AI in Drug Discovery: 2025 in Review — Insilico Medicine Hits Phase IIa Milestone
Insilico Medicine achieved the first positive Phase IIa results for a fully AI-designed drug, while the Recursion-Exscientia merger created an end-to-end AI drug discovery platform. Over 200 AI-discovered drugs are now in development.
2025 marked a turning point for AI-driven drug discovery. Insilico Medicine reported positive Phase IIa results for ISM001-055 (rentosertib), a Traf2- and Nck-interacting kinase inhibitor for idiopathic pulmonary fibrosis. This represents the first time a fully AI-designed drug has demonstrated clinical efficacy in a randomized trial — a milestone the field has been working toward for years.
The industry landscape shifted dramatically with the Recursion-Exscientia merger, which combined phenomic screening capabilities with automated precision chemistry into a full end-to-end AI drug discovery platform. Meanwhile, Schrödinger's physics-enabled approach reached late-stage testing with zasocitinib (TAK-279) advancing into Phase III for autoimmune conditions.
The numbers tell the story of a maturing field: over 200 AI-discovered drug candidates are now in various stages of development, with 15-20 expected to enter pivotal trials in 2026. AI platforms have demonstrated the ability to compress early discovery timelines by 30-40% and reduce preclinical candidate development to 13-18 months.
Yet the ultimate test remains ahead. No AI-discovered drug has achieved FDA approval as of early 2026, and Phase III — where historically over 40% of candidates fail — will provide the first large-scale evidence of whether AI meaningfully improves clinical success rates beyond the pharmaceutical industry's persistent ~90% overall failure rate.