Artificial intelligence is emerging as a potential tool to help pharmaceutical companies predict whether experimental drugs are likely to succeed before costly human trials. Novartis shares fell 11% after its late-stage trial of experimental muscular dystrophy drug del-desiran missed its goal, wiping out around $30 billion in market value. Copenhagen-based AI startup BioinvestGPT had previously simulated the trial using virtual patients and predicted only an insignificant clinical benefit. The company says its model has correctly predicted five of six major trial outcomes it assessed, highlighting the potential of AI to identify clinical risks earlier.
The global biopharmaceutical industry spends about $140 billion annually on human clinical trials, yet only around 12% of drug candidates ultimately receive regulatory approval. AI companies such as BioinvestGPT and QuantHealth are developing virtual trial platforms that simulate patient responses using genetic and real-world data. These tools can help drugmakers decide which candidates are worth advancing, evaluate acquisition targets and potentially reduce the number of failed trials. However, the technology remains imperfect. BioinvestGPT also incorrectly predicted success for Novartis’ pelacarsen, later acknowledging that it had failed to account for genetically determined variations affecting the drug’s mechanism.
Pharmaceutical companies remain cautious, stressing that AI cannot yet replace human clinical trials. Biogen and Takeda are among companies whose upcoming trials have been predicted by BioinvestGPT to produce weak or negative results, while executives argue that biological mechanisms and patient populations are too complex for definitive AI forecasts. Experts agree that human trials will remain essential, but say simulations could increasingly help determine which studies should proceed. With AI investment in drug discovery reaching $8.4 billion in 2025, the technology could eventually shift the industry’s focus from simply conducting trials faster to avoiding trials that are most likely to fail.
Pic Courtesy: google/ images are subject to copyright
