Insilico Medicine (Rentosertib / ISM001-055): AI-Discovered TNIK Inhibitor From Target to Phase IIa IPF Readout
📌Key Takeaways
- 1Insilico Medicine (Clinical Pipeline) (Biotech & Pharmaceutical R&D, 300+ scientists & AI researchers) deployed Insilico Medicine.
- 2Target-to-Preclinical Candidate Timeline: 65% faster cycle time (now 18 months).
- 3Wet-Lab Compounds Synthesized: 95% fewer syntheses (now <80 compounds synthesized).
- 4Implementation timeline: 18 months from target discovery to preclinical candidate; Phase IIa readout completed.
Target Identification to Preclinical Candidate in 18 Months
Idiopathic pulmonary fibrosis (IPF) is a progressive, fatal lung disease with median survival of 3 to 5 years where standard-of-care antifibrotics (pirfenidone and nintedanib) slow decline but don't reverse lung function loss. As detailed in Nature Biotechnology (Ren et al., 2024), Insilico Medicine trained PandaOmics multi-omics and biomedical text transformers on lung fibrosis patient datasets to rank novel intracellular targets. The platform prioritized TRAF2- and NCK-interacting kinase (TNIK), a regulator of TGF-beta, Wnt, and Hippo fibrotic signaling that had never been clinically tested in IPF. What's the takeaway for your own architecture? We've analyzed this rollout closely, and it's clear you can't separate compute from wet-lab execution. You'll get the best ROI when your scientists and modelers share one data loop.
Generative Chemistry & Positive Phase IIa Human Efficacy
Using Chemistry42's structure-based generative reinforcement learning modules, medicinal chemists generated novel small-molecule scaffolds optimized simultaneously for nanomolar TNIK inhibition, selectivity against homologous kinases, and lung tissue distribution. Insilico synthesized and wet-lab tested fewer than 80 compounds to nominate the preclinical candidate ISM001-055 (now USAN-named rentosertib) in 18 months from target launch. In the randomized, double-blind, placebo-controlled Phase IIa trial (NCT05938920) across 71 IPF patients treated for 12 weeks, patients receiving 60 mg once daily achieved a mean forced important capacity (FVC) improvement of +98.4 mL compared to a -62.3 mL decline in the placebo cohort. What's the takeaway for your own architecture? We've analyzed this rollout closely, and it's clear you can't separate compute from wet-lab execution. You'll get the best ROI when your scientists and modelers share one data loop.
Target-to-Preclinical Candidate Timeline
65% faster cycle time
Wet-Lab Compounds Synthesized
95% fewer syntheses
Phase IIa 12-Week FVC Lung Function Delta
+160.7 mL vs placebo
The Challenge
Traditional target discovery and lead optimization for idiopathic pulmonary fibrosis (IPF) require 4.5 years, 1,000 to 3,000 synthesized molecules, and tens of millions of dollars before reaching IND-enabling toxicology. That's a massive bottleneck for any R&D team. If you're running similar programs, you know legacy workflows simply don't scale.
The Solution
Combined PandaOmics multi-omics target discovery (identifying TNIK) with Chemistry42 generative reinforcement learning to design an orally bioavailable, selective small-molecule inhibitor (ISM001-055 / rentosertib). Here's how they fixed it. By pairing Insilico Medicine with automated wet-lab feedback, we see how you can compress months of manual screening into weeks.
Implementation
Timeline
18 months from target discovery to preclinical candidate; Phase IIa readout completed
Team
Insilico Medicine AI target biology, generative medicinal chemistry, and clinical development teams
- 1Trained PandaOmics omics and literature graph models on IPF transcriptomic and methylation cohorts to nominate TNIK as a master fibrotic node.
- 2Seeded Chemistry42 generative models on the TNIK ATP-binding hinge pocket to generate novel small-molecule chemotypes with kinome selectivity.
- 3Synthesized and assayed fewer than 80 compounds across biochemical IC50, cellular EMT/FMT fibrotic assays, and rodent bleomycin models.
- 4Completed Phase I healthy volunteer studies and a 71-patient, 12-week randomized Phase IIa IPF clinical trial (NCT05938920).
Results
| Metric | Before | After | Change |
|---|---|---|---|
| Target-to-Preclinical Candidate Timeline | 42 to 54 months industry average | 18 months | 65% faster cycle time |
| Wet-Lab Compounds Synthesized | 1,500 to 2,500 compounds typical | <80 compounds synthesized | 95% fewer syntheses |
| Phase IIa 12-Week FVC Lung Function Delta | -62.3 mL placebo decline | +98.4 mL FVC improvement (60mg QD) | +160.7 mL vs placebo |
"To our knowledge, ISM001-055 is the first therapeutic candidate with both a novel AI-discovered target and an AI-generated molecular design to report positive Phase IIa clinical efficacy in patients."
Dr. Alex Zhavoronkov, Founder & CEO, Insilico Medicine
Key Learnings
- 1Coupling target discovery AI (PandaOmics) directly with structure-based generative chemistry (Chemistry42) compresses handoff friction between biology and medicinal chemistry.
- 2Multi-parameter reward functions in generative chemistry must penalize synthetic complexity and CYP450/hERG liabilities early to keep wet-lab synthesis under 100 compounds.
- 3Publishing full preclinical and clinical trial data in peer-reviewed journals (Nature Biotechnology) is essential for establishing credibility with pharma licensing partners.
Frequently Asked Questions
Verified References & Citations (2)
- [1]
A small-molecule TNIK inhibitor targets fibrosis in preclinical and clinical models discovered via PandaOmics and Chemistry42.
- [2]
Phase IIa randomized clinical trial readout of rentosertib (ISM001-055) in idiopathic pulmonary fibrosis.
ClinicalTrials.gov NCT05938920: Study of ISM001-055 in Patients With Idiopathic Pulmonary Fibrosis (U.S. National Library of Medicine)