Quantitative Structure–Activity Relationships Study on the Ah Receptor Binding Affinities of Polybrominated Diphenyl Ethers Using a Support Vector Machine

Abstract

AbstractPolybrominated Diphenyl Ethers (PBDEs), which are widely distributed in the environment due to their use as flame retardant, may cause long‐term health problems in humans. Their structural similarity to Polychlorinated Biphenyls (PCBs) implies their possible dioxin‐like toxicity. By the use of Partial Least Square Regression (PLS ) with Leave‐One‐Out (LOO) Cross‐Validation (CV), seven net atomic charge descriptors have been extracted from more than 80 quantum descriptors for predicting the Aryl hydrocarbon Receptor Relative Binding Affinities (RBA) of PBDEs. Using Support Vector Machines (SVM) and Radial Basis Function Neural Networks (RBFNs), the RBAs of 15 PBDE congeners have been correlated with the extracted seven quantum chemical descriptors. The SVM models generalize better than the RBFN models. The CV correlation coefficients (q2) for the SVM models based on three‐ and five‐fold CVs are 0.889 and 0.896, respectively. The good performance of the QSAR models based on net atomic charges suggests that electrostatic interactions may play important roles in the Aryl hydrocarbon Receptor (AhR) binding of PBDEs.