论文部分内容阅读
基于定量结构-性质相关(QSPR)研究多氯联苯化合物(PCBs)的性质具有重要意义。用分子负电性距离矢量(MEDV)表征209个PCBs的分子结构,同时用多元线性回归(MLR)技术和逐步回归结合留一法交叉检验筛选模型变量,建立多氯联苯类化合物的水溶性(LgS_w)、土壤吸附性(LgK_(oc))、色谱保留指数(RRI)、水溶液活度系数(LgY_w)、总分子表面积(TSA)与MEDV的QSPR模型,其线性相关系数(R)分别为0.9651,0.9692,0.9968,0.9111,0.9960。继用留一法和外部样本检验模型稳定性能,其相关系数(R_(CV),Q_(ext))分别为0.9611、0.9812;0.9572、0.9845;0.9941、0.9984;0.9980、0.9412;0.9953、0.9998。结果表明:所建QSPR模型均稳定性和预测能力良好。
It is of great importance to study the properties of polychlorinated biphenyls (PCBs) based on quantitative structure-property correlation (QSPR). The molecular structure of 209 PCBs was characterized by MEDV, and the model variables were also screened by multivariate linear regression (MLR) and stepwise regression with a one-residue crossover test to establish the water-soluble LgS_w, LgK_oc, RRI, LgY_w, TSA and MEDV QSPR model, the linear correlation coefficients (R) were 0.9651 , 0.9692, 0.9968, 0.9111, 0.9960. Following the method of using the left one method and the external sample to test the stability of the model, the correlation coefficients (R CV and Q ext) are 0.9611, 0.9812, 0.9572, 0.9845, 0.9941, 0.9984, 0.9980, 0.9412, 0.9953, 0.998 respectively. The results show that the proposed QSPR model has good stability and prediction ability.