AI companies run virtual drug trials, aim to improve success of human studies | Reuters
Copenhagen startup BioinvestGPT runs virtual versions of drug trials with its AI models, and it says its forecasts were right in five of six cases so far. The companies behind these tools hope that testing drugs in simulation first will help more of them succeed in real human studies.
- BioinvestGPT shared its predictions for several high-profile drug trials with Reuters in July, before those results were public.
- The firm has been correct in five of six of those forecasts so far, a small sample that still needs more track record to judge.
- The pitch is that better virtual trials could cut the failure rate of expensive human studies.
I can't see the full article text from the link, only the headline and one subhead: "AI companies run virtual drug trials, aim to improve success of human studies," with a line saying Copenhagen-based BioinvestGPT shared detailed predictions of several high-profile trial outcomes with Reuters in July, and has been right in five of six so far. So the summary below rests on that, and I'd be guessing at the rest.
What the piece appears to be about is a shift in how AI firms are being used in drug development: running simulated, or "virtual," versions of clinical trials before people are enrolled in the real ones. The pitch is that better prediction of outcomes could lead to better-designed human studies, which matter because most drugs that enter trials never reach patients, and the cost of failure is enormous. Reuters' framing puts the emphasis on whether these tools can actually improve the odds, not just on the technology itself.
The one concrete data point is BioinvestGPT's track record: five correct calls out of six, on trials whose predictions were recorded in advance and checked against results afterward. That's an encouraging number, but six cases is a very small sample. A 5-of-6 record could reflect real skill, a lucky stretch, or the choice of which trials got included, and a read-it-later reader should treat it as a starting signal rather than proof. The useful follow-up questions are which six trials were called, how the predictions were made, and whether the firm's methods have been independently checked. If the full article answers those, it's worth saving for a closer read.