However, a major new collaboration led by the University of Michigan, alongside Los Alamos National Laboratory and other industry partners, is exploring how AI combined withs human-derived organoids could transform how drug safety is assessed before trials even begin. Their approach aims to replace traditional reliance on animal testing with systems that more closely reflect human biology.

This work is enabled by human liver organoids — miniature, 3D liver tissues grown from stem cells derived from patients who have experienced DILI. Unlike animal models, these organoids can metabolise drugs and respond to toxicity in ways that closely mirror real human physiology.

The scale, speed and success of this system are striking. In just one afternoon, it can process around 20,000 toxicity tests, with predictive models achieving approximately 90% accuracy, compared to around 50% using traditional methods that can take months.

This approach is part of a broader industry shift toward making drug safety decisions using computation and human-relevant data rather than animal testing. The long-term ambition is even greater: enabling regulatory approval of early-stage trials based on these models.

Earlier and more accurate toxicity prediction could reduce costly late-stage failures, accelerate development timelines and improve patient safety. It also opens the door to more personalised approaches, particularly for niche cases and underrepresented patient groups.

While challenges remain — including regulatory validation and scaling — the system shows significant promise. Drug development may be moving toward a future where safety is assessed not through approximation, but through direct biological simulation.

“This new paradigm of AI and drug discovery will allow us to solve the unsolvable, treat the untreatable and engage directly with patients by modelling their own tissues in the laboratory to find out which drugs work best for them.”
Johnathan Sexton, Professor of Internal Medicine at Michigan Medicine