Decoding Mechanisms Underlying Chemical Pollution Risks to Biodiversity: A Computational Workflow for Constructing Adverse Outcome Pathways (AOPs)

Chemical pollution is a pervasive driver of biodiversity loss but is under-studied compared to other drivers like climate change. One challenge in characterizing chemical impacts on biodiversity is forming the link between chemicals and subsequent adverse effects on individuals, populations, and ecosystems. Adverse Outcome Pathways (AOPs) describe the link between the trigger of a chemical effect (the molecular initiating event, MIE), key events (KEs) in the cause-effect chain, and ultimately an adverse outcome (AO) in a species. However, the number of existing AOPs is limited (≈550 across chemicals, outcomes, and species), and they are predominantly for humans. Therefore, to form the link between chemicals and adverse outcomes across species to protect biodiversity, more non-human AOPs are needed.

In our project, we develop and test a computational workflow that links existing publicly available data to develop AOPs that may have cross-species relevance, thus serving as potential mechanistic indicators for chemical impacts on biodiversity.  Starting from existing databases of chemical-gene interactions in the Comparative Toxicogenomics Database, we link chemical-specific gene sets and associated pathway/gene ontology terms using enrichment analysis with the DAVID Knowledgebase, we map associations to specific anatomical contexts with the Bgee Gene Expression Database, and link phenotypes via the Monarch Initiative. Anatomical entities are harmonized with the UBERON multispecies anatomy Ontology and Cell Ontology. Through this, we build per-pesticide networks and extract recurring AOP chains (MIE → cellular/tissue KE → adverse outcome) that were predicted across multiple pesticides per species.

For example, across 173 pesticides, we identified 611,551 recurring AOPs that describe potential pesticide-induced neurotoxicity in zebrafish, and the fungicide iprodione was implicated across the most AOPs. By following this AOP-development method for other species (e.g., extrapolating zebrafish AOPs across more widely distributed fish species), these cross-species AOPs provide individual chemical-level mechanistic predictions into chemical impacts that may contribute to adverse outcomes underlying biodiversity loss. These AOPs can complement established biodiversity monitoring initiatives by predicting causal linkages that signal earlier warnings of biodiversity impacts and enable targeted mitigation.