30种多氯有机物的沉积物吸附系数定量结构性质关系研究  

QSPR study on sediment sorption coefficient of thirty polychlorinated organic compounds

作　　者：焦龙[1] 李华[2] 

机构地区：[1]西安石油大学化学化工学院,陕西西安710065 [2]西北大学化学与材料科学学院,陕西西安710069

出　　处：《计算机与应用化学》2012年第4期409-412,共4页Computers and Applied Chemistry

基　　金：陕西省自然科学基金项目(2010JQ2003;0);国家自然科学基金项目(20975081);西安石油大学博士启动基金(YS29030819);西北大学合成与天然功能分子化学教育部重点实验室开放基金项目(kf09014)

摘　　要：研究了30种含氯有机物沉积物吸附系数(Koc)的定量结构性质关系(QSPR)模型。模型的自变量由遗传算法从多个结构描述符中选择得到,校正模型则采用多元线性回归方法建立。应用留一交叉验证及外部测试集验证对所建立的模型进行了检验。留一交叉验证中模型的预测均方相对误差为8.13,外部测试集验证中模型的预测均方相对误差为8.88。这说明所建立模型对于未知样本有较好的预测能力,可用于预测含氯有机物的Koc值;遗传算法结合多元线性回归方法是一种合理可行的建立含氯有机物沉积物吸附系数QSPR模型的方法。A quantitative structure property relationship （QSPR） model for predicting the sediment sorption coefficient, Koc, of 30 polychlorinated organic compounds was developed. Multivariate linear regression （MLR） was used to develop the calibration model for Koc. The two independent variables of MLR were selected from six structural descriptors of the investigated compounds by using genetic algorithm （GA）. Leave one out cross validation and external test set validation were carried out to assess the predictive ability of the developed model. In the leave one out cross validation, the RMS%RE of prediction is 8.13. In the external test set validation, the RMS%RE of prediction is 8.88. It is demonstrated that the developed model is practicable for predicting Koc of the investigated compounds. In addition, MLR combined with GA is shown to be a practicable method for developing QSPR model for Koc of polychlorinated organic compounds.

关 键 词：定量结构性质关系 沉积物吸附系数 含氯有机物 遗传算法 多元线性回归 

分 类 号：TQ015.9[化学工程] TP391.9[自动化与计算机技术—计算机应用技术]