Insilico eyes fibrosis drug in 18 months

Drug discovery is usually a long, ugly slog. You spend years figuring out what biological target even matters, then more years trying to design a molecule that hits it without causing new problems. Insilico Medicine’s pitch is that generative AI can compress that front end hard enough to change the economics of the whole process. The reason people are paying attention now is simple — this is no longer just a claim about lab speed. The company’s fibrosis program has moved into human testing and produced mid-stage data. ### What drug are we talking about? The drug is rentosertib, previously called ISM001-055 or INS018_055. It is being developed for idiopathic pulmonary fibrosis, or IPF — a progressive lung-scarring disease that gets worse over time and has limited treatment options. Insilico says both the biological target and the molecule itself were generated through its Pharma.AI workflow. (insilico.com) ### What does “18 months” actually mean? It does not mean a finished medicine reached patients in 18 months. The shorter clock refers to the preclinical discovery stretch — from starting the program to identifying a novel fibrosis target and nominating a drug candidate. Insilico has said that step took just under 18 months, and that the path from program start to the beginning of Phase I took under 30 months. (insilico.com) ### Why is that a big deal? Because this is the slowest, messiest part of drug R&D for a lot of programs. You are searching for a useful lock, then trying to forge a key that fits. Insilico’s claim is that AI helped with both — target discovery on the biology side and molecule generation on the chemistry side. The company has also argued that its candidates from 2021 to 2024 typically reached nomination in 12 to 18 months, versus several years in more traditional workflows. (prnewswire.com) ### So did the drug actually work? Early signs are encouraging, but this is not a finished verdict. In results published in *Nature Medicine* in June 2025, a Phase IIa trial in 71 patients suggested rentosertib was safe and showed efficacy signals in IPF. That matters because it appears to be one of the first peer-reviewed mid-stage clinical readouts for a drug whose target and molecule were both identified with generative AI. (nature.com) ### What target did AI find? The target is TNIK, a kinase Insilico linked to fibrosis biology. That is important because IPF drug development has been constrained not just by chemistry, but by target selection — picking the wrong mechanism can waste years. Rentosertib is designed as a TNIK inhibitor, so the AI story here is not only “we made a molecule faster,” but “we chose a target faster too.” (insilico.com) is the program now? Insilico has been preparing for later-stage development and also expanding the franchise. In April 2026, the company said an inhaled formulation of rentosertib received IND clearance from China’s CDE for a Phase I study. Insilico also described that as the 13th program from its AI-driven pipeline to receive IND clearance. (insilico.com)same thing as approval. Drugs still have to survive larger trials, manufacturing scale-up, regulatory review, and the basic reality that many promising Phase II programs fail later. So the real test is no longer whether AI can generate candidates quickly — it is whether those candidates keep holding up as the evidence gets harder. ### Bottom line? Insil(insilico.com)ete. The 18-month figure is real, but it applies to candidate creation, not the whole journey to market. The bigger story is that rentosertib has now carried that speed claim into human data — and that is where the industry starts taking it seriously.