AOP-Wiki EMOD 3.0 uses agentic AI to unify toxicology data models

A new paper from Virginia K. Hench and colleagues introduces AOP-Wiki EMOD 3.0, the third iteration of an evidence model prototype designed to modernize the global repository for Adverse Outcome Pathways (AOPs). AOPs are logic models that link molecular measurements (in vitro or in silico) to adverse outcomes used for chemical regulatory decisions. The existing AOP-Wiki data model and infrastructure have struggled to keep pace with growing data and the need to integrate New Approach Methodologies (NAMs), which replace traditional animal testing.

EMOD 3.0 tackles these limitations head-on by incorporating agentic AI to automate evidence structuring, internal quality improvement, and semantic integration between AOPs and NAMs. The framework focuses on making data more FAIR—Findable, Accessible, Interoperable, and Reusable—while also preparing it for AI-driven analysis. This foundation allows for computationally-generated AOPs and quantitative AOPs (qAOPs), which model dose-response relationships. Ultimately, the work aims to revolutionize regulatory toxicology, biomedical research, and One Health applications by delivering faster, more reliable chemical safety assessments with fewer animals.