AI companies see success in predicting drug failures after building virtual human trials

Copenhagen-based BioinvestGPT says it can predict whether drugs will fail in clinical trials by building virtual human trials. The company shared detailed analyses with Reuters in July that set out its expected outcomes for several high-profile drug trials before the results were released.

So far, BioinvestGPT has correctly predicted five out of six outcomes. That is the main evidence offered for the approach in the text available here.

The source is only a fragment of the original article. It contains the headline, the opening lines and the track record, and the rest is cut off and replaced by unrelated headlines from the page, such as stories about a Bentley on sale in Doncaster, a statue in France and Disney ticket prices. It gives no detail on how the virtual trials work, which drugs or trials were predicted, what the one miss was, or whether other AI companies are doing similar work, even though the headline refers to "AI companies" in the plural.

The significance is in what the headline implies. Late-stage drug trials are costly and often fail, so a method that can flag likely failures in advance would be valuable to developers and investors. The six-prediction sample is small, and the available text does not say how the company defines or checks a correct prediction. For a fuller picture, readers would need the original Reuters-sourced article.