【作者】 贺晓莹；

【导师】 杨国生；

【作者基本信息】 山东大学 ， 分析化学， 2008， 硕士

【摘要】 化合物的定量结构-色谱保留相关（OSRR）是色谱学基础理论研究的重要课题,它将被分析物的色谱保留数据与其分子结构特征相关联。QSRR方法在研究分子结构参数与色谱保留性质的递变规律、实现色谱过程中保留数据的预测、选择分离条件及探索色谱保留机制等方面有重要意义,如今越来越受到色谱工作者的重视。本论文选取意义明确的结构参数,运用多种数理统计方法,构建了含有氮、氧、硫等杂原子的几类化合物集合的气相色谱保留指数模型,结果令人满意。根据建立的QSRR方程探讨了化合物在不同极性色谱柱上的保留机理。论文主要包括以下几方面的工作:（一）构建了烷基吡啶类化合物在两种不同极性固定相（SE-30和PEG-351）上的气相色谱保留指数的QSRR模型,运用遗传函数近似法得到的QSRR方程具有高度的相关性（R₁²=0.990,R₂²=0.989）,优选出的结构参数包括了量子化学参数和物理化学参数,经方程得到的保留指数估计值与实验测定值较为一致:还构建了硝基苯甲酸酯类化合物在两种不同极性固定相（SE-30和OV-351）上的OSRR模型,运用多元线性回归方法也得到了高度相关的QSRR方程（R₁=0.9975,R₂=0.9910）,采用留一法（LOO）进行交互检验证明了模型的稳定性。（二）选用烷基苯酚类化合物和饱和酯类化合物作为研究对象,研究了化合物定量结构与色谱保留的相关关系。用遗传函数近似法建立了烷基苯酚类化合物的气相色谱保留指数的QSRR方程,相关性良好（R²=0.972）,经方程得到的保留指数估计值与实验测定值较为一致;又分别用遗传函数近似法和多元线性回归法,构建了在四种不同极性固定相（SE-30、DC-710、XE-60、Silar 5CP）上的饱和酯类化合物的QSRR模型,得到的QSRR方程均具有高度的相关性。由多元线性回归得到的方程其相关系数分别为:R₁=0.9969,R₂=0.9947,R₃=0.9959,R₄=0.9935;由遗传函数近似法得到的方程其相关系数平方分别为:R₁²=0.9994,R₂²=0.9972,R₃²=0.9985,R₄²=0.9974。对比可知对于该体系用遗传函数近似法得到的OSRR方程相关关系更好。（三）研究了不饱和取代的硫醚类化合物在四种不同极性固定相（ApiezonM、OV-17、TritonX-305、PEG-1000）上的定量结构与色谱保留的相关关系。用多元线性回归法建立的QSRR方程具有高度的相关性（R₁=0.9984,R₂=0.9982,R₃=0.9940,R₄=0.9911）,用留一法验证了模型的稳定性。通过对含氮、氧、硫等杂原子的几类化合物的QSRR研究,均得到相关性显著的QSRR模型,证明了该方法的可行性。可以用来有效地预测化合物的色谱保留值,并为选择化合物最佳的色谱分离条件提供依据,从而减少实验的盲目性,同时节省了时间和大量的人力物力。更多还原

【Abstract】 Quantitative structure-retention relationship（QSRR）of compounds is an impotant topic in chromatographic basis study.It relate chromatographic retention data of analyte with its molecular structure.QSRR has significant meaning in many way.It can give information on the relation between molecular structure parameters and chromatographic retention mechanism.It can also help us to predict chromatographic retention data and to select the condition of separate.Nowadays QSRR is more and more recognized by the chromatographic experts.This paper begins with selecting structure parameters which have specific chemical meaning as molecular descriptors,and then using statistical method,such as multiple linear regression（MLR）and genetic function approximation（GFA）,to develop QSRR models.The compounds studied in this paper contain O-、N-、Sgroups. The results indicate that the QSRR models developed are very satisfactory. The major work is described as follows:（1）QSRR models for the gas chromatographic retention indices of alkyl pyridines on two different polarity stationary phases（SE-30 and PEG-351）have been developed. By using genetic function approximation method,we obtained empirical equations with high correlation（R²₁ =0.990,R²₂=0.989）.RIpre came from QSRR equation was close to RIref came from reference.Then nitryl-benzoic esters were studied. QSRR models for the gas chromatographic retention indices of nitryl-benzoic esters on two different polarity stationary phases（SE-30 and OV-351）have been developed by using multiple linear regression method.Empirical equation with high correlation（R₁=0.9975,R₂=0.9910）was obtained.Use leave-one-out method to prove the stability of models.（2）Choosing alkylphenols and saturated ester as object,QSRR of compounds contained O- were studied.By using genetic function approximation method,we obtained QSRR functions of alkylphenols with high correlation（R²=0.972）.Then the retention indices of saturated ester on four different polarity stationary phases（SE-30. DC-710,XE-60,Silar 5CP）was studied.QSRR equations were respectively built with multiple linear regression method and genetic function approximation method.QSRR equations were both good.But by contrast the two method,we found that genetic function approximation method fit to this series of compounds and the equations built by genetic function approximation method were better.The statistical result came from multiple linear regression method was:R₁=0.9969,R₂=0.9947,R₃=0.9959, R₄=0.9935.The result came from genetic function approximation method was: R²₁=0.9994,R²₂=0.9972,R²₃=0.9985,R²₄=0.9974.They were all considered as high correlation.（3）QSRR models for the gas chromatographic retention indices of sulfur ether on four different polarity stationary phases（Apiezon M,OV-17,Triton X-305,PEG-1000） were developed.The equations were built with multiple linear regression method. Empirical equation with high correlation（R₁=0.9984,R₂=0.9982,R₃=0.9940, R₄=0.9911）was obtained.Then use leave-one-out method to prove the stability of models.By studying the compounds contained O-,N-,S- groups,QSRR models with high correlation were obtained.So the feasibility of QSRR method can be proved. QSRR models can be used to predict the chromatographic retention datas and to select the best separation conditions.It can help us save time of experiment and reduce the blindness in the experiment.更多还原