Quantification of an Adverse Outcome Pathway Network by Bayesian Regression and Bayesian Network Modeling

Area of Science:

• Environmental Toxicology
• Computational Toxicology
• Risk Assessment

Background:

• The Adverse Outcome Pathway (AOP) framework systematically links mechanistic processes to toxicity endpoints.
• Implementation in risk assessments is hindered by a lack of quantitative AOP models (qAOPs) and uncertainty assessment.
• Existing qAOP models often rely on data-intensive systems biology approaches.

Purpose of the Study:

• To propose a less data-demanding approach for quantifying AOPs and AOP networks using regression modeling and Bayesian networks (BNs).
• To demonstrate the utility of this approach through a case study of AOP #245 involving pesticide exposure in Lemna minor.
• To develop a quantitative AOP-Bayesian Network (AOP-BN) model for predicting adverse outcomes.

Main Methods:

• Employed Bayesian regression modeling to quantify causal relationships within AOPs, including dose-response and key event relationships.
• Utilized fitted regression models with associated uncertainty to simulate response values.
• Parameterized Bayesian Network (BN) models using simulated values to create a quantitative AOP-BN.

Main Results:

• The developed AOP-BN model successfully reflected the network structure of AOP #245, encompassing molecular initiating events (MIEs), key events (KEs), and adverse outcomes (AO).
• The model demonstrated high accuracy in internal validation, particularly when run from intermediate nodes and when lower resolution for the AO was acceptable.
• Despite limitations due to a small dataset, the study proved the concept of combining Bayesian regression and BN modeling for AOP quantification.

Conclusions:

• The proposed approach offers a viable, less data-intensive alternative for quantifying AOPs and AOP networks.
• The AOP-BN model provides a flexible tool for prognostic, diagnostic, and omnidirectional inference within toxicological pathways.
• This proof-of-concept study paves the way for more robust quantitative risk assessments using AOPs.