of ageing. Insilico’s analysis drew on twelve weeks of serum proteomic data covering 2,841 proteins from 42 patients.
The numbers, and who reports them
All of the following come from Insilico’s own announcement of its own paper. In the 30mg twice-daily group, the clocks showed a peak reversal of three to four years at week four, reaching roughly six years on some of the models. On lung function, the 60mg once-daily group recorded a mean gain of 98.4 mL in forced vital capacity against a 20.3 mL decline on placebo, per the release.
Why six clocks and not one
A proteomic aging clock is a model that reads a blood proteome and returns a predicted age. Different groups train them on different cohorts against different outcomes, so a single clock’s verdict is easy to dismiss as an artefact of how that particular model was built.
The six here were developed separately: ProtAge from Harvard Medical School, two variants of OrganAge from Oxford, and PAC, ipfP3GPT and PAOPAC from Peking University and Westlake University. Insilico’s point is agreement — the models disagree about a great deal, and did not disagree here.
The caveat is in the announcement itself
The most useful line in Insilico’s own release comes from Nobel laureate Michael Levitt, who is quoted saying the trial “cannot yet separate slower aging from a treated lung” and that a study in healthy volunteers is what he wants to see next.
That is the whole problem in one sentence. Rentosertib treats a disease of the lung. If it works, the patient’s blood proteome will change because the lung is less inflamed and less fibrotic — and an aging clock reading that blood will report a younger number. Distinguishing “this drug slows ageing” from “this drug treats the illness that was making these people look old” requires participants who are not ill.
Where the AI comes in
Rentosertib inhibits TNIK. Insilico says it identified that target with its PandaOmics platform and generated the molecule with Chemistry42, its generative chemistry system. That is the claim the company has been building toward for a decade: not that AI wrote a paper about a drug, but that an AI-nominated target and an AI-designed molecule reached patients and moved a clinical endpoint.
Forty-two patients is a small number to carry that weight, and a secondary proteomic analysis is a weaker thing than a trial designed to test the hypothesis it now supports.
What to watch
The healthy-volunteer study Levitt asked for. Until something like it runs, the honest reading of this result is that a drug which appears to help a fibrotic lung also makes a blood-based age model read younger — which is interesting, consistent, and not the same claim as reversing ageing.
