Studies On Computational Analysis Methods For Biological Samples With Capillary Electrophoresis

Recently, chemometric techniques have played an increasingly important role in the field of analytical chemistry research. In this thesis, the application of chemometrics in capillary electrophoresis (CE) analysis of nucleosides is studied.Firstly, radial basis function network (RBFN) was applied to build the calibration model for the quantitative analysis of nucleoside (guanosione) in CE, which can take the nonlinear response of detector into account eventually. Compared with traditional linear regression (LR) and back-propagation artificial neural network (BP-ANN), the increase of accuracy in capillary electrophoresis could be achieved by RBFN approach. The results show that RBFN approach is easy to be used and can effectively improve the accuracy of quantitative analysis. RBFN might become a promising alternative to the existing calibration methods for the quantitative analysis in CE.Secondly, a method based on a powerful pattern recognition tool, Discriminant PLS (DPLS), was developed to identify the Cordyceps Sinensis from different producing areas. Ten kinds of nucleosides and other compounds in Cordyceps Sinensis samples from three areas in Qinghai province were analyzed by CE. DPLS was used to distinguish the producing areas of Cordyceps Sinensis samples. The results show that the combination of CE and DPLS is an accurate and effective tool to identify sources of Cordyceps Sinensis samples and could be widely used in quality control of Traditional Chinese Medicine (TCM).At last, as a recently developed classification method, support vector machine was used to diagnose breast cancer according to the levels of nucleosides in human urine. Using CE, eleven nucleosides in urine samples were analyzed from 17 healthy persons and 17 breast cancer patients. Then the urine samples were classified by SVM, and the results were better than that of BP-ANN. The combination of CE and SVM was expected to be an assisted tool for the clinical diagnosis of breast cancer.In summary, the application of chemometrics shows great potentialities in nucleoside analysis with CE.