Foundations for AI-assisted Adverse Outcome Pathways (AOPs) in radiation research

Area of Science:

• Radiation Protection
• Toxicology
• Computational Biology

Background:

• Adverse Outcome Pathways (AOPs) are crucial for radiation risk assessment but are hindered by manual data integration and complexity.
• Fragmented data across diverse sources complicates the development of robust AOPs.

Purpose of the Study:

• To explore the application of Artificial Intelligence (AI) and Machine Learning (ML) in overcoming challenges in radiation AOP development.
• To propose a phased, AI-driven plan for accelerating AOP construction and data consolidation in radiation protection.

Main Methods:

• Utilizing AI/ML for extraction, annotation, and integration of heterogeneous data sources.
• Applying natural language processing to mine scientific literature for mechanistic insights.
• Employing supervised and unsupervised ML for identifying Key Events (KEs) and Key Event Relationships (KERs).
• Constructing causal models using knowledge graphs and probabilistic inference for mechanistic directionality.

Main Results:

• AI/ML can identify KEs, infer KERs, and suggest AOP structures by mining literature and datasets.
• The proposed AI-driven plan facilitates automated narrative generation, evidence scoring, and model refinement.
• AI/ML integration promises to accelerate data consolidation and systematic organization in radiation AOP development.

Conclusions:

• AI/ML offers a methodological foundation for developing radiation AOPs more efficiently.
• This approach facilitates systematic organization, integration, and prioritization of biological and experimental data for radiation protection.