以电性距离矢量 Mt表征165种非离子性有机物(NOC)的分子结构. 利用多元回归方法建立了122种NOC的生物富集因子(lg BCF)与32个 Mt的数学模型,其相关系数R为0.976. 经逐步回归建立最佳四变量(M15, M17, M36, M91)模型,其R为0.960;并以Jackknife法检验,其LOO交互检验系数(q2)为0.915. 这4个 Mt参数揭示了影响NOC生物富集因子的主要因素,结果表明该模型具有良好的估算能力与鲁棒性. 以111个NOC为训练集的生物富集模型作用于11个NOC的检验集,显示出良好的预测能力.
Molecular electronegativity-distance vector (Mt) was used to describe the molecular structures of 165 nonionic organic compounds in this paper. A quantitative linear relationship between 32 Mt descriptors and bioconcentration factor (lg BCF) of 122 NOC was set up using the method of multiple linear regression (MLR). The satisfactory model of 4 parameters (such as M15, M17, M36, M91) was built by stepwise multiple regression and the Jackknifed method (the leave-one-out, LOO), of which the calibrated correlation coefficient (R) was 0.960 and the validated correlation coefficient (q2) was 0.915. The 4-variable model has shed light on some main structural factors influencing the BCF on NOC, which shows the good estimation ability and robustness. A prediction power for the external sample of 11 NOC was validated by the model built from the training set with 111 NOC.