Clemens Wittwehr is a Group Leader in the Systems Toxicology Unit of the European Commission’s Joint Research Centre (JRC). His work focuses on fostering international collaboration to advance New Approach Methodologies (NAMs) and on promoting the Adverse Outcome Pathway (AOP) Framework as a bridge between mechanistic science and regulatory decision-making.
At the JRC, Clemens leads initiatives linking OECD Harmonised Templates (OHTs) with AOP development to make mechanistic data FAIR and regulatory-ready. Within the AI4AOP initiative, he explores how artificial intelligence - particularly Large Language Models and Knowledge Graphs - can enhance AOP discovery, reasoning, and usability while keeping AI outputs anchored in scientific evidence.
Earlier in his JRC career, Clemens managed the IUCLID project, which delivered the software now used globally for REACH regulatory submissions. He holds a degree in Computer Science and Business Administration from the University of Linz, Austria.
Smarter Toxicology Ahead: The Role of AI in Shaping Adverse Outcome Pathways (AOPs)
Clemens Wittwehr, European Commission, Joint Research Centre (JRC)
The convergence of artificial intelligence (AI), New Approach Methodologies (NAMs), and mechanistic toxicology is opening new avenues toward smarter, animal-free safety assessment. As the volume of biological and omics data continues to grow, AI provides the analytical capacity to uncover meaningful, decision-relevant patterns from this complexity. Within this changing landscape, the Adverse Outcome Pathway (AOP) framework remains central - linking molecular perturbations to adverse health outcomes and providing a mechanistic and transparent foundation for regulatory science.
The AOP-Wiki, an OECD-endorsed global knowledgebase, captures this mechanistic understanding in a standardized form. To keep pace with evolving data and computational tools, the AI4AOP initiative explores how modern AI approaches, including large language models (LLMs), can assist in AOP development, curation, and communication. LLMs can process and summarize scientific literature, identify mechanistic relationships, and support ontology-based annotation, helping experts focus on interpretation and validation rather than data triage.
Beyond technology, the AI4AOP Reverse Mentoring initiative exemplifies how human-AI collaboration can be shaped through community participation. It connects early-career scientists with experienced AOP developers to jointly explore how AI tools can support mechanistic reasoning, key event relationship building, and knowledge translation. This exchange not only tests practical AI applications but also strengthens mutual understanding between generations of researchers and the emerging digital tools of toxicology.
Importantly, AOP knowledge itself provides a safeguard against unverified or implausible AI outputs. Because AOPs describe causally linked, peer-reviewed biological mechanisms, they offer a robust framework that can guide and validate AI-assisted reasoning. In this way, AOPs become both a knowledge source and a quality filter - a foundation for “trustworthy AI” in regulatory contexts.
The presentation highlights three domains where AI can contribute most effectively:
- AOP Development and Review - AI-assisted identification of mechanistic evidence, literature screening, and ontology alignment.
- AOP Usage - employing AI as a co-pilot to enhance discoverability, multilingual access, and consistency within the AOP-Wiki.
- AOP Community Building - leveraging AI to support collaborative learning and engagement, as demonstrated by the Reverse Mentoring initiative.
Ultimately, AI will not replace scientific judgment but complement it - turning the AOP framework into a more adaptive, continuously learning system that evolves with new evidence and community insight.