AI companies run virtual drug trials, aim to improve success of human studies
BIIB•AI companies are simulating drug trials to help assess experimental medicines before costly human studies, which cost the industry about $140 billion annually and result in regulatory approval for around 12% of candidates. BioinvestGPT says it correctly predicted five of six high-profile trial outcomes and forecasts failures for upcoming Biogen and Takeda studies.
1. Simulations assess drug prospects
AI companies say they are working with drugmakers on virtual trials to identify clinical risks and help decide which programs merit investment. Some simulations take a month or less, compared with years for traditional trials, which proceed through Phase 1, 2 and 3 studies.
2. Predictions include misses
BioinvestGPT simulated Novartis’ del-desiran trial in July and predicted an insignificant clinical benefit; the late-stage study later missed its goal. The company says it has been correct in five of six predictions, but it also forecast positive results for Novartis’ pelacarsen, which failed to reduce major heart attack or stroke risk in a late-stage trial. Co-founder Bragi Lovetrue said the model missed genetically set variations in lipoprotein size.
3. Upcoming trial forecasts
BioinvestGPT predicts failure for two Phase 3 trials of Biogen’s litifilimab in lupus and Phase 2 studies of Takeda’s zasocitinib in Crohn’s disease and ulcerative colitis. Takeda research chief Andy Plump said he does not think AI can yet make definitive predictions; Biogen’s Diana Gallagher said historical-data algorithms could be prone to negative forecasts. Experts said human trials will remain necessary.
