【作者】 韩萍；

【导师】 刘惠涛；

【作者基本信息】 烟台大学 ， 分析化学， 2011， 硕士

【摘要】 定量结构——性质关系(QSPR)主要研究化合物的结构与其各种物理、化学、生物活性等性质之间的定量函数关系模型,借以指导新化合物的合成及预测未知化合物的各种性质,是计算机在化学中应用的一个特别活跃的领域。近年来,QSPR已经广泛地应用于化学、生物、环境、工程技术等各个领域,并与各种实验设计结合用于色谱分离。本文用QSPR建立模型预测了化合物的光学、电化学及生物性质并将其与响应面实验设计结合应用于毛细管电泳中优化了中药中活性组分的分离条件,结果表明实验设计和QSPR方法与毛细管电泳相结合在毛细管电泳分离方面是一个非常有力的工具。本论文共分六个部分,主要综述了QSPR的研究现状、原理、研究方法、建模方法应用;实验设计与QSPR联合用于毛细管电泳分离测定中药及其制剂中的两种活性组分;启发式方法(HM)和径向基函数神经网络(RBFNN)用于预测化合物的电化学性质、光学性质、对环境的毒性以及生物活性。第一章综述了QSPR的研究现状、研究方法、建模方法及其与实验设计结合在分析化学中的应用。论文第二章将响应面实验设计和毛细管区带电泳结合,建立了一种分离测定厚朴酚、和厚朴酚的简单快速的方法。优化了实验条件并成功用于厚朴和藿香正气水中厚朴酚、和厚朴酚的分析。另外,基于由Box-Behnken设计给出的实验条件及分离度与后出峰组分的迁移时间比(Rs/t),用径向基函数神经网络建立了“3-7-1”结构的人工神经网络非线性模型,并对最佳分离条件下的实验结果进行了预测,预测结果与数学软件的计算值及实验值都一致,相对偏差<5%。。在第三章中,基于所建立的线性和非线性模型,研究了染料中间体9,10-蒽醌的半波电位(E1/2)和其结构之间的定量构效关系。仅由9,10-蒽醌的分子结构计算的描述符就可得到其E1/2。用启发式方法选择最合适的分子描述符并用于线性QSPR模型的建立,用径向基函数神经网络建立非线性模型,并对影响E1/2的结构因素进行了探讨。第四章中,基于50个香豆素的结构及其紫外最大吸收波长(λmax),以启发式方法和径向基函数神经网络建立了线性及非线性QSPR模型。统计结果表明两模型均具有很好的稳定性及预测能力。通过对模型中出现的描述符的讨论,有助于理解影响香豆素类化合物紫外最大吸收波长的重要结构因素。在第五章中,将第四章中的建模方法用于光动力治疗癌症中光敏剂的可见光吸收波长(λ)及其结构关系模型的建立。获得了142个光敏剂的定量结构-可见光吸收波长线性关系模型。交互检验及对外部测试集的预测结果表明模型具有令人满意的稳定性及预测能力。对于新光敏剂的合成具有指导意义。第六章,研究了48个苯胺化合物的梨形四膜虫毒性(-logIGC50)与其结构之间的关系。分别由启发式方法和径向基函数神经网络建立了线性和非线性的定量结构-活性关系(QSAR)模型。并进行交互检验用于评价模型,结果表明两种模型都具有好的稳定性和预测能力。更多还原

【Abstract】 The study of QSPR is to construct models between the structure of compounds and their physical, chemical properties and biological activity so as to develop new compounds and predict the properties of the compounds. It is a very active field in the application of computer in chemistry. In recent years, QSPR has been widely used in many fields of chemistry, biology, environment and technology, and combine with various experimental designs to the separation of many components. In this paper, QSPR was used to develop models to predict some signigicative properties of chemicals. QSPR and response surface experimental design were applied to capillary electrophoresis to optimize the separation conditions of active components in Chinese medicines. The results indicated that the combination of experimental design and QSPR was found to be a powerful tool in predicting separation conditions in CE.There are six parts in this dissertation, including a review on the background, theory, research method, modeling method and applications of quantitative structure property relationship (QSPR); the application of response surface experimental design and QSPR to the separation and determination of two active components of Chinese traditional medicine and preparations by capillary electrophoresis (CE); QSPR study on electrochemistry, optical properties, toxicity to environmental and biological activity of organic compounds by heuristic method (HM) and radial basis function neural network (RBFNN).In the first chapter, we reviews the background, research method, modeling method and the applications in analytical chemistry of QSPR .In chapter 2, a simple and rapid method for the separation and determination of honokiol and magnolol in Magnolia officinalis and its medicinal preparation is developed by response surface methodology and capillary zone electrophoresis. The condition was optimized and successfully applied to the analysis of honokiol and magnolol in Magnolia officinalis and Huoxiang Zhengqi Liquid. In addition, an artificial neural network with“3-7-1”structure based on the ratio of peak resolution to the migration time of the later component (Rs/t) given by Box-Behnken design is also reported and the predicted results are in good agreement with the values given by the mathematic software and the experimental results, RSD<5%. In chapter 3, Quantitative structure-property relationship (QSPR) models correlating half-wave potentials (E1/2) of dye intermediate 9, 10-anthraquinones and their structures were developed based on linear and non-linear modeling methods. Descriptors calculated from the molecular structures alone were used to represent the E1/2 of 9, 10-anthraquinones. Heuristic method (HM) was used to select the most appropriate molecular descriptors and a linear QSPR model was developed. Using the selected descriptors, radial basis function neural networks (RBFNN) was used in the non-linear model development. And the structural factors which have effect on E1/2 were discussed.In chapter 4, in an attempt to develop predictive tools for the determination of the UV maximum absorption wavelength (λmax), QSPR models forλmax of 50 coumarins were developed based on their structures alone. A six-descriptor linear correlation by heuristic method (HM) and a non-linear model using radial basis function neural network (RBFNN) approach were reported. Both of the models indicated satisfactory stability and predictive ability. The descriptors appearing in these models are discussed. This QSPR approach can contribute to a better understanding of structural factors of the organic compounds responsible for the analysis of maximum absorption wavelength.In chapter 5, the modeling methods used in chapter 4 was applied to develop quantitative structure-visible light absorption wavelength (λ) relationship model for photosensitizers in photodynamic therapy for cancer. A linear quantitative structure-visible light absorption wavelength (λ) relationship model of 142 photosensitizers was obtained. The results of cross-validation algorithm and external validation set indicated that the model has a satisfactory statistical stability and predictivity. It could be a potential way for instructing synthesis of this kind of new photosensitizers.In chapter 6, a quantitative structure-activity relationship (QSAR) was developed by the Heuristic method (HM) and Radial basis function neural networks (RBFNN) to study the tetrahymena pyriformis toxicity (explained as–logIGC50) of 48 aniline compounds. Cross-validation was used to evaluate the linear and non-linear models. The results show that both the two models have good stability and predictability for tetrahymena pyriformis toxicity of aniline derivatives.更多还原