【作者】 张巧霞；

【导师】 李志良；

【作者基本信息】 重庆大学 ， 分析化学， 2007， 硕士

【摘要】 组合合成是建立在高效、快速、平行合成基础之上,这种合成新方法步骤较少,但底物和试剂较多样,故生成的化合物数量巨大,因此组合化学以其制备能力大的优点在生物学合成和化学合成中发挥了巨大作用,尤其在新药的研究与开发领域。多组分反应是组合合成的有力工具,是组合合成中模块合成策略的有用手段;得到的产物是在单一反应过程中生成,并且目标产物产率较高;多组分反应技术还节省时间、劳动、费用、资源,是高效地寻找到较理想新药的有力工具。本文以现代组合化学中多组分反应的研究与应用为主线,探索了常规液相平行组合合成技术在三组分有机合成中的应用,重点进行了Mannich反应研究及其定量构效关系研究。定量构效关系研究是建立在实验室已有的良好基础上,基于分子二维结构信息,结合逐步回归、多元线性回归变量筛选方法和QSAR建模技术,采用原子电性作用矢量(AEIV)结合原子杂化状态指数(AHSI)进行定量构谱关系研究(QSSR),将新型分子电性作用矢量(MEIV)、启发式算法(HM)应用于化合物合成产率预测,取得了较好的结果。本文开展的工作主要有以下几个方面:(1)常规液相平行组合合成Mannich碱。采用常规液相平行组合合成技术,进行多次实验合成大量产物,得到一个较小的化合物库。主要探讨了由直链酮、环酮以及芳香酮在95%乙醇中参与的Mannich反应,并考察了反应条件、催化剂及试剂结构对反应的影响。得到了50余个新型Mannich碱,合成产率在13.25%-80.48%,固体产物均进行熔点测定,部分化合物进行谱图分析。(2)采用反映原子化学微环境的原子电性作用矢量(AEIV),并结合原子杂化状态指数(AHSI),对参与Mannich反应的反应物:醛、胺、酮分别进行定量构谱关系研究。借助多元线形回归技术建立起6参数模型,醛、胺、酮QSAR模型复相关系数分别为0.965、0.981、0.949,对应交互校验的复相关系数分别是:0.830、0.978、0.944;构建外部预测集对模型进行检验,得到的预测复相关系数分别是:0.965、0.982、0.946,LOO-CV结果分别是:0.835、0.978、0.927。结果表明,采用原子电性作用矢量结合原子杂化状态指数方法用于醛、胺、酮定量构谱研究是可行的,模型的预测能力较强,稳定性良好。(3)新型分子电性作用矢量用于化合物产率定量构效研究。分别对二组分Wittig反应和三组分Mannich反应的目标化合物产率进行预测。并借助于多元线形回归技术,逐步回归分析方法对变量进行筛选。42个Wittig目标化合物样本建立起7参数QSAR模型,模型复相关系数以及交互校验的复相关系数分别为0.985、0.973;选用外部集对模型健壮性、预测能力进行考察,得到的模型相关系数、留一法交互检验相关系数分别是:0.984、0.965,模型预测能力较高、稳定性良好。对67个Mannich碱产率进行预测,建立4变量QSAR模型,结果如下:R=0.578,RCV=0.470;外部集检验,模型相关系数、交互检验相关系数分别是0.565、0.428;模型质量较好。(4)启发式算法(HM)用于Mannich碱合成产率预测。采用一种新的化学计量学方法对Mannich碱合成产率进行预测,得到良好的结果。启发式算法得到10参数QSAR模型结果:R=0.735、RCV=0.520。剔除2个异常值之后模型结果:R=0.860,RCV=0.770。为了检验采用启发式算法建立模型稳定性及其预测能力,进行了两次外部集检验:①外部集对模型检验,从42个化合物中随机抽出10个化合物作为预测集,其余作为训练集进行建模。QSAR模型复相关系数0.926,留一法交互检验相关系数是0.820,模型优良。②实验检验,自行设计了5个Mannich碱组成预测集进行检验,得到理论产率,再进行实验合成。试验结果表明,模型有较好的预测能力。采用HM方法对Mannich碱合成产率进行预测结果良好。论文在分子结构表征上,主要采用2D-QSAR分子结构表征技术,从定量构效关系建模的结果上看,文中采用的方法能够较好地表征相关体系的分子结构,预测化合物合成产率,对化学合成生产实践具有重要的指导意义。更多还原

