Rentosertib is the first AI discovered and AI designed drug to enter Phase III, after a 71 patient, 12 week Chinese trial separated it from placebo on lung capacity.
Insilico Medicine's rentosertib entered Phase III this month. Insilico and the Nature Medicine publication call it the first drug whose target and molecule were both picked by artificial intelligence to reach a large, registration-style efficacy trial. The milestone is real. It is also a beginning, not a verdict: the data that put rentosertib on this path came from 71 patients in China, observed for 12 weeks, against placebo alone.
The Phase IIa study, published in Nature Medicine and posted on Insilico's results page, enrolled patients with idiopathic pulmonary fibrosis (IPF), a progressive lung-scarring disease with a median survival of two to five years after diagnosis. Patients on the 60 mg once-daily dose of rentosertib gained 98.4 mL of forced vital capacity (FVC), a standard measure of how much air the lungs can move, over 12 weeks. Those on placebo lost 20.3 mL. The 95% confidence interval on the treatment arm ran from 10.9 to 185.9 mL; on placebo, from −116.1 to 75.6. The drug separated from placebo at the top dose and not at lower ones, a clean dose-response signal.
Per the Phase III initiation release, the safety profile was comparable across arms: treatment-emergent adverse event rates of 72.2%, 83.3%, and 83.3% for the 30 mg daily, 30 mg twice-daily, and 60 mg daily arms, against 70.6% on placebo. The compound targets a kinase Insilico and its collaborators describe as previously undrugged in fibrosis, per the Nature Medicine paper. There were no new safety red flags.
The target, TNIK, was selected by Insilico's PandaOmics platform, trained on multi-omics data and an aging-biology framework that links fibrosis to inflammation and cellular senescence. Chemistry42, the company's generative chemistry engine, then produced 79 physical molecules; the lead emerged from the 55th iteration and was described in the medicinal chemistry literature. The FDA granted rentosertib Orphan Drug Designation in February 2023, a regulatory status that follows precedent rather than efficacy.
The market rentosertib would enter has only two approved antifibrotics: nintedanib (Ofev) and pirfenidone (Esbriet). Both slow lung function decline; neither reverses scarring, and both carry tolerability burdens that limit real-world use. A third option at a different biological node, where TNIK sits at the intersection of fibrotic signaling, ECM remodeling, and senescence, would matter clinically if the late-stage data hold up.
Phase III is a 320-patient, 52-week, placebo-controlled trial running at Chinese sites, led by Prof. Zuojun Xu of Peking Union Medical College Hospital, per the company release. Insilico has not disclosed a US or EU regulatory pathway or a comparator arm against nintedanib or pirfenidone, the only approved antifibrotics for IPF. A peer-reviewed analysis cited in the Nature Medicine paper noted that AI-discovered drugs to date have failed in Phase 2 at rates comparable to non-AI candidates, and none had previously completed Phase 3. The current Phase III is the first large-scale efficacy test of an AI-discovered, AI-designed candidate.
The Phase IIa readout is a placebo-controlled signal at 12 weeks in a narrow population, not a treatment. Whether an oral AI-designed drug can hold its FVC gain for a year, and how it stacks up against the standard of care, is roughly four years of enrollment and follow-up away.