Rentosertib takes AI drug discovery into a 52-week lung-disease trial

Rentosertib has reached a stage where the most important evidence will come from patients followed for a year. On 10 September 2026, Insilico Medicine announced the first dosing in GENESIS-IPF-3, a placebo-controlled trial of its experimental oral treatment for idiopathic pulmonary fibrosis. The company plans 320 participants across 47 centres in China.[1]

Idiopathic pulmonary fibrosis, or IPF, progressively scars lung tissue and makes breathing harder. The NHS describes current treatment as a combination of medicines that can slow scarring and support for symptoms and everyday function. That creates a demanding question for a new drug: can it preserve enough lung function, for long enough, to make a useful difference to people living with the disease?[6]

AI worked on the target and the molecule

The discovery programme used two linked systems. PandaOmics ranked possible biological targets using patient tissue data, biological networks and published research. With filters for protein kinases, novelty and suitability for small-molecule drugs, TNIK came top. Protein kinases are enzymes involved in cellular signalling. TNIK was already a known protein; the proposal was to inhibit it as a treatment strategy for IPF.[2]

Chemistry42 then generated candidate molecules using the available structure of TNIK. Early compounds bound the target but had problems with properties such as solubility and metabolic clearance. Further design and laboratory testing produced the candidate later named rentosertib. The reported 18-month development interval runs from target discovery to preclinical candidate nomination. It does not include the subsequent years of patient trials.[2]

This is a more specific contribution than an AI system proposing an interesting chemical shape. The programme connected a disease hypothesis to a molecule, then used experiments to improve properties that a drug would need inside the body. Those are separate problems: identifying a relevant protein does not automatically produce a usable inhibitor.[2]

What the 12-week trial established

The registered phase 2 study made treatment-emergent adverse events its primary outcome. Changes in forced vital capacity, or FVC, were secondary outcomes. FVC is a breathing-test measure of the volume a person can forcefully exhale after a full inhalation. The trial randomised 71 people in China across four groups for 12 weeks.[4]

The highest-dose result was the encouraging signal, but the study was small and primarily designed around safety. Sixteen participants stopped treatment early. Seven withdrawals involved liver injury or dysfunction; four of those seven participants were also taking nintedanib. The authors identified drug interactions as a question requiring further study.[3]

A treatment-emergent event is an event arising or worsening during the specified study period; it is not automatically caused by the medicine. That distinction matters here because a useful safety assessment considers both the overall event count and whether investigators judged particular events related to treatment. The trial register records those definitions alongside the reported outcomes.[4]

The next study measures the course of the disease

The phase 3 register lists an actual start date of 9 September 2026 and recruiting status in its September update. Its primary endpoint is the annual rate of FVC decline over 52 weeks. A key secondary endpoint tracks the first disease-progression event, including a substantial fall in predicted FVC, an acute exacerbation, respiratory hospitalisation or respiratory death. Other measures cover CT-assessed fibrosis and patient-reported quality of life.[5]

These endpoints widen the question from a short-term breathing-test signal to the trajectory of illness. Following people through repeated visits can show whether an early difference persists, disappears or is offset by problems that take longer to emerge. Patient-reported outcomes also ask whether a laboratory-style measurement is accompanied by changes people experience in daily life.[5]

The registry currently estimates completion in October 2029, while Insilico’s announcement explicitly describes rentosertib as investigational and unapproved. The coming evidence will therefore test two connected achievements: whether the proposed TNIK mechanism helps patients, and whether an AI-assisted discovery process has produced a medicine with a sustained clinical benefit.[5][1]

Report an error or suggest a correction