| The scientific analysis of chemical data sets is to discover the hidden relationships between the chemical properties and their constitutions. The recent development of high-dimensional chemometric resolution methods provides a more powerful tool for us to explore the complex chemical systems as opposed to conventional analytical methods. The primary purpose of this work is to develop new methodology in herbal fingerprinting research and spectral analysis and to test the adequacy of chemometric resolution techniques and multivariate analysis in areas related to the chemical sciences. Hence, a fair part of the thesis describes methodological developments related to herbal fingerprinting and chemometric methods, as well as their application to chemical systems. In all, this thesis covers four aspects as below.1. The similarity algorithm study on fingerprint of traditional Chinese medicine (chapter 3 to chapter 4). A new method based on fuzzy mathematics and information theory is developed to solve the similarity comparison of fingerprint. The relative entropy has become an excellent method to measure the difference of two probability distributions; the method has been applied to process the question of pattern recognition, etc. Based on it, divergence is proposed to measure difference of distribution. In this paper, the principle of determining similarity of chromatographic fingerprint of Chinese herbal medicine is calculated by using of the good character of divergence.Partial least-squares (PLS) is developed to solve the similarity comparison of fingerprint too. The PLS was used for extracting the most useful information from the chromatograph or spectrum data of Traditional Chinese medicine, in order to predict a response variable. In this paper, the principle of determining similarity of chromatographic fingerprint of Chinese herbal medicine is calculated by using of PLS.A better result is obtained by compared with conventional method and the two methods have less computation quantity than other algorithms.2. Analysis of the catalytic process of enzyme by using heuristic evolving latent projection (HELP) (chapter 5). A complex bio-chemistry reaction process was studied by sectional moving window factor analysis (SMWFA) - heuristic evolving latent projection (HELP) methods, the process information was resolved to concentration and spectrum profiles. Compared with multivariate curve resolution (MCR-ALS) method, the reasonable mechanism was obtained.3. Resolution and quantitative analysis of overlapped HPLC signals by three-way PARAFAC model (chapter 6). Alternating Least Squares (ALS) technique is applied to PARAFAC in order to add some constraints in the spectrum and chromatograph model to optimize the parameter estimates from the PARAFAC model. An example of HPLC-DAD data is used to show the applicability of the method.4. Analysis of the traditional Chinese medicine with hyphenated chromatography and chemometric resolution (chapter 7 to chapter 8). Multivariate curve resolution (MCR) - alternating least squares (ALS) and sectional moving window factor analysis (SMWFA) - heuristic evolving latent projection (HELP), are used to resolve the overlapped two-dimensional chromatogram-spectrum data of the Traditional Chinese Medicine into estimated pure component chromatograms and spectrum profiles. In this paper, the two methods are used to resolve a heavily overlapped components cluster of the compound Haqing injection. On the basis of comparison of the two methods, an improved deducted order of magnitude of the fixed size moving window evolving factor analysis (DOMFSMWFA) procedure is proposed and applied in the resolution of the compound Haqing injection.Peak shift correction (PSC) is applied to HPLC signals of traditional Chinese medicine in order to strengthen the trilinearity of the data, which is particularly useful in analyzing three-way data. The usefulness of PSC on PARAFAC solutions is examined by means of an example of HPLC-DAD data of the traditional Chinese medicine Yanhusuo.
|
PDF Full Text Request