应用偏最小二乘回归技术建立了102种酚类化合物气相色谱保留指数与分子全息结构间的相关模型,在最佳建模条件下得到非交叉验证相关系数(r2)为0.965,交叉验证相关系数(qLOO2)为0.963.从102种酚类化合物中随机选出68种作为训练集,其余作为测试集,来验证分子全息QSRR模型的预测能力和稳健性.在最佳建模条件对训练集进行偏最小二乘回归分析,r2为0.967,qLOO2为0.927.用训练集数据所建立的QSRR模型预测测试集中酚类化合物的色谱保留指数,结果表明,基于训练集所建立的QSRR模型具有很高的预测能力和稳健性,可以对测试集酚类化合物的气相色谱保留指数进行很好的预测.此外,利用最佳全息定量结构保留关系(HQSRR)模型的色码图,探讨酚类化合物中的不同侧链基团对其色谱保留性质的影响,及其在固定相上的色谱保留机理.
A quantitative structure-retention relationship (QSRR) model was constructed for gas chromatographic retention indices (GC-RI) and a molecular hologram of 102 phenols using a partial least-squares (PLS) regression analysis. The model showed high statistical quality with non-cross validation correlation coefficient r2 of 0. 965, and cross validation correlation coefficient qtoo^2 of 0. 963. In order to veri(y robustness and prediction capacity of the model, 68 phenols were selected randomly from the data set as the training set, while the rest acted as the testing set. The result of PLS regressive analysis of the training set yielded r2 of 0. 967 and qLoo^2 of 0.927, and the model showed excellent ability to predict the GC-RIs of phenols in the testing set. Furthermore, the retention behavior of the compounds in GC stationary phase was discussed, and the effects of different groups in the side-chain of phenols on the interaction between the phenols and the stationary phase were explored using color codes of hologram QSRR. This provided useful guidelines for the retention rules of phenols and related compounds.