In Silico QSAR and qRASAR Approaches for Danio rerio Toxicity: A Comparative Study Across 2-, 3-, and 4‑Hour Exposures with Mechanistic and Data Gap Filling Analyses
Lu Li
Co-Presenters: Individual Presentation
College: The Dorothy and George Hennings College of Science, Mathematics and Technology
Major: Biology
Faculty Research Mentor: Supratik Kar
Abstract:
The Toxic Substances Control Act (TSCA) mandates that the US EPA documents all chemicals in the US. With nearly 86,000 chemicals and many added annually, obtaining toxicity data for each substance is impractical. In silico approaches like quantitative structure-activity relationship (QSAR) and quantitative read‐across structure-activity relationship (qRASAR) offer strategic alternatives to predict aquatic toxicity, crucial for protecting aquatic species and human health. In this study, we curated acute LC₅₀ (median lethal concentration) toxicity data for Danio rerio (zebrafish), a well-established model organism for ecotoxicity testing from the US EPA’s ToxValDB. The dataset was stratified by experimental covariates including study type (mortality), study duration (2, 3, and 4 hours), exposure route (static and renewal), exposure method (drinking water), and chemical type (industrial chemicals and pharmaceuticals). The curated datasets comprised 97 compounds for 2-hour studies, 45 compounds for 3-hour studies, and 356 compounds for 4-hour studies. Using these data, we developed 6 sets of robust QSAR and qRASAR models to predict zebrafish aquatic toxicity. Our results indicate that, across all study durations, the qRASAR models outperformed the QSAR models in external predictivity. Moreover, for the 3- and 4-hour models, qRASAR approaches also demonstrated superior internal predictivity, although the 2-hour model showed comparable Q²_LOO values between the two methods. To further assess the utility of our models, we predicted the toxicity endpoints for over 1,100 external chemicals lacking zebrafish toxicity data, thereby effectively filling significant ecotoxicity data gaps. Overall, the integration of QSAR and qRASAR methodologies across multiple exposure durations not only enhances the predictive capability for aquatic toxicity but also provides mechanistic insights into the toxic action of chemicals. These models offer a valuable tool for regulatory prioritization and risk assessment, contributing to more efficient environmental protection strategies.