以非氢原子自身及非氢原子之间的关系为分子结构描述符,对35个醛酮类化合物进行了结构表征。采用多元线性回归和偏最小二乘回归的方法建立了该类化合物结构与色谱保留指数之间的关系模型,两模型的复相关系数(R)分别为0．987和0．981,标准偏差(SD)分别为25．259和32．240。采用“留一法”交互检验和外部样本预测的方法对模型的稳健性和预测能力进行了评价,交互检验的复相关系数( RCV )分别为0．981和0．958,标准偏差(SDCV)分别为34．292和39．652；外部预测的复相关系数(Rtest)分别为0．991和0．990,标准偏差( SDtest )分别为25．749和26．776。结果表明所构建的分子结构描述符能够恰当表现该类化合物结构特征,所建模型具有良好的稳健性和预测能力。%The non-hydrogen atoms themselves and the relationships between them as structural descriptors and then 35 compounds of aldehydes and ketones were characterized. Two quantitative structure retention relationship models of these compounds were es-tablished through multiple linear regression and partial least squares regression methods. The correlation coefficients( R) of the two models were 0. 987 and 0. 981,and the standard deviations(SD) were 29. 259 and 32. 240. The“leave-one-out”cross validation and external sample test were used to evaluate stability and predictive ability of the two models. The multiple correlation coefficients of cross-validation(RCV)were 0. 981 and 0. 958,and the standard deviations(SDCV)were 34. 292 and 39. 652;the multiple correlation coefficients of external tests(Rtest) were 0. 991 and 0. 990,and the standard deviations(SDtest)were 25. 749 and 26. 776. The results showed that the molecular descriptor constructed in this paper can properly express structural characteristics of these compounds, and the models have good stability and predictability.
展开▼