【Abstract】 Combinatorial synthesis is one of combinatorial chemical techniques based on high efficient, fast and parallel synthetic methodology, which has less experimental steps, however, owe to all sorts of reactants, enormous compounds library can be gotten by one experimental trial. Therefore, combinatorial chemistry has a great superiority on synthetic biology and chemistry, especially in the study and development of novel drugs. A multiple-component reaction is a considerably useful tool for important instrumentality of module synthesis strategy in combinatorial synthesis; while higher target product can be obtained by a single reaction process than other chemical reaction. One can save time, labor, money, resource, and so on by using a multiple-component reaction skill, which is an important and perfect tool to search and find novel drugs.In this thesis, the research and application of multiple component reaction in modern combinatorial chemical is taken the mainline, the realization of general parallel solution combinatorial synthesis is explored in the three components reaction (3CR) preliminarily. Some aspects are discussed in details on Mannich reactions for organic syntheses and quantitative structure activity relationships (QSAR). Based on some good studies fulfilled on our laboratory, all chemometrics research in the paper are based on the two dimensional information of molecular structure, multiple linear regression (MLR) and stepwise multiple regression (SMR) are used to correlate the two dimension vector of molecules with their data. Beginning with the two dimensional structure of molecular, a quantitative structure-spectroscopy relationship (QSSR) method based on both novel atomic electronegativity interaction vector (AEIV) and atomic hybridation state index (AHSI) is developed for expression of local chemical microenvironment and atomic hybridation state. In the prediction to yields of compounds, novel molecular electronegativity interaction vector (MEIV) and heuristic method (HM) are applied and extended, respectively, most obtained models are favorable. The main contents are as follows:(1) General parallel solution combinatorial synthesis method is used to synthesize Mannich bases. Three components reaction Mannich condensation system is taken as the templet, a small amount of structurally similar target molecules, considered as a mini-size combinatorial compound library, were synthesized under almost coincident reaction conditions through varying the three substrates or reagents. The reaction is studied by classifying chain ketone, cycle ketone and aroma ketone, which is realized in the 95% ethanol medium. In this way, the combinatorial synthesis might be simulated. Meanwhile, the effect of reaction conditions, catalyst and reagent structure on the synthetic reaction were discussed in detail. About 50 new compounds were synthesized with yields from 13.25 to 80.48%, solid compounds are mensurated melting points, and some molecule structures of compounds are analysed by spectroscopy methods.(2) Quantitative structure-spectroscopy relationship (QSSR) research of 13C Nuclear Magnetic Resonance of aldehyde, ketone and amine. Beginning with the topologic indexes of two dimensional structure of molecular, which are bond length and electronegativity, a quantitative structure-spectroscopy relationship method based on both novel atomic electronegativity interaction vector (AEIV) and atomic hybridation state index (AHSI) is developed for expression of local chemical microenvironment and atomic hybridation state. By using these ways, we successfully model carton atoms 13C NMR chemical shift from aldehyde, amine and ketone. The correlation coefficient(R) values of QSSR model estimation with 6 variables based on multiple linear regression analysis (MLR) are 0.965, 0.981 and 0.949, the leave-one-out (LOO) cross-validation (CV) RCV are 0.830, 0.978, 0.944, respectively. Afterwards, these models are tested by 13C NMR chemical shifts of aldehyde, ketone and amine at random with the prediction correlation coefficients being 0.965, 0.982, 0.946, the leave-one-out cross-validation RCV are 0.835, 0.978, 0.927. The results show that the novel vector AEIV is an excellent structural index with satisfactory estimation stability and favorable generalization.(3) Prediction of yields of synthesis based on novel molecular electronegativity interaction vector (MEIV). Quantitative structure-produce relationship (QSPR) research are carried out for 2 different groups of Wittig and Mannich bases. Multiple linear regression analysis (MLR) is used to built model, and some variables are selected by stepwise multiple regression (SMR). A 7 variables which are selected by SMR model is obtained, The correlation coefficient(R) values of QSPR model estimation based on multiple linear regression analysis is 0.985, the leave-one-out cross-validation correlation coefficient(RCV) is 0.973. Besides, to test the prediction ability and the stability of the model, an exterior set is built at random, the prediction correlation coefficients being RMM=0.984 and RCV=0.965. The results show the QSPR model is favorable. The same way is used to predict produce of 67 Mannich bases, which model is made of 4 variables by SMR. The correlation coefficient(R) values is 0.578, RCV is 0.470, exterior set estimation results are R=0.565, RCV=0.428, the model is stability. Above discussions indicate MEIV method is a good way to predict produce of Wittig and Mannich bases ultimately, and the development of MEIV will be applied to predict produce of other compounds.(4) Prediction of yields of Mannich bases based on Heuristic method (HM). Heuristic method is a novel way to predict compounds yield, which is utilized to construct the line prediction model of Mannich bases, leading to a correlation coefficient R and cross-validation RCV of 0.735 and 0.520 by 10 variables model. 2 abnormal compounds are eliminated after analysis, new model results are: R=0.860, RCV=0.770. 2 exterior sets are constructed to test the prediction ability and stability.①The exterior sets is used to test the model, datum of 42 compounds are obtained form references, 10 compounds which are from the above 42 compounds at random, are made of the test set, which R and RCV are 0.926, 0.820, and the results are favorable.②In the second predicted model, 5 Mannich bases are designed, then, which yields are predicted by the model constructed by 42 compounds above. afterwards, we synthesize them in the lab, and get experimented yields, which are compared with the predicted results. The results indicate the model are favorable, and has good prediction ability and stability. The results show that the novel heuristic method can be applied to predict the yields of compounds in the organic synthesis reaction, and HM is an excellent method with satisfactory estimation stability and favorable generalization.In the paper, molecular structure is described by 2D-QSAR skill, and the results of the models show that all ways are favorable to characterized molecular structure, which are used in the thesis. Predicting the yields of compounds provides a guide tool in the experiments of synthesis in reality.更多还原