CAS OpenIR  > 研究所(批量导入)
多组分同时分析与构效关系研究
石乐明
Subtype博士
Thesis Advisor许志宏
1991-06-30
Degree Grantor中国科学院研究生院
Abstract化学计量学(Chemometrics)是一门由数学、统计学以及计算机技术与化学相结合而产生的新型边缘科学。 它的迅速兴起为化学工作者更好地设计实验、获取数据、解析日益增长的大量化学数据提供了有力的手段和方法。本文针对化学计算学的两个主要研究领域,即“多组分同时分析”与“构效关系研究”,开展了一系列工作。本文第一篇介绍了作者在多组分同时分析领域所进行的工作。1、对卡尔曼滤 波方法在多组分同时分析中的应用进选取深入研究。利用紫外可见光谱进行多组分氨基酸及多种金属离子混合物的同进分析;将卡尔曼滤波方法在分析化学中的应用领域推广到分子荧光光谱研究,实现了多种染料的同时测定;提出了一种基于卡尔曼滤波方法的解析重叠双晶X-射线荧光光谱图的新方法,使得利用双晶X-射线荧光光谱进行元素的价态分析成为可能。2、将目标因子分析法与紫外光谱相结合,进行了多种氨基酸比芳香化合物的同时分析,可同时进行化合物的定性鉴别与定量分析。3、将偏最小二乘法与紫外光谱法相结合进行多组分有机物同时分析;借偏最小二乘方法与付立叶变换近红外漫反射光谱法实现了农产品中蛋白质的快速分析,与传统的化学方法相比,效率大为提高。4、利用卡尔曼滤波方法对滴定分析、化学平衡与色谱分析中重叠峰的解析进行了研究,拓广了卡尔曼滤波在化学中的应用领域。本文第二篇介绍了作者在构效关系研究方面所开展的工作。1、对镧系元素化合物结构参数(基态角量子数L)与性质之间的关系进行了研究。2、将对应因子分析法用于磺酰脲类除草剂的构效关系研究,基于化合物的结构参数信息实现其除草活性的分类与预测。3、将神经网络方法用于农药分子设计,对化学运河雄剂的构效关系与磺酰脲类除草剂的构效关系进行了研究,对影响神经网络预测能力的因素进行了探讨。对于这两个不同的体系,均获得了令人满意的分类与预测结果。利用神经网络对保幼激素的定量构效关系进行了初步研究,并与Hansch分析法的结果进行了比较。对神经网络的优越性进行了讨论,表明神经网络在农药分子筛选与设计中具有广阔的应用前景。
Other AbstractChemometrics is a newly developed interdisciplinary branch of chemistry, in which mathematical and statistical methods are used, (a) to design or select optimal measurement procedures and experiments; and (b) to provide maximum chemical information by analyzing chemical data. In this thesis, studies have been concentrated on the 'Simultaneous Multicomponent Analysis' and the 'Structure-Activity Relationships (SAR)', which are two of the most important chemometric topics. In Part One, the work on the 'Simultaneous Multicomponent Analysis', is described. 1. Numerous studies have been made on the application of Kalman filter (KF) algorithm in spectral multicomponent analysis. In ultra-violet and visible (UV/vis) spectrophotometry, the Kalman filter was used for the simultaneous determination of multicomponents, up to six amino acids and five inorganic metallic ions, respectively. The Kalman filter was extended for solving overlapped molecular fluorimetric spectra and used for the simultaneous determination of three dyestuffs. By using the Kalman filter algorithm, a new method for chemical valence analysis based on the resolution of overlapped two-crystal X-ray fluorescence spectra was proposed. This method was applied to the chemical valence analysis of sulfur. The accuracy and precision of the method was acceptable for multicomponent analysis. Detailed results were presented in this thesis. The characteristics of the Kalman filter for simultaneous multicomponent analysis were extensively examined. 2. Target factor analysis (TFA) method combined with UV spectrophotometry was successfully used for the simultaneous multicomponent analysis of amino acids and aromatic compounds. The relative error is less than 10%. It is able to use this method for qualitative identification and quantitative determination simultaneously. 3. Partial Least-Squares (PLS) method was used for the simultaneous multicomponent analysis of aromatic compounds. The PLS-based Fourier transform near infrared diffuse reflectance spectroscopy was proposed and used for the fast determination of protein content in agricultural products. The standard error of prediction is 0.279%. This method is much faster than the traditional Kjeldhl nitrogen determination method. 4. In fast or non-equilibrium titration analysis, the traditional methods were not able to find the correct end-point. The Kalman filter was also applied to determine the end-point in such titration systems and the relative error is less than 1.2%. In the resolution of overlapped chromatograms, the relative error of the Kalman filter is less than 7% for four different multicomponent systems. The Kalman filter and factor analysis methods combined with UV/vis spectrophotometry were proposed for estimating the chemical equilibrium constants of some analytical reagents. The results are in consistent with those reported in literature. Part Two of this thesis is on the 'Structure-Activity Relationship (SAR)' study. 1. The correlation analysis between some properties of the lanthanides and their total angular quantum number L at ground state was made with satisfactory results. 2. Correspondence factor analysis (CFA) method was applied to the SAR study of 70 compounds of Sulfonyl Urea Herbicides (SUH). Thirty of them were used as the training set. It was found that the first two principal factors were able to characterize more than 83% of the total variance of the data matrix. In the display diagram, all samples of class one were located in the area above the X axis, while all samples of class two under the X axis. The rate of correct classification of the 30 training samples is 100%. The rate of correct prediction of the 40 prediction samples is 77.5%. 3. Artificial Neural Network (ANN) was for the first time used for the SAR study of Chemical Hybridizing Agents (CHA) and SUH. For the CHA system, 47 samples were collected. In the classification test, the ANN method was able to classify all samples correctly. In the 'leave-n-out' experiment, the rate of correct prediction is 93.6% (for n = 10). For the SUH system, 97 samples were collected. Sample No. 8 was detected as an 'outlier' by the ANN method, so it was deleted from the sample set and the remaining 96 samples were used as the sample set. After 20,000 cycles of training, the rate of correct classification is 100%. The rate of correct prediction is 82.3% by using the 'leave-n-out' method (for n = 5). The factors affecting the prediction ability of the ANN method, including the numbers of hidden nodes and training cycles, were studied. The advantages of the ANN method over multivariate statistical methods for SAR study were discussed. The ANN method was also applied to the Quantitative Structure-Activity Relationship (QSAR) study of Juvenile Hormone Analog (JHA) pesticides. The results were compared with those obtained by the Hansch QSAR method. It is concluded that the ANN method may become a powerful tool for both SAR and QSAR study.
Pages108
Language中文
Document Type学位论文
Identifierhttp://ir.ipe.ac.cn/handle/122111/5814
Collection研究所(批量导入)
Recommended Citation
GB/T 7714
石乐明. 多组分同时分析与构效关系研究[D]. 中国科学院研究生院,1991.
Files in This Item:
File Name/Size DocType Version Access License
多组分同时分析与构效关系研究.pdf(6953KB) 限制开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[石乐明]'s Articles
Baidu academic
Similar articles in Baidu academic
[石乐明]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[石乐明]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.