Ecotoxicological QSAR study of fused/non-fused polycyclic aromatic hydrocarbons (FNFPAHs): Assessment and priority ranking of the acute toxicity to Pimephales promelas by QSAR and consensus modeling methods
Shuo Chen,Guohui Sun,Tengjiao Fan,Feifan Li,Yuancong Xu,Na Zhang,Lijiao Zhao,Rugang Zhong
DOI: https://doi.org/10.1016/j.scitotenv.2023.162736
2023-06-01
Abstract:Fused/non-fused polycyclic aromatic hydrocarbons (FNFPAHs) have a variety of toxic effects on ecosystems and human body, but the acquisition of their toxicity data is greatly limited by the limited resources available. Here, we followed the EU REACH regulation and used Pimephales promelas as a model organism to investigate the quantitative structure-activity relationship (QSAR) between the FNFPAHs and their toxicity for the aquatic environment for the first time. We developed a single QSAR model (SM1) containing five simple and interpretable 2D molecular descriptors, which met the validation of OECD QSAR-related principles, and analyzed their mechanistic relationships with toxicity in detail. The model had good degree of fitting and robustness, and had better external prediction performance (MAE<sub>test</sub> = 0.4219) than ECOSAR model (MAE<sub>test</sub> = 0.5614). To further enhance its prediction accuracy, the three qualified single models (SMs) were used for constructing consensus models (CMs), the best one CM2 (MAE<sub>test</sub> = 0.3954) had a significantly higher prediction accuracy for test compounds than SM1, and also outperformed the T.E.S.T. consensus model (MAE<sub>test</sub> = 0.4233). Subsequently, the toxicity of 252 true external FNFPAHs from Pesticide Properties Database (PPDB) was predicted by SM1, the prediction results showed that 94.84 % compounds were reliably predicted within the model's application domain (AD). We also applied the best CM2 to predict the untested 252 FNFPAHs. Furthermore, we provided a mechanistic analysis and explanation for pesticides ranked as top 10 most toxic FNFPAHs. In summary, all developed QSAR and consensus models can be used as efficient tools for predicting the acute toxicity of unknown FNFPAHs to Pimephales promelas, thus being important for the risk assessment and regulation of FNFPAHs contamination in aquatic environment.
environmental sciences