Science Inventory

Predicting the probability that a chemical causes steatosis using Adverse Outcome Pathway Bayesian Networks (AOPBNs)

Citation:

Burgoon, L., M. Angrish, N. Garcia-Reyero, N. Pollesch, A. Zupanic, AND E. Perkins. Predicting the probability that a chemical causes steatosis using Adverse Outcome Pathway Bayesian Networks (AOPBNs). RISK ANALYSIS. Blackwell Publishing, Malden, MA, 40(3):512-523, (2020). https://doi.org/10.1111/risa.13423

Impact/Purpose:

Adverse outcome pathways (AOP) provide evidence to support linkages between molecular level changes due to a chemical or stressor and adverse outcomes. AOPs provide directed causal information that, with additional information linking events, can be used in a probabilistic predictive sense. This paper presents a method using Bayesian networks to develop an adverse outcome pathway Bayesian network (AOPBN) for predicting liver steatosis.

Description:

Adverse Outcome Pathway Bayesian Networks (AOPBNs) are a promising avenue for developing predictive toxicology and risk assessment tools based on Adverse Outcome Pathways (AOPs). Here we describe a process for combining AOPs with existing literature on disease pathways into AOPBNs. AOPBNs use causal networks and Bayesian statistics to integrate evidence across key events. Investigators can use these AOPBNs to design test batteries that replace animal tests in the future. Investigators can also identify data gaps key events that need to be measured to increase their confidence in a prediction. In addition, it allows investigators to make informed hypotheses about potential modes of action. To illustrate these ideas, we use our AOPBN to predict the occurrence of steatosis under different chemical exposures. Since it is an expert-driven model, we use external data (i.e., data not used for modeling) from the literature to validate predictions of the AOPBN model. The AOPBN accurately predicts steatosis for the chemicals from our external data. In addition, we demonstrate how to use Pearl’s backdoor algorithm, commonly used with causal networks, to identify the minimal set of sufficient key events that is, those key events that are the only ones that need to be measured to be predictive. We then demonstrate empirically using our model what sufficiency means probabilistically. We also walk through a process for identifying how to test combinations of key events to identify how they impact confidence in steatosis predictions. We then close with a discussion of how the model can be used to predict potential effects of mixtures and how to model susceptible populations (e.g., where a mutation or stressor may change the conditional probability tables in the AOPBN). Using this approach for developing expert AOPBNs will facilitate the identification of minimally sufficient set of key events (MinSSKE) to predict an adverse outcome and greatly improve chemical hazard screening strategies.

Record Details:

Record Type:DOCUMENT( JOURNAL/ PEER REVIEWED JOURNAL)
Product Published Date:03/01/2020
Record Last Revised:05/14/2020
OMB Category:Other
Record ID: 348806