ExpoPath: A method for identifying and annotating exposure pathways from chemical co-occurrence networks

Improving risk evaluation for environmental and human health is of paramount concern for the U.S. Environmental Protection Agency (EPA). This includes the identification and assessment of chemical transport from commercial and industrial sources to environmental and ecological media, where repeated patterns are often categorized as exposure pathways. Utilizing network analysis techniques paired with graph machine learning tools allows for the construction and analysis of a global chemical co-occurrence network with which to identify sets of overlapping or distinct communities that represent likely exposure pathways. Data from several chemical source databases were aggregated and used to generate a chemical co-occurrence network that encoded linkages between source categories and environmental and receptor categories within the EPA's Multimedia Monitoring Database (MMDB). Multiple algorithms were used to detect communities of chemicals within this network, while enrichment of the resulting communities based on presence-in-media information, physicochemical properties, and functional use information helped to annotate likely exposure pathways. This research identified communities of chemicals associated with various pharmaceutical, consumer, pesticide, and persistent chemical pathways. This novel approach to the study of chemical co-occurrence demonstrates the applicability of network analyses and graph machine learning methods for identifying empirical patterns of connectivity within the domain of exposure science. SYNOPSIS: Network analysis and community detection algorithms help reveal linkages among environmental monitoring data and chemical sources while providing supporting evidence for empirically derived exposure pathways.