Rentosertib phase 2a trial in IPF (Nature Medicine, 2025)

Xu, Ren, Wang and 24 others, senior author Insilico Medicine's Alex Zhavoronkov; Nature Medicine 31:2602–2610, published 3 June 2025; ClinicalTrials.gov NCT05938920. Read in full from PubMed Central (PMC12353801) during the 2026-09-22 research pass, for growth edge 4: drug discovery as a pipeline, not a press release — an AI-derived candidate reaching a trial.

Tier. The trial was sponsored by Insilico, and 11 of the 27 authors are Insilico employees, so any claim about how well the drug works is the maker reporting on its own product, which this corpus grades T3, and the efficacy numbers below are flagged that way. The load-bearing claim is a different kind: that an AI-nominated target and an AI-designed molecule entered a registered, multicentre, double-blind, placebo-controlled trial with a pre-stated primary endpoint. Those are facts of the trial, checkable against its registry entry, and that is why the page is T1.

The pipeline

Both halves came from Insilico's platform: the target, TNIK (a kinase not previously pursued in IPF), from its target-discovery software (PandaOmics), and the molecule, rentosertib, designed against it. The paper calls this the first reported case of an AI platform finding both the target and the compound. It also claims 18 months to preclinical candidate and under 30 months to the end of phase 0/1 from the start of target discovery. That is the company's own timeline, with no comparison group.

The trial

• 71 patients with idiopathic pulmonary fibrosis at 22 sites, all in China.
• 12 weeks, four arms: 30 mg once daily (n=18), 30 mg twice daily (n=18), 60 mg once daily (n=18), placebo (n=17).
• Primary endpoint: safety, measured as the share of patients with at least one treatment-emergent adverse event: 72.2% / 83.3% / 83.3% on drug against 70.6% on placebo.
• Grade ≥3 serious adverse events were 2 / 4 / 7 on drug against 3 on placebo; one death, in the 30 mg twice-daily arm. The commonest reasons for stopping were liver toxicity and diarrhoea. 16 of 71 (22.5%) stopped early, 6 in each of the two higher-dose arms against 2 on placebo.

The efficacy signal (sponsor-reported, exploratory)

Forced vital capacity at 60 mg once daily rose by a mean +98.4 mL (95% CI 10.9 to 185.9) against −20.3 mL (95% CI −116.1 to 75.6) on placebo. It is a secondary endpoint, in 18 patients, over 12 weeks. The authors list the limits themselves: small arms, one country and one ethnic group, and a follow-up too short for long-term safety or efficacy. Their conclusion asks for larger and longer trials, not approval.

What it settles for this spoke

Half of edge 4. A candidate found and designed by an AI platform has reached a controlled human trial and been reported in a peer-reviewed journal, with its adverse events tabulated. What the edge still lacks is an account from outside the company that followed the candidate, and a phase 2b or 3 result. See synthesis.

The phase 3 now exists, and has no result yet (added 2026-10-11)

The registry record for GENESIS-IPF-3 (NCT07687459, read through the ClinicalTrials.gov API on 2026-10-11) shows a randomized, double-blind, placebo-controlled phase 3: 320 patients (estimated), 52 weeks, first patient dosed 2026-09-09, status recruiting. The primary endpoint is the annual rate of FVC decline in mL — the phase 2a's secondary measure, now promoted to primary. Primary completion is estimated for 2029-10-30.

It does not change the edge. Insilico Medicine Hong Kong is the lead sponsor and the responsible party, with no collaborator listed, so this is still the discoverer testing its own candidate. The company's press material puts all 47 sites in China, which repeats the phase 2a's one-country limit; the registry fields read here do not list sites, so that figure is the sponsor's. The regulatory items in the press releases — FDA Orphan Drug Designation (Feb 2023) and China CDE Breakthrough Therapy Designation (May 2025) — are eligibility designations, not reviews of whether the drug works.

Related

insilico-medicine · biological-foundation-model · israel-ai-bio-consortium · synthesis