Insilico Medicine

Idiopathic pulmonary fibrosis (IPF) is a devastating age-related lung disease with no curative treatment — approved therapies only slow progression, and patients typically die within 2–5 years of diagnosis. Traditional drug discovery to address this unmet need takes over 10 years and exceeds $2 billion. Identifying a disease-associated molecular target for a poorly understood fibrotic disease required integrating vast multi-omics repositories beyond human scale.

Insilico Medicine deployed its end-to-end generative AI platform Pharma.AI, using its Biology AI module (PandaOmics) to identify TNIK as a priority molecular target for IPF in 2019 by integrating omics datasets, publication texts, patents, and grants. The Chemistry AI module (Chemistry42) then aided medicinal chemists in designing, optimizing, and synthesizing ISM001-055, a novel first-in-class small molecule TNIK inhibitor, achieving preclinical candidate nomination just 18 months after initial target identification.

ISM001-055 met its primary Phase 2a endpoint of safety and tolerability across all dose levels in a 71-patient, 21-site randomized controlled trial in China. Patients receiving the highest dose (60 mg once-daily) showed a mean FVC improvement of +98.4 mL over 12 weeks, compared to a –62.3 mL decline in the placebo group — demonstrating not only slowed disease progression but actual improvement in lung function. Results were published in Nature Biotechnology and represent the first AI-designed drug for an AI-discovered target to demonstrate clinical efficacy.

• Generative AI can compress the target identification-to-preclinical candidate timeline to 18 months, dramatically below the industry norm of several years for this phase alone.
• An end-to-end AI platform (target discovery → molecule design → clinical prediction) enabled a fully integrated drug discovery workflow without reliance on external vendors.
• Positive dose-dependent FVC signal at 12 weeks in a heterogeneous disease like IPF validates AI-driven target selection, not just molecule design.