Georgetown Law
Technology Review

The pharmaceutical industry confronts a persistent bottleneck: despite clearing preclinical testing, more than 90% of drug candidates fail in human trials. Unanticipated toxicity remains a major contributor to attrition across therapeutic areas. This failure rate reflects fundamental deficiencies in conventional preclinical toxicity testing methodologies, which are often unable to reliably predict human responses to new drugs.

More sophisticated toxicity predictions are possible through the integration of complementary methodologies, enabling insights that remain out of reach for any approach in isolation. A combinatorial approach integrating stem cell-derived organoids with innovations in machine learning can accelerate drug development, reduce reliance on animal testing, and circumvent obstacles that have traditionally plagued preclinical toxicity prediction.

This Technology Explainer examines the constraints of established preclinical safety testing tools, highlights the distinctive advantages of stem cell-derived organoids, and describes how machine learning converts organoid data into actionable toxicity risk predictions. It concludes with an assessment of the regulatory infrastructure and data governance required for widespread adoption.