A workflow to extend Adverse Outcome Pathway knowledge and find supporting experimental data

An Adverse Outcome Pathway (AOP) is a knowledge framework that captures mechanistic descriptions of toxicological responses of a biological system upon exposure to a stressor. Its purpose is to reuse existing knowledge and assist in the replacement of animal testing in risk assessments by sets of in vitro or in silico methods that measure or predict the so-called Key Events (KEs). AOPs are generally stored in the AOP-Wiki and contain mostly qualitative descriptions, while there is a demand for quantitative understanding and supporting data. Our research aims to tackle that demand by combining databases, tools and services into reproducible, adaptable workflows, finding supporting data for AOPs of interest. Several services that have been developed in OpenRiskNet can provide information, data, and predictions related to toxicological assessments. For example, several SPARQL endpoint services exist, exposing RDF for the AOP-Wiki, AOP-DB and WikiPathways. Furthermore, the EdelweissData Explorer allows querying and extracting data from ToxCast and TG-GATES. Together with the identifier mapping of BridgeDb, a reproducible workflow has been developed as a Jupyter notebook. Starting with an AOP ID, the workflow gathers information, various types of relevant identifiers, and experimental data, of which TG-GATES data is used for pathway analysis using WikiPathways. The developed workflow integrates a variety of services into a single Jupyter notebook that allows for rapid exploration of the AOP of interest, and finding, analyzing and evaluating supporting data. Ongoing work focuses on the identification of KEs in AOPs based on experimental data. However, it remains a challenge to automatically link molecular pathways to the mechanisms described in AOPs. Poster presented at BioSB 2020 (2020).