Rager J, Isaacs K, Biryol D, Newton S, Strynar M, Sobus J. 2017. Innovative screening methods to identify chemical exposure signatures and linkages to toxicity: Case study with house dust. Presented at Society of Toxicology Annual Meeting, March 13, Baltimore, MD.
Non-targeted and suspect screening analysis methods are analytical approaches that can be used to identify large numbers of chemicals in environmental and biological samples and thus unravel the human exposome. Suspect screening involves the detection of analytes using existing chemical inventories and matching algorithms, while non-targeted screening uses no a priori information—that is, no list of suspected or targeted chemicals. Here, we present a case study for implementing such approaches through the chemical analysis of house-dust samples. House dust was selected for this analysis, because it’s an important indicator of indoor chemical exposures, particularly for infants and young children. A preliminary suspect screening analysis tentatively identified thousands of chemicals in dust; high-throughput bioactivity and exposure data (from EPA’s ToxCast and ExpoCast programs, respectively) were used to prioritize chemicals for confirmatory analysis. This proof-of-concept study identified a number of previously unstudied chemicals in dust, and demonstrated the potential of 21st century environmental monitoring methods in the evaluation of complex mixtures. Despite this innovative screening strategy, only a small percentage (<5%) of the total constituents of dust were successfully identified. More recent efforts have therefore built upon this work using a more holistic approach. After screening for the presence of individual chemicals, all sample constituents were characterized using supervised and unsupervised clustering and classification methods in order to detect human exposure signatures. Specific approaches evaluated include: 1) hierarchical clustering of co-occurring constituents, 2) analysis of the correlation of constituents with specific exposure sources (e.g., smoking within the home, cleanliness, pesticide use), and 3) identification of candidate “sentinel” markers of specific exposure sources using random forest classification modeling. These methods collectively aid in characterizing individual sample components (i.e., chemical structures) and complex exposure pathways. This research serves as groundwork for rapid exposure screening methods that can be used to connect aggregate exposure pathways to adverse outcome pathways, with the goal of identifying individual chemicals and chemical mixtures that contribute to human disease.