The First AI-Designed Drug in Phase III: Rentosertib Case Study

What Is Rentosertib?

Rentosertib (INS018_055) is a small-molecule inhibitor of TNIK (TRAF2 and NCK-interacting kinase) developed by Insilico Medicine for the treatment of idiopathic pulmonary fibrosis (IPF). It holds the distinction of being the first drug candidate where both the target and the compound were identified and designed using artificial intelligence, reaching Phase II clinical trials in 2023.

The development of rentosertib represents a watershed moment in AI-driven drug discovery, demonstrating that an end-to-end AI pipeline—from target identification to lead compound—can produce viable clinical candidates. The drug was generated using Chemistry42, Insilico Medicine's generative chemistry platform, after the target TNIK was identified using PandaOmics.

Data: Rentosertib Development Timeline and Metrics

Source: Insilico Medicine press releases and clinical trial registry (NCT05154240, NCT05975983).

Traditional drug development typically takes 4-5 years to reach IND filing, meaning the AI-driven approach saved approximately 2.5-3 years in the preclinical phase.

How: The AI-Driven Development Pipeline

Step 1: Target Identification with PandaOmics

1. Input: IPF disease phenotype data, including 13 fibrotic disease datasets
2. PandaOmics analyzed multi-omics data (transcriptomics, proteomics, genomics)
3. Applied AI algorithms to rank 20+ potential targets
4. TNIK emerged as top candidate based on:
   • High expression in fibrotic tissue
   • Druggable kinase domain
   • Novel IP position
   • Role in Wnt/β-catenin and TGF-β signaling pathways

Step 2: Molecule Generation with Chemistry42

1. Used structure-based generative chemistry (TNIK crystal structure available)
2. Chemistry42 generated ~80 novel molecular scaffolds
3. Multi-objective optimization: TNIK potency, selectivity, ADMET, synthetic accessibility
4. After 6 rounds of optimization, INS018_055 was selected:
   • IC50 = 2.4 nM (TNIK)
   • Selectivity > 100x over 300 kinases
   • Favorable oral bioavailability (F = 65%)
   • Synthetic route: 5 steps

Step 3: Preclinical Validation

1. In vitro: TNIK inhibition confirmed, anti-fibrotic activity in human lung fibroblasts
2. In vivo: Efficacy in bleomycin-induced lung fibrosis mouse model
3. Toxicology: 28-day repeat dose in rats and dogs
4. Safety pharmacology: hERG, genotoxicity, phototoxicity panels

Step 4: Clinical Development

1. Phase I: Single and multiple ascending doses (SAD/MAD) in healthy volunteers
2. Favorable safety and PK profile at doses up to 120 mg
3. Phase IIa: Proof-of-concept in IPF patients
4. Phase IIa results: Dose-dependent FVC improvement, acceptable tolerability

Comparison: AI-Driven vs. Traditional Drug Development

Summary: Key Takeaways

1. Rentosertib is the first AI-designed drug to reach Phase II/III, validating end-to-end AI drug discovery.
2. The development timeline was reduced by ~2.5-3 years compared to traditional approaches.
3. The AI pipeline (PandaOmics → Chemistry42 → InClinico) demonstrated that both target and compound can be AI-identified.
4. Phase IIa results showed clinical proof-of-concept, supporting further development.
5. This case study establishes a template for future AI-driven drug discovery programs.

References

1. Zhavoronkov, A. et al. "Deep learning enables rapid identification of potent DDR1 kinase inhibitors." Nature Biotechnology 37, 1038-1040 (2019).
2. Insilico Medicine. "Insilico Medicine initiates Phase II clinical trial of INS018_055 for IPF." Press release (2023).
3. ClinicalTrials.gov. NCT05975983. "A Study to Evaluate INS018_055 in Participants with Idiopathic Pulmonary Fibrosis."
4. Pun, F.W. et al. "PandaOmics: An AI-driven platform for therapeutic target identification and biomarker discovery." Drug Discovery Today (2023).
5. FDA. "Artificial Intelligence in Drug Development." Draft guidance (2024).