Beyond adverse outcome pathways: making toxicity predictions from event networks, SAR models, data and knowledge

Area of Science:

• Toxicology
• Computational chemistry
• Risk assessment

Background:

• Adverse outcome pathways (AOPs) provide a framework for understanding toxicity mechanisms.
• Key events (KEs) and their relationships are fundamental components of AOPs.
• Predicting adverse outcomes (AOs) is crucial for chemical safety assessment.

Purpose of the Study:

• To construct networks using AOP building blocks (KEs and relationships) for predicting AOs.
• To integrate diverse data sources, including assay data and structure-activity relationship (SAR) predictions, into a unified framework.
• To develop an information-rich display for visualizing knowledge, data, and AO predictions for both general cases and specific chemicals.

Main Methods:

• Utilized key events (KEs) and key event relationships from AOPs to build predictive networks.
• Augmented KE networks with data from toxicological assays and SAR predictions linking chemical classes to KEs.
• Employed a reasoning framework to combine these inputs for generating AO predictions.

Main Results:

• Successfully constructed networks of KEs for predicting adverse outcomes (AOs).
• Demonstrated the integration of assay data and SAR predictions within the network framework.
• Provided illustrative examples for skin sensitization, reprotoxicity, and non-genotoxic carcinogenicity, showcasing the predictive capability for individual chemicals and abstract cases.

Conclusions:

• Networks of key events (KEs) offer a robust method for predicting adverse outcomes (AOs).
• The integration of diverse data sources enhances the predictive power and utility of AOP-based networks.
• This approach provides valuable tools for chemical risk assessment and regulatory decision-making.