【作者】 孙立力；

【导师】 李志良；

【作者基本信息】 重庆大学 ， 生物医学工程， 2004， 博士

【摘要】 定量构效关系(QSAR)是重要的现代药学与化学基础理论研究与开发应用领域,其核心内容是考察和分析基本分子结构特征与物化性质或生物活性之间的定量相关关系。分子结构参数化是进行定量构效关系研究的重要组成部分和关键前提所在。随着化学拓扑图论的发展和应用,分子拓扑指数方法不断丰富和迅速发展。由于拓扑指数毋需实验测定,可以直接从分子结构获得,因此是一种简便而直接的分子结构参数化方法。本实验室提出的分子电性距离矢量(MEDV)是一种描述分子二维结构的拓扑描述子,其中引入了原子属性和原子类型的概念,适用于含多个杂原子、饱和键与不饱和键、环和非环等结构特征,可适用于药物与生物及有机与无机等多种类型的分子。本文在进一步探讨MEDV-4的适用性的基础上,分别从分子微环境对电性的影响、相对距离和电性效应等方面对原有的MEDV-4进行了改进和扩展。本文开展的工作有以下几个方面:考虑分子中杂化对原子轨道电负性的影响,建立了新的分子电距矢量MEDVh。分别采用MEDV-4和MEDVh对几组雌激素的分子结构进行了表征。MEDV-4与两组雌二醇化合物和一组2-苯基吲哚化合物生物活性的定量相关模型的复相关系数R均达到0.92以上,留一法交互检验RCV均在0.85以上。MEDVh与黄体酮化合物生物活性的定量相关模型的复相关系数R=0.9546,交互检验RCV=0.8826。按连接氢原子数目划分原子类型的H-MEDV比MEDV-4能更好地表征两组抗艾滋病药物HEPT和一组二氢叶酸还原酶抑制剂的分子结构,与HEPT生物活性建立的定量相关模型复相关系数均在0.90以上,留一法交互检验预测值与实验值的相关系数也都在0.85以上,与二氢叶酸还原酶生物活性建立的模型,R和RCV在0.80以上。进一步探讨MEDV-4的适用性。将MEDV-4应用到对醇、醛酮化合物和多氯二苯醚的多种理化性质的QSPR研究中,应用多元线性回归技术,建立了MEDV-4与这些化合物的多种理化性质的定量相关模型,以模型估计值与实验值之间的复相关系数R和标准偏差SD评价模型对内部样本的估计能力,留一法交互检验预测的外部样本的性质与实验值之间的相关系数RCV和标准偏差SDCV评价模型的稳定性和对外部样本的预测能力,并采用逐步多元回归方法筛选变量。对醇的沸点、水溶性、醇/水分配系数、摩尔体积、摩尔折光率、分子(摩尔)表面积等6种理化性质所建立的模型,其R值均在0.99以上,留一法交互检验RCV都在0.98<WP=6>以上;对醛酮化合物沸点建立的模型结果为R=0.9911和RCV=0.9881;对多氯二苯醚的蒸气压(P)、水溶性(SW)和醇/水分配系数的模型的R达到了0.98以上,RCV则都在0.97以上。表明MEDV-4能够良好表征这些化合物的分子结构,具有优良的估计与预测能力。将原子电距矢量AEDV对88个芳香醛酮化合物中羰基碳原子的13C NMR化学位移相关联,建立的定量相关模型具有良好的对内部样本的估计能力和对外部样本的预测能力,复相关系数R和留一法交互检验RCV均在0.90以上。而对原子局部微环境较复杂的体系,确定了划分5种原子类型的新型原子电性距离矢量AEDV-5,并将其应用于对一组雄甾酮分子中所有碳原子的13C NMR化学位移的定量相关研究中,模型估计值与实验值的相关系数R和交互检验RCV都达到了0.99以上。与AEDV相比,AEDV-5能更好地表征较为复杂的原子局部微环境。对存在多路径的含环化合物,建立了等效电阻法、类效电阻法,以改进多路径下相对距离的计算方法。含环分子中两原子之间存在有多条路径,常规MEDV-4中以最短路径计算相对距离,这种简化可能带来的某些信息漏失可以采用等效电阻法或类效电阻法加以修正。通过对两组烷烃化合物的QSPR研究,得出的初步结论是:与常规MEDV-4相比,等效电阻法和类效电阻法计算的相对距离包含的信息量更全面,由此计算出的分子电距矢量与两组化合物沸点建立的定量相关模型具有至少相当甚至更好的估计能力和预测能力。而类效电阻法较之等效电阻法具有更好的应用前景。引入电负性效应的概念,建立了能反映电负性随化学键传递而逐渐衰减这一特点的新型分子电距矢量——衰减电距矢量(AMEDV-4)。对两组烷烃的沸点的QSPR研究表明,AMEDV-4能良好表征这两组分子的结构特征,并且通过调整衰减指数,可以使模型的效果得到一定程度的改善。与采用常规MEDV-4建立的模型相比,对两组烷烃化合物建立的与最佳衰减指数对应的定量相关模型,R值有明显的提高,分别达到了0.9946和0.9966,表明含衰减指数的AMEDV-4具有比常规MEDV-4更为优越的性能。这种新型的描述子可望得到进一步完善和应用。更多还原

