Breaking

Insilico Medicine Says AI-Designed Fibrosis Drug Reduced Biological Age on Six Clocks

September 7, 2026 at 07:36 EDT

Healthcare
Research & Papers
Foundation Models

An experimental idiopathic pulmonary fibrosis drug designed with generative artificial intelligence reduced predicted biological age on six independent proteomic aging clocks, according to a Nature Biotechnology analysis of a completed mid-stage trial published Monday. Insilico Medicine, the Hong Kong-listed developer, said the finding points to a broader use for rentosertib beyond the scarring lung disease the pill was built to treat.

September 7, 2026 · Insilico Medicine

AI-designed lung drug turns back aging clocks

Rentosertib, a generative-AI TNIK inhibitor for idiopathic pulmonary fibrosis, reduced predicted biological age on six independent proteomic clocks in a 12-week Phase 2a analysis.

Independent proteomic clocks, all lower after treatment

3–4 yr

Peak predicted-age reversal at week 4 (30 mg BID)

IPF patients in the proteomic clock analysis

+98.4 mL

FVC gain at 60 mg once daily vs −20.3 mL on placebo

Predicted biological-age change

Column height is years of clock reversal. Placebo barely moved; treated patients read younger, peaking at about 4 years and up to 6 on one clock.

~0 yr

Placebo
little change

~4 yr

30 mg BID, week 4
peak across clocks

6 yr

Single-clock peak
upper reported reversal

From AI target to clock readout

PandaOmics flags TNIK

AI ranks a dual fibrosis and aging kinase

→

Chemistry42 designs the pill

Generative chemistry yields rentosertib in ~18 months

→

12-week Phase 2a in IPF

71 patients dosed; FVC rose at 60 mg daily

→

Six clocks, one direction

Serum proteomes from 42 patients read younger

What holds up

Clock consensus, not just one model. Six independently trained proteomic clocks all moved younger on treatment, and lung volume improved at the highest daily dose.

What does not

A treated lung is not proof of slower aging. Only IPF patients were studied, the cohort is small, clocks are imperfect, the signal plateaued after week 4, and healthy volunteers have not been tested.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.