Chen Shu-Da, Zeng Xiao-Lan, Wang Zun-Yao, Liu Hong-Xia
School of Biological and Chemical Engineering, Jiaxing University, Zhejiang Jiaxing, 314001, People's Republic of China.
Sci Total Environ. 2007 Aug 15;382(1):59-69. doi: 10.1016/j.scitotenv.2007.04.014. Epub 2007 May 24.
The number of Cl substitution positions (N(PCS)) of all 209 possible molecular structure patterns of polychlorinated diphenyl ethers (PCDEs) were correlated with their partition properties n-octanol/water partition coefficient (lgK(ow)) and sub-cooled liquid water solubilities (-lgS(w,l)). The correlation coefficients (R) and the leave-one-out (LOO) cross-validation correlation coefficients (R(cv)) of all the 6-descriptor models for lgK(ow) and -lgS(w,l) are more than 0.98. By using stepwise multiple regression (SMR), the best two models of lgK(ow) with three descriptors (R=0.9,913) and lgS(w,l) with four descriptors(R=0.9,818) were obtained respectively. Based on these equations, the two properties of all 209 PCDEs were predicted. Model validation showed that the two models exhibit optimum stability and high predictive power. Moreover, results of prediction were further compared with data from similar studies by molecular connectivity indices (MCIs) and PM3 methods.
对多氯代二苯醚(PCDEs)所有209种可能分子结构模式的氯取代位置数(N(PCS))与其分配特性正辛醇/水分配系数(lgK(ow))和过冷液态水溶解度(-lgS(w,l))进行了相关性研究。lgK(ow)和-lgS(w,l)的所有6描述符模型的相关系数(R)和留一法(LOO)交叉验证相关系数(R(cv))均大于0.98。通过逐步多元回归(SMR),分别得到了lgK(ow)的最佳三描述符模型(R = 0.9913)和-lgS(w,l)的最佳四描述符模型(R = 0.9818)。基于这些方程,预测了所有209种PCDEs的这两种性质。模型验证表明,这两个模型具有最佳稳定性和较高的预测能力。此外,还将预测结果与分子连接性指数(MCIs)和PM3方法的类似研究数据进行了进一步比较。