应用量子化学参数预测多氯有机物的分配性质

应用量子化学AM1算法求得多氯有机物的量化参数，以此来构建定量结构－性质相关（QSPR）模型。该模型能较好地预测多氯有机物的正辛醇／水分配系数（Kow)和沉积物吸附系数（Koc).Kow和Koc平均误差分别为0．31和0．33个对数单位。模型的稳健性检验结果表明：模型中自变量不存在相关关系；残差呈正态分布；对Koc来说，PCB和PCDD／F在构建模型中起核心作用，剔除该类化合物导致模型相关系数明显下降，农药类及氯硝基苯类化合物的存在降低了模型的精度。

Prediction of n-octanol/water partition coefficients and soil sorption coefficients for polychlorinated organic compounds (PCOCs) using quantum chemical parameters

Quantum chemical parameters based on AM1 molecular orbital method were calculated for the polychlorinated organic compounds (PCOCs) and used to develop quantitative structure|property relationships models.The models were used to predict effectively the octanol/water partition coefficients ( K ow ) and sediments sorption coefficients ( K oc ) for all of the test compounds.The mean residuals of the estimated K ow and K oc of the compounds were 0 31,0 33 log units,respectively.Robustness of models was tested by several methods.Results demonstrated that the independent variables in the models did not interrelated mutually.Residuals appeared normal distribution.PCB and PCDD/F are important in developing models,correlation coefficients were decreased by eliminating these compounds;the accuracy of models were decreased due to the present of pesticides and chlorinated nitrobenzenes compounds in the models.