RESUMO
The molecular geometries of 209 polybrominated diphenyl ethers (PBDEs) were optimized at the B3LYP/6-31G level with Gaussian 98 program. The calculated structural parameters were taken as theoretical descriptors to establish two novel QSPR models for predicting supercooled liquid vapor pressures (P(L)) and octanol/air partition coefficients (K(OA)) of PBDEs based on the theoretical linear solvation energy relationship (TLSER) model, respectively. The two models achieved in this work both contain three variables: most negative atomic partial charge in molecule (q(-)), dipole moment of the molecules (mu) and mean molecular polarizability (alpha), of which R(2) values are both as high as 0.997, their root-mean-square errors in modeling (RSMEE) are 0.069 and 0.062 respectively. In addition, the F-value of two models are both evidently larger than critical values F(0.05) and the variation inflation factors (VIF) of variables herein are all less than 5.0, suggesting obvious statistic significance of the P(L) and K(OA) predicting models. The results of Leave-One-Out (LOO) cross-validation for training set and validation with external test set both show that the two models obtained exhibited optimum stability and good predictive power. We suggest that the QSPRs derived here can be used to predict accurately P(L) and K(OA) for non-tested PBDE congeners from Mono-BDEs to Hepta-BDEs and from Mono-BDEs to Hexa-BDEs, respectively.