A species-specific QSAR-based framework for PFAS hazard identification and ecological risk mapping

The rapidly expanding chemical space of per- and polyfluoroalkyl substances (PFAS) poses a major regulatory challenge due to limited toxicological data for many environmentally relevant compounds. To address this gap, we developed species-specific quantitative structure-activity relationship (QSAR) models to predict acute and chronic toxicity of untested PFAS across four representative aquatic taxa: fish, water fleas, midges, and green algae. The models demonstrated strong goodness-of-fit, robustness, and predictive performance, outperforming existing predictive tools. Nineteen PFAS were identified as high-potency candidates, whose sub-micromolar toxicity across multiple taxa was linked to structural characteristics captured by 2D descriptors, thereby providing critical structure-based early warning signals. Based on predicted chronic toxicity across taxa, water-column predicted no-effect concentrations (PNEC) were derived for 188 PFAS and subsequently applied to assess the ecological risks of 25 commercially used and environmentally detectable PFAS in surface freshwater of China. Emerging PFAS, such as perfluorooctadecanoic acid (PFODA) and perfluorohexadecanoic acid (PFHxDA), posed greater chronic risks than legacy PFAS even at low environmental concentrations. The risk hotspots mainly concentrated in highly urbanized and industrial reaches of the lower Yangtze and Yellow River basins. Overall, this study provides a scalable computational framework for the rapid screening of ecological hazards of data-poor PFAS, facilitating evidence-based environmental risk management.