KronosRx: How Cedars-Sinai is Using AI ‘Avatars’ to Replace Animal Testing”

Cedars-Sinai has been awarded a $5.05 million contract from the Advanced Research Projects Agency for Health (ARPA-H) to develop KronosRx, an AI-driven platform designed to predict drug toxicity more accurately in humans. The initiative aims to address a major bottleneck in drug development: the high failure rate of clinical trials caused by adverse drug reactions that are often missed by animal models. By combining large-scale electronic health record (EHR) data with lab-grown human “patient avatars,” Cedars-Sinai hopes to reduce costly late-stage failures and accelerate access to safer therapies.

KronosRx pairs human stem-cell–derived organoids and organ-on-chip systems—miniature, functional models of organs such as the heart and brain—with deep-learning models trained on millions of anonymized, longitudinal EHR records. These patient avatars allow researchers to observe how experimental drugs affect human tissue in real time, while the AI layer learns from real-world clinical data. According to project lead Nicholas Tatonetti, PhD, this approach enables dynamic modeling that accounts for age, comorbidities, and polypharmacy, rather than relying on static animal tests that often fail to translate to human outcomes.

The multidisciplinary effort brings together experts in regenerative medicine and computational biomedicine. Researchers will study neurotoxicity using stem-cell models, assess cardiotoxicity—a leading cause of drug withdrawal—with cardiac organoids, and mine unstructured EHR data to identify molecular and clinical patterns linked to adverse reactions. Cedars-Sinai leaders say the platform targets the so-called “valley of death” in drug development, where promising therapies are abandoned after significant investment due to unexpected toxicity in early human trials.

By improving early toxicity prediction, KronosRx could significantly lower development costs, reduce trial failures, and help more effective drugs reach patients faster.