Molecular Adverse Outcome Pathways: towards the implementation of transcriptomics data in risk assessments

Abstract

Adverse Outcome Pathways (AOPs) are designed to provide mechanistic insights into toxicological processes after exposure to a stressor and facilitate the replacement of animal studies with in vitro testing systems. Starting with a Molecular Initiating Event (MIE) and ending with an Adverse Outcome, the sequence of Key Events (KEs) spans across biological levels, but the majority of KEs in AOP-Wiki describe molecular and cellular processes. That suggests that established transcriptome-wide studies can be used to validate or measure a multitude of KEs simultaneously and be a goldmine of useful information about cellular responses. Currently, in toxicology, omics technologies are not widely applied in risk assessments of chemicals because of their complexity and lack of consensus on aspects such as standardization, analysis and interpretation. Given their value in hypothesis generation and ability to provide a vast amount of information about a biological response, the challenge lies in the acceptance of using transcriptomics data in regulatory risk assessments. We here introduce molecular AOPs to define the connections between KEs of the AOP-Wiki and curated biological pathways in WikiPathways, thereby providing a new method for the analysis and interpretation of transcriptomics data to identify KE activation. To study this, we performed case studies on liver steatosis and mitochondrial complex I inhibition, for which molecular AOPs were developed and public transcriptomics datasets were selected. Upon extension of the molecular AOP networks in Cytoscape, we mapped and analysed transcriptomics data, and calculated an enrichment score for individual KEs. Further interpretation of the data was done through the visualisation of the data on the specific molecular pathways. Two molecular AOPs were developed and KEs were linked to the appropriate molecular pathways, allowing a detailed exploration of molecular processes with the selected transcrip-tomics datasets. This has shown us that we can verify the activation of specific MIEs and KEs, and assess progression across the AOP in the steatosis case study through variables in exposure time and dose. These case studies have shown that transcriptomics data can be used for identifying the potential activation of KEs. However, it is also clear that extensive datasets are required to fully test the capabilities of molecular AOPs, and the process of linking molecular pathways and KEs can be challenging, not always allowing one-to-one mapping. While proven valuable to analyse and understand transcriptomic data, pathways linked to KEs appear to show inconsistent levels of activation and should be looked into and refined. More case studies are required to optimize the approaches used for the development and use of molecular AOPs with transcriptomics datasets.