\n\n
Virtual‑patient simulations are gaining traction as a way to de‑risk expensive human studies. In July, BioinvestGPT used a simulated trial to predict that the experimental drug del‑desiran would deliver only a modest benefit, a forecast later confirmed by a missed late‑stage outcome. The startup, along with firms such as QuantHealth, is partnering with pharma companies to run virtual trials that highlight which programs merit costly progression. QuantHealth’s chief medical officer, Francisco Beca, stresses the need to avoid “trials that are going to fail,” reflecting a broader industry push to cut waste.
Investment in AI‑driven drug discovery more than doubled to $8.4 billion in 2025, underscoring rapid adoption. AI models now claim to correctly predict five out of six real‑world trial results, including upcoming failures for Biogen’s litifilimab lupus trials and Takeda’s zasocitinib studies in inflammatory bowel disease. Such predictive power could help the sector, which spends about $140 billion annually on human testing but approves only roughly 12% of candidates.
Regulators are also moving. U.S. health officials recently announced initiatives to speed up clinical trials, opening a potential pathway for AI forecasts to be used in regulatory decision‑making. Concurrently, major tech players—Meta, Google DeepMind, and Isomorphic Labs—are pouring $300 million into the Virtual Biology Initiative to generate massive cell‑response datasets and build models that may further shrink development timelines.
Why it matters: By flagging likely failures early, AI virtual trials promise to save billions, focus lab resources on the most promising molecules, and accelerate patient access to effective therapies. Watching next: (1) how regulators integrate AI predictions into trial approvals, (2) the real‑world performance of AI‑forecasted outcomes across more diverse indications, and (3) the impact of the Virtual Biology Initiative’s data platform on next‑generation drug design. [[1, 2, 3]]