Idiopathic pulmonary fibrosis scars the lungs until breathing itself becomes the fight, and no approved drug has ever reversed that scarring rather than merely slowing it. On July 7, 2026, Insilico Medicine said it had enrolled and dosed the first patients in GENESIS-IPF-3, a Phase III trial of rentosertib, a molecule the company says was identified and designed end-to-end by its own AI platform rather than discovered through conventional medicinal chemistry. If the claim holds up to scrutiny, it would be among the furthest any fully AI-originated drug has advanced through human trials.
A disease that has outrun its existing treatments
Idiopathic pulmonary fibrosis, or IPF, causes progressive, irreversible scarring of lung tissue with a cause doctors still can’t pin down, and median survival after diagnosis is commonly measured in just a few years. The two antifibrotic drugs currently approved for IPF, pirfenidone and nintedanib, are understood to slow the rate of lung-function decline, but neither is considered capable of reversing fibrosis that has already formed — which is the specific unmet need rentosertib is designed to address.
What “AI-designed” means for this particular molecule
Rentosertib is a small-molecule, oral, once-daily drug that targets TNIK, a protein kinase implicated in the inflammation-and-scarring feedback loop that drives IPF. Insilico describes it as a first-in-class TNIK inhibitor and markets it as evidence that its Pharma.AI platform can take a target from computational identification through molecule generation to clinical testing without the years of manual chemistry that conventional drug discovery typically requires. Independent observers note that how much of that pipeline was truly “end-to-end AI” versus AI-assisted is still largely a claim made by Insilico itself, since the company controls how that term gets defined in its own announcements.
The Phase IIa data that earned the leap to Phase III
The case for advancing rentosertib rests on a randomized Phase IIa study published in Nature Medicine and presented at the American Thoracic Society’s 2025 International Conference. In that trial, patients on the 60 mg once-daily dose saw their forced vital capacity — a standard measure of how much air the lungs can hold — improve by a mean of 98.4 milliliters over 12 weeks, while the placebo group’s lung function declined by a mean of 20.3 milliliters over the same period. A nearly 120-milliliter gap between an improving treatment arm and a declining placebo arm, in a disease where the baseline trajectory is one-directional decline, is the kind of signal that drug developers rarely see and is what investors and pulmonologists point to as the real justification for the size and cost of a Phase III commitment.
How the Phase III trial is built
GENESIS-IPF-3, registered as NCT07687459 and CTR20262475, is a randomized, double-blind, placebo-controlled study designed to enroll 320 IPF patients across 47 clinical centers in China over 52 weeks. The first patient was dosed at Peking Union Medical College Hospital, with Shanghai Pulmonary Hospital completing its own first enrollment the same day; Professor Zuojun Xu of Peking Union Medical College Hospital is serving as the trial’s leading principal investigator. The 52-week duration matters clinically — IPF’s decline plays out over months to years, so regulators and pulmonologists alike want to see whether rentosertib’s early lung-function gains persist, or merely delay the same eventual decline other therapies already slow.
Why skeptics are waiting for Phase III, not celebrating Phase IIa
The broader track record of AI-discovered drugs argues for caution rather than certainty. An analysis presented at ASCO in 2026 tracked 117 AI-enabled therapeutic assets across 63 companies that had entered human trials, and found that while 51.3 percent had completed Phase 1, only 6.8 percent had made it through Phase 2 — a steep drop-off. A separate Boston Consulting Group analysis of roughly two dozen AI-discovered molecules found Phase 1 safety success rates of 80 to 90 percent, well above the industry’s historical rate near 50 percent, but Phase 2 efficacy success falling back to the ordinary industry rate of roughly 40 percent, since AI has mainly accelerated how fast molecules get designed, not resolved the underlying biological uncertainty of picking the right target. That pattern is part of why Recursion Pharmaceuticals’ 2025 decision to discontinue its own lead AI-discovered program is frequently cited alongside rentosertib’s progress as the other side of the same coin.
What’s next
Insilico expects GENESIS-IPF-3 to run through its full 52-week course before readout, meaning a verdict on whether rentosertib’s early lung-function gains hold up is still well over a year away. Pulmonologists following the trial say the real test isn’t whether an AI can generate a plausible drug candidate — Insilico and its peers have shown that repeatedly — but whether that candidate can clear the same Phase 2-to-Phase 3 graveyard that has quietly ended most AI-discovery stories before they reached patients at scale. For a disease with no reversal therapy on the market, even a partial answer from this trial will shape how much capital the rest of the industry is willing to put behind AI-originated molecules going forward.
Photo: Tima Miroshnichenko / PEXELS via Pexels
