Abstract
This study investigated whether the Endocrine Disruptome and VirtualToxLab in silico platforms are suitable for predicting the endocrine disrupting effects of per- and polyfluoroalkyl substances (PFASs)-in particular, for interactions with oestrogen receptors (ERs) and androgen receptor (AR). Compounds included in the U.S. Environmental Protection Agency's PFAS working list were analysed with both models, and the results were compared with the available in vitro data regarding their modulation of nuclear receptors. Based on the identified prediction parameters, such as sensitivity, specificity, accuracy, and Mathews' correlation coefficient, VirtualToxLab was found to be a reliable model for predicting the reactivity of PFASs with AR, while a positive consensus approach of both platforms provided reliable predictions of the PFAS reactivity with ERα and ERβ. This study provides the evidence that Endocrine Disruptome and VirtualToxLab can be used as a tier 1 screening tool for assessment of the endocrine disrupting effect of PFASs. Furthermore, it demonstrates that the likelihood of endocrine disrupting properties increases with the lipophilicity of PFASs and identifies the understudied PFHpS, PFNS, PFDS, 9-Cl, NMeFOSAA, NEtFOSAA, 4:2 FTS, 6:2 FTS, 8:2 FTS, 6:2 monoPAP, 8:2 monoPAP, and 5:3 acid as potential ligands of AR and/or ERs.