Chemometric Methods Applied Research, In The Qing Kai Ling Injection Nationals Wine Spectroscopy

This thesis describes chemometric methods and their use in Ultraviolet (UV) and Near Infrared spectroscopy (NIR). The particular applications in this study include the UV and NIR analysises of qingkailing injection and Guogongjiu medicinal wine using ultraviolet and transmittance near infrared spectrophotometers. The calibration set was constructed from using the spectra and HPLC data, which was used to predict the content of unknown samples. Different calibration techniques, including Partial Least Squares Regression (PLS), Artificial Neural Network (ANN), Support Vector Machines (SVM) and Least Squares Support Vector Machines (LS-SVM) were investigated and applicability for pharmaceutical analyses was discussed. In addition, the techniques for calibration subset selection and mathematical sample pre-treatments were explored.In Chapter 1, a brief description of chemometric methods and NIR spectrometry was given. Recent progresses for both the theoretic aspect and practical applications were reviewed.In Chapter 2, Qingkailing injections intermediate was studied by SVM and UV spectrometry. The first derivative and wavelet compression methods were used to eliminate the slope-background and reduce variables for the measured UV spectra of Qingkailing injections intermediate. Then, SVM is used for building the classification model to discriminate qualified and unqualified samples. The effects of spectral preprocessing and model parameters were investigated. Under optimized conditions,correct classifications of 100%, 95.4%, 97.3%, and 100% were obtained for the four batches of the intermediate of Qingkailing injection samples,respectively. 97.3% of the intermediate samples were correctly classified for the mixture samples of four batches together. Results showed that SVM combination to UV spectrometry can be a useful means for quality control of Chinese medicinal injections owing to its good accuracy and good generalization.In Chapter 3, the first derivative spectra with selected wavelengths were used to eliminate the slope-background and reduce variables for the measured UV spectra of Qingkailing injections intermediates. Then, LS-SVM was used for building the classification model to discriminate 196 injections intermediate samples. The modeling parameters were optimized by using two-grid searching and ten-fold cross-validation methods. Under the optimized conditions, the predictive ability of the testing set and the area under Receiver Operating Characteristic curve curves (AUR) can reach 98.0% and 0.983, respectively. Comparing with the conventional SVM, LS-SVM was proved to give better accuracy and generalization for the classification problem investigated. Results showed that LS-SVM technique can be a useful means for quality control of Chinese medicinal injections in the production process and other Chinese medicines.In Chapter 4, NIR spectrometry combined with chemometric methods is used for establishing a new method to determine the content of total nitrogen and geniposide in Qingkailing injections. Firstly training set and testing set are partitioned by Kernard-Stone algorithm, synergy interval Partial Least Squares (siPLS)) is used for selecting effective spectral regions and building quantitative calibration models of total nitrogen and geniposide. Spectral retreating methods are discussed in detail. Correlation coefficient is 0.999, Root Mean Square Error of Cross-Validation (RMSECV) is 0.023, Root Mean Square Error of Prediction (RMSEP) is 0.074 for total nitrogen; Correlation coefficient is 0.708, RMSECV is 0.023, RMSEP is 0.159 for geniposide. The predictive results show that the method proposed is rapid, non-destructive and credible, which can be applied to online control the quality of Chinese medicine injection in industry scenes extensively.In Chapter 5, NIR spectrometry combined with LS-SVM is used for establishing a new method to determine the hesperidin content of guogongjiu medicinal wine. Training set is partitioned by Kernard-Stone algorithm. Spectral pretreating methods are discussed in detail. Smoothing, first derivative and range-scaling were used for the pretreating methods of the NIR spectra of guogongjiu medicinal wine. Then, the effective interval is selected for 8211～8312 cm-1 and 9712～9808 cm-1 by siPLS, The model is established by LS-SVM, RMSECV is 0.001, and RMSEP is 0.004. Comparing with siPLS, Radial Basis Function Neural Network (RBF-NN), and SVM, The result shows that the method is rapid, non-destructive, and credible. It is an effective measurement for determining the hesperidin content of guogongjiu medicinal wine.