Insilico Medicine has registered a Phase III clinical trial for rentosertib, a drug for idiopathic pulmonary fibrosis whose molecular structure was originally generated by the company’s AI drug-discovery platform, marking one of the furthest advances yet for a medicine designed with the help of artificial intelligence rather than purely human medicinal chemistry. The trial, registered on July 7, 2026 under the identifier NCT07687459, is expected to enroll roughly 320 patients over 52 weeks, with an estimated start date of August 30, 2026.
Idiopathic pulmonary fibrosis is a progressive lung-scarring disease with no cure, and existing treatments mainly slow decline rather than reverse it. Rentosertib targets the disease through a different mechanism than currently approved drugs: it inhibits an enzyme called TNIK, which the company’s research links to fibrosis-promoting signaling pathways and chronic lung inflammation, an approach Insilico’s AI platform flagged as a promising drug target years before any molecule existed to test it.
Why This Trial Matters for the AI Drug Discovery Field
Thousands of AI-discovered drug candidates have entered early testing over the past several years, but very few have survived the attrition of clinical trials long enough to reach Phase III, the final and most expensive stage before a company can seek regulatory approval. A peer-reviewed analysis presented at the American Society of Clinical Oncology’s 2026 meeting counted 117 AI-enabled therapeutic assets from 63 companies that had entered human trials, of which just over half had completed Phase 1 and fewer than 7% had completed Phase 2. Rentosertib’s advance to Phase III puts it in a small, closely watched group that includes Recursion’s REC-4881 for a rare polyp-forming condition and Generate:Biomedicines’ GB-0895, an antibody for severe asthma now enrolling roughly 1,600 patients in its own Phase III program.
The Phase 2a Data Behind the Decision
The case for moving rentosertib forward rests on results from the GENESIS-IPF trial, a 71-patient study across 22 sites in China. Patients on the high dose of rentosertib saw their forced vital capacity, a standard measure of lung function that reliably declines as pulmonary fibrosis progresses, improve by a mean of 98.4 milliliters over 12 weeks, compared with a 20.3 milliliter decline among patients on placebo. For a disease where the expected trajectory is steady lung function loss, any measured improvement rather than just a slower decline drew significant attention from pulmonologists when the results published in Nature Medicine.
The Safety Signal That Complicates the Picture
The Phase 2a results were not without warning signs. Seven patients on the high-dose regimen discontinued treatment due to liver-related adverse events, and only 67% of patients completed the full high-dose course compared with 88% in the placebo group. That tolerability gap means the Phase III trial will need to closely monitor liver function and treatment discontinuation rates alongside its primary lung-function endpoints, and some pulmonologists caution that the efficacy signal, while promising, was measured in a relatively small, geographically narrow patient population that may not fully represent the disease’s global variability.
What AI Actually Contributed to This Drug’s Origin
Rentosertib’s development history is frequently cited in the industry as one of the clearer examples of AI meaningfully compressing drug discovery timelines rather than simply assisting human researchers at the margins. Insilico’s platform was used both to identify TNIK as a viable target and to generate and prioritize candidate molecule structures computationally, a process the company says took significantly less time than traditional target discovery and lead optimization, though independent researchers note that translating a faster discovery phase into a genuinely safer or more effective drug still depends entirely on how the molecule performs in human trials, which unfold on the same timeline and face the same risks regardless of how it was designed.
A Sector Still Proving Itself
The broader AI drug discovery industry has faced growing scrutiny over whether computationally designed molecules actually succeed in the clinic at higher rates than traditionally discovered ones, given that most AI-flagged candidates still fail for the same reasons any experimental drug fails: unexpected toxicity, insufficient efficacy, or unfavorable side-effect profiles relative to existing treatments. Industry observers describe the sector as moving into a more sober phase, focused less on the discovery process itself and more on rigorously proving out results in later-stage trials, exactly the phase rentosertib is now entering.
What Comes Next
If rentosertib’s Phase III trial confirms both the lung-function benefit and an acceptable safety profile at scale, it would become one of the first AI-originated small molecules to reach regulatory submission for a chronic disease, a milestone the drug discovery industry has anticipated for years. Enrollment is expected to begin in the coming weeks, with the 52-week trial duration meaning meaningful readouts are unlikely before 2028, a reminder that even AI-accelerated drug discovery still runs into the same slow, patient-by-patient pace of human clinical testing that governs every other experimental medicine.
