摘要
In this study,DFT-B3LYP level of theory with the 6-311G basis set was used to calculate a set of molecular descriptors of 36 nitrobenzenes.Quantitative structure-activity relationship（QSAR） models of the toxicity（pIGC50） of 36 nitrobenzenes were established using some of the following calculated descriptors：HOMO energy（EHOMO）,LUMO energy（ELUMO）,the difference between LUMO and HOMO energy values（ΔE=ELUMO–EHOMO）,average polarizability（P）,dipole moment（μ）,molecular volume（V） and the charge of nitro group（Q-NO2）.The QSAR models obtained by employing stepwise multiple regression techniques are aimed at correlating the structures with their reported experimental toxicity values. Among the models presented in this study, statistically the most significant one is a two-parameter linear equation with the correlation coefficient （R2） of 0.896 and cross-validated correlation coefficient （R2cv） of 0.878. To further validate the predictive ability of the resulting model, external validation was carried out with R2~xt and Q^2,ext values of 0.894, 0.900, 0.909, 0.874, 0.885 and 0.903, respectively. This study also reconstructed the obtained models using AMl-based calculated descriptors to demonstrate the superiority of DFT over AM1 for quantum calculations of mechanical descriptors. The results were discussed in the light of the main factors that influence the toxicity of nitrobenznenes.
In this study,DFT-B3LYP level of theory with the 6-311G basis set was used to calculate a set of molecular descriptors of 36 nitrobenzenes.Quantitative structure-activity relationship（QSAR） models of the toxicity（pIGC50） of 36 nitrobenzenes were established using some of the following calculated descriptors：HOMO energy（EHOMO）,LUMO energy（ELUMO）,the difference between LUMO and HOMO energy values（ΔE=ELUMO–EHOMO）,average polarizability（P）,dipole moment（μ）,molecular volume（V） and the charge of nitro group（Q-NO2）.The QSAR models obtained by employing stepwise multiple regression techniques are aimed at correlating the structures with their reported experimental toxicity values. Among the models presented in this study, statistically the most significant one is a two-parameter linear equation with the correlation coefficient （R2） of 0.896 and cross-validated correlation coefficient （R2cv） of 0.878. To further validate the predictive ability of the resulting model, external validation was carried out with R2~xt and Q^2,ext values of 0.894, 0.900, 0.909, 0.874, 0.885 and 0.903, respectively. This study also reconstructed the obtained models using AMl-based calculated descriptors to demonstrate the superiority of DFT over AM1 for quantum calculations of mechanical descriptors. The results were discussed in the light of the main factors that influence the toxicity of nitrobenznenes.
基金
supported by the Natural Science Foundation of Yancheng Teachers University (08YCKL077)
the Jiangsu Provincial Key Laboratory of Coastal Wetland Bioresources & Environmental Protection (JLCBE 07025)
the Educational Bureau of Jiangsu Province (08KJD150021)
