AI-designed drugs reach human trials

AI is moving from research tool to active drug-design engine as companies use machine learning to analyze biological and chemical datasets before lab testing begins. The shift is aimed at replacing slow trial-and-error screening with systems that can simulate and evaluate millions of molecular possibilities, narrowing the field of candidates faster and at lower cost.

Insilico Medicine’s Rentosertib, an experimental treatment for idiopathic pulmonary fibrosis, is being evaluated in Phase IIa clinical trials in the U.S. and China. A 12-week randomized, double-blind study recently showed dose-dependent improvement in lung function, and the company says its Pharma.AI platform identified TNIK as a target and used generative models to optimize the molecule. The discovery process was completed in roughly 30 months, about half the time of traditional methods.

Other companies, including DeepMind, Recursion Pharmaceuticals and BenevolentAI, are pursuing related systems for protein prediction, clinical data analysis and compound discovery. DeepMind’s AlphaFold has helped researchers predict protein structures, while AlphaFold 3 expands modeling to interactions involving DNA, RNA and ligands.

Broader use could support personalized medicine, faster clinical trials and more work on rare diseases, but reliability remains tied to data quality. Biased datasets, opaque black-box models, patient trust, regulatory scrutiny and job disruption concerns mean human scientists and physicians remain central to validating AI-generated results.