论文部分内容阅读
20 Quantum chemical parameters of chlorophenol compounds were fully optimized by using B3LYP method on both 6-31G* and 6-311G* basis sets. These structural parameters are taken as theoretical descriptors, and the experimental data of 20 compounds’ aquatic photogen toxicity (–lgEC50) are used to perform stepwise regression in order to obtain two predicted –lgEC50 correlation models whose correlation coefficients R2 are respectively 0.9186 and 0.9567. In addition, parameters of chlorine atom’s substitutive positions and their correlations (NPCS) are taken as descriptors to obtain another predicted –lgEC50 model with the correlation coefficient R2 of 0.9444. Correlation degree of each independent variable in the three models is verified by using variance inflation factors (VIF) and t value. In the cross-validation method, cross-validation coefficients q2 of 3 models are respectively 0.8748, 0.9119 and 0.8993, which indicates that the relativity and prediction ability of this model are superior to those of the model obtained by topological and BLYP methods.
20 Structural chemical parameters of chlorophenol compounds were fully optimized by using B3LYP method on both 6-31G * and 6-311G * basis sets. These structural parameters are taken as theoretical descriptors, and the experimental data of 20 compounds’ aquatic photogen toxicity (- lgEC50) are used to perform stepwise regression in order to obtain both predicted -lgEC50 correlation models whose correlation coefficients R2 are respectively 0.9186 and 0.9567. In addition, parameters of chlorine atom’s substitutive positions and their correlations (NPCS) are taken as descriptors to obtain another Predicted -lgEC50 model with the correlation coefficient R2 of 0.9444. Correlation degree of each independent variable in the three models is verified by using variance inflation factors (VIF) and t value. In the cross-validation method, cross-validation coefficients q2 of 3 models are respectively 0.8748, 0.9119 and 0.8993, which indicates that the relativity and prediction ability of this mod el are superior to those of the model obtained by topological and BLYP methods.