大气中痕量挥发性有机物的结构表征和保留时间的估计与预测

ESTIMATION AND PREDICTION OF CHROMATOGRAPHIC RETENTION TIME OF TRACE VOLATILE ORGANIC COMPOUNDS IN ATMOSPHERE

采用分子电性距离矢量(MEDV)表征大气中痕量挥发性有机物的分子结构,同时采用逐步回归结合统计检测对模型进行变量筛选,建立了大气中痕量挥发性有机物定量结构-色谱保留(QSRR)关系的8个变量和5个变量模型,两种QSRR模型的建模计算值复相关系数(R)分别为0.937和0.931;留一法(leave-one-out)交互校验复相关系数(RCV)分别为0.901和0.906,表明模型具有良好的估计能力与稳定性.

Molecular electronegativity-distance vector（MEDV） is used to describe chemical structures of 39 volatile organic compounds （ VOC ） . The QSRR model of 8 MEDV variables is built by multiple linear regression （MLR） with the correlation coefficients（R） being 0. 937. A new model of 5 MEDV descriptors can be obtained by stepwise multiple regression （ SMR ） with the correlation coefficient （ R ） being 0. 931. The predicted and observed gas chromatographic retention times are proved to be in accord with cross-validation （CV）, performed by leave-one-out （LOO） procedure with good results of correlation coefficients （ RCV ） being 0. 901 and 0. 906 for all 8 MEDV variables and 5 MEDV variables respectively. It is concluded that the models can provide satisfactory estimation stability and favorable predictive ability.