【Abstract】 Quantitative Structure-Activity Relationship (QSAR), which investigates the quantitative relationship between the molecular structural parameters and biological activities or dependent functions, is one of the most important fundamental fields in pure and applied chemistry and pharmacy. Molecular structural characterization is an important technique in QSAR studies. With development and application of graphic theory, more and more topological indices have been developed rapidly and widely. One of merits for topological indices is that they can be derived directly from molecular structure without experiments, and so they have been applied in QSAR as simple, direct and effective molecular structure parameterization methods. Both molecular and atomic electronegativity distance vector (MEDV/AEDV), which have been developed in our laboratory to describe two-dimensional (2D) molecular topology and local microenvironment around various chemically equivalent carbon atoms, are applicable to various molecular global and local structures which contain the multiple bonds and/or heteroatoms by introducing the concepts of relative electronegative (REN) and relative bond length (RBL). In this dissertation, investigation on applicability of MEDV-4/AEDV as the extended and improved 2D descriptors are performed as viewed from the molecular microenvironment, relative distance and electronegativity effect.The main contents are as follows:Taking the effects of various hybridization on atomic electronegativities into account, a set of novel structural descriptors, called molecular electronegativity- distance vector with hybridization (MEDVh), has been developed. MEDV-4 and MEDVh have been used to express the structures of several groups of estrogens. The correlation coefficients R of the QSAR models that related MEDV-4 to biological activities of two sets of estradiols and a set of 2-phenylindoles are higher than 0.92, and the RCV are higher than 0.85. The QSAR models related MEDVh to 21 progesterones has R=0.9546 and RCV=0.8826.Hydrogen-association classified molecular electronegativity-distance vector (H-MEDV) can be used express molecular structures of two sets of anti-HIV HEPT agents and one set of dihydrofolate reductase inhibitors and the results are better than that of MEDV-4. The R and RCV for QSAR models for HEPT are higher than 0.90 and <WP=8>0.85, respectively. Both the R and RCV of the QSAR models for activities of dihydrofolate reductase inhibitors are higher than 0.80.The applicable fields of MEDV-4 are extended. MEDV-4 has been related to various of physicochemical properties of alcohols, aldehydes, ketones and polychlorinated diphenyl ethers, and the quantitative relationship models have been constructed by using multiple linear regression technique. The correlation coefficient R and standard deviation SD between the estimated properties and observed experimentally properties have been used to evaluate the estimation abilities for internal samples. The correlation coefficient RCV and standard deviation SDCV between the properties observed experimentally and the properties predicted by leave-one-out crossvalidation technique have been used to evaluate the predictive abilities of models. The R of the models for boiling points, water solubility, octane/water partition, molecular total surface area, molecular volumes and molecular refractions are higher than 0.99, and RCV are higher than 0.98. The models which related MEDV-4 to boiling point of aldehydes and ketones has R=0.9911 and RCV=0.9881. The R and RCV of models for liquid vapor pressures, n-octanol/water partition coefficients and liquid water solubilities of polychlorinated diphenyl ethers are higher than 0.98 and 0.97, respectively. The results show that MEDV-4 can be used to well express the structures of these organic compounds.The descriptors, AEDV, are used to express the local microenvironment of various chemically equivalent carbon atoms in 88 aromatic aldehydes and ketones and the quantitative relationship model exhibits good estimation ability fo更多还原