OpenTox 2.0: A Perspective on the Principles for Predictive Toxicology and Risk Assessment Applications in the Era of Integrated New Approach Methods, Computational Modelling and Artificial Intelligence

This perspective revisits and extends the OpenTox framework to address current and emerging challenges in predictive toxicology and chemical risk assessment. Since the original OpenTox principles were formulated in 2010, the field has evolved substantially with the rapid expansion of New Approach Methodologies (NAMs), advances in computational modelling, and the emergence of artificial intelligence (AI)-driven applications. At the same time, persistent challenges remain, including access to high-quality data, interoperability across heterogeneous resources, transparency and reproducibility of models, and the integration of evidence into decision-ready frameworks.

We propose an updated and structured set of OpenTox principles that reflect these developments and support the design and deployment of interoperable, FAIR, and AI-enabled knowledge infrastructures. These principles are aligned with and enable modern regulatory and scientific frameworks, including Integrated Approaches to Testing and Assessment (IATA), Next Generation Risk Assessment (NGRA), and Safe and Sustainable by Design (SSbD). We extend the original framework by incorporating requirements for data and model integrity, provenance, uncertainty communication, and deployment-ready workflows, and further introduce AI-specific principles addressing transparency, correctness, robustness, explainability, and safeguards for high-stakes applications.

A central argument of this work is that the next phase of predictive toxicology requires a transition from isolated models and data silos towards integrated systems that support traceable, reproducible, and decision-oriented evidence generation. We illustrate how these principles can be operationalised through contemporary applications and knowledge infrastructures, including ontology-driven integration, knowledge graphs, and AI-assisted workflows.

 The resulting OpenTox 2.0 framework provides a foundation for building trustworthy, scalable, and regulatory-relevant predictive toxicology systems. By combining open science principles with advances in AI and NAMs, it supports the development of future-ready solutions for human and environmental risk assessment and sustainable innovation.