【作者】 周丽平；

【导师】 李志良；

【作者基本信息】 重庆大学 ， 分析化学， 2002， 硕士

【摘要】 计算机辅助分子设计(CAMD)与定量构效关系(QSAR)的研究是药物设计与研制的关键环节和重要基础。如何产生出表征分子结构特征的结构参数是定量构效关系研究成败的关键所在。作者根据分子二维拓扑结构，以原子电性和原子间距描述不同分子环境中各类原子之间的相互作用，并结合原子类型划分方案，应用了用于表征药物主要是有机药物分子结构的新型矢量型结构描述子：分子电性距离矢量(MEDV)。在分子电距矢量的基础上进行拓展并对几类生物活性物质进行了定量构效关系及定量构性关系研究。本文的主要内容和研究结果如下： (1)将分子距边矢量(MDEV)用于100种多环芳烃的化学结构表征，对其毛细管气相色谱保留值进行定量相关性研究，发现存在良好的线性相关关系，相关系数R=0.9899，采用留一法进行交互检验，得相关系数R=0.9856，可见模型稳定性很好，可用于多环芳烃色谱保留指数的分子建模与定量预测。 (2)根化路径数矢量能良好表征151个烷烃碳原子C-13核磁共振化学位移，相关系数达R=0.944。另外，对伯仲叔季类碳原子分别进行建模，相关系数分别为0.993，0.983，0.961，0.992，经交互检验说明模型稳定。 (3)对几类抗艾滋病类药物分子包括肽类HIV-1蛋白酶抑制剂、一组电子等排衍生物和环尿素类化合物采用新型的分子电距矢量进行结构表征，并与其抗艾滋病病毒活性进行相关研究，分别取得良好结果。对寻找高活性抗HIV药物有一定的指导意义。 (4)将以分子电性距离(MED)矢量为基础的新型矢量型结构描述子按氢分类的分子电性距离矢量(H-MEDV)对28个具有肝细胞毒性的苯胺类药物的结构参数和对三种肝细胞逸出酶活性进行定量构效关系建模，模型相关系数分别为0.924，0.874，0.818。所建立的三组多参数模型稳定并具有良好的估计预测能力。 (5)应用表征药物主要是有机药物分子结构的新型矢量型结构描述子：按氢分类的分子电性距离矢量(H-MEDV)和分子全息距离矢量(MHDV)对40个氨基喹啉类化合物进行了结构参数表征和抗疟原虫活性相关关系建模分析。并运用多元线性回归方法和逐步回归方法来构造数学模型(R＞0.9)。所建立的多参数模型稳定并具有良好的估计预测能力。更多还原

【Abstract】 Computer-assisted molecular design (CAMD) and quantitative structure-activity relationship (QSAR) are a key fundament and an important procedure of drug design, research and development. Molecular structural characterization heavily decides the success of the QSAR study. In this thesis, a type of novel structural vectorial descriptors, called molecular electronegativity-distance vector (MEDV), which was first developed by our laboratory, have been extended to characterize several complex biological molecules. The main contents and some conclusions are given as follows:1. The original MEDV was used to characterize the chemical structures of 100 polycyclic aromatic hydrocarbons (PAHs) and quantitative relationship between structure and the gas chromatographic retention index has been established by using multiple linear regression (MLR) technique. The correlation coefficients is R =0.9899. The stability and prediction capacity for external samples of the model have been tested by the cross validations method with leave-one-out (LOO), the correlation coefficients is also as high as R cv=0.9856.2. The atomic rooted path vector (ARPV) of various lengths was developed to relate the local circumstance of the carbon atom in 151 alkanes and their chemical shifts of C-13 nuclear magnetic resonance (NMR). For all types of the atoms, the correlation coefficient of atomic modeling equation is 0.944. For primary, secondly, tertiary and quaternary carbon atoms, the correlation coefficients are 0.993, 0.983, 0.961, 0.992, respectively. All of the models were tested to have quite good stability through cross-validation (CV) with the leave-one-out (LOO) procedure.3. The molecular electronegativity-distance vector was extended into several modified forms, and then has been used to express the chemical structures of three kinds of anti-HTV drug molecules including the peptide HTV protease inhibitor, a set of electronic isostere derivatives and cyclic ureas derivatives. Good results could be usefully in the research of new anti-HIV agents with higher anti-HIV activity.4. A novel set of structural descriptors with ten elements, called the hydrogen-association classified molecular electronegativity-distance vector (H-MEDV), was developed to describe the chemical structure of 28 hydropyridine derivatives and reasonable molecular modeling results were achieved with a relatively high correlation coefficient (#=0.924, 0.874, 0.818, respectively) by a multiple linear regression (MLR). For three kinds of biological activities, the models are all have good stability and prediction ability.5. The hydrogen-association classified molecular electronegativity-distance vector (H-MEDV) and a novel descriptor, MHDV, were used to express the structures of 40 aminoquinolines with good relationship (R>0.9). The correlation coefficients (R) between the antiplasmodial activities observed experimentally and the antiplasmodial activities predicted by leave-one-out (LOO) method of cross validations are also higher than 0.8.更多还原