酚类化合物色谱保留指数的估计与预测  

Estimation and Prediction of Chromatographic Retention Index of Phenols

作　　者：周原[1] 刘新玲[1] 曾辉[2] 李志良[2] 

机构地区：[1]湖南工程学院化学化工系,湖南湘潭411104 [2]重庆大学生物工程教育部与重庆市重点实验室,重庆400044

出　　处：《湖南工程学院学报（自然科学版）》2006年第2期63-67,共5页Journal of Hunan Institute of Engineering(Natural Science Edition)

摘　　要：用根据非氢原子类型分类、基于非氢原子相对电负性和非氢原子间距离等进行计算得到的分子电性距离矢量(MEDV)为描述子,对酚类化合物结构进行表征.用多元线性回归(MLR)方法,研究并建立了酚类化合物定量结构-色谱保留(QSRR)关系的10变量模型,其相关系数为0.9512;采用SPSS统计处理软件对变量进行逐步回归(SMR)筛选后,建立了8变量的QSRR模型,其相关系数为0.9508.上述模型对102种酚类化合物色谱保留的预测值与实验值能较好吻合,留一法交互检验的结果分别为Rcv=0.9269和Rcv=0.9377.结果表明所建模型具有良好的稳定性和预测能力.A novel molecular electronegativity - distance vector （ MEDV ）, which has been developed according to classification of the types of non - hydrogen atoms,is used to describe the chemical structure of a series of phenols. The MEDV descriptors can be calculated through a relative electronegativity and a relative distance of the non - hydrogen atoms. The QSRR model of 10 varieables of the phenols is built by multiple linear regression （MLR） with the correlation coefficient （R） of molecular modeling being 0. 9512, and a new model of 8 descriptors can be obtained by stepwise multiple regression （SMR） using SPSS software with the correlation coefficient （R） of modeling being 0. 9508. The predicted and observed gas chromatographic retention indices of studied compounds have been proved to be in accord. Cross-validation of the models, which contain selected vectors, are performed by leave - one - out procedure （LOO） and the satisfied results with correlation coefficients of 0. 9269 and 0. 9377 respectively. The results show that the models constructed can provide estimation stability and favorable predictive ability.

关 键 词：描述子 分子电距矢量 酚类化合物 定量结构色谱保留相关 

分 类 号：O641[理学—物理化学]