Quality Analysis For Some Food And Chinese Herbal Medicine With The Use Of Chromatography And Near Infrared Spectroscopy With The Aid Of Chemometrics

Fingerprinting techniques have been widely used in the quality control of food and Chinese herbal medicine, the quality control including the determining of origins, active components and adulterations of food and traditional. Fingerprinting techniques had the features of otherness, fuzziness, and integrity. Every food and traditional Chinese herb had its unique chemical fingerprint, and their fingerprints could comprehensively reflect their chemical characteristics. The similarity and difference between the samples could be revealed just by the signal response of their fingerprints so as to achieve quality control and quality evaluation. In this study, high performance liquid chromatography(HPLC) and gas chromatography –mass spectrometry(GC-MS) fingerprints of Flos chrysanthemum were built, and near-infrared reflectance spectroscopy fingerprints of dehydrated tomato and purple sweet potato were built, then combined with chemometrics to extract the information of fingerprints data and process the data, so that to obtain quality control and quality evaluation of Flos chrysanthemum, dehydrated tomato and purple sweet potato. The main contents of this study were as follows:1. A rapid and non-destructive near-infrared(NIR) spectroscopy technique combined with chemometric was built to discriminated dehydrated tomatoes from different producers. In this research, NIR spectra of tomatoes from four different producers were collected, and then NIR spectra were pretreated by multiplicative scatter correction(MSC). Take the whole spectra and the characteristic wavelength variables selected by genetic algorithm-partial least square(GA-PLS) as input data to do principal component analysis(PCA) respectively, the results demonstrated that full spectrum contained noise and other disadvantageous spectrum region which will be against for discriminating dehydrated tomato samples, and GA-PLS combining the genetic algorithm with the partial least squares method had a strong powder to select the efficient regions that provided good PCA clustering result. Establish an NIRS multivariate calibration model with the use of partial least squares regression(PLSR) and radial basis function-neural network analysis(RBF-NN) methods for quantitative analysis of lycopene and several other key parameters in dehydrated tomatoes- total acid, sugar, phenols and antioxidant activity. Compare the statistical results of both prediction models, it could be found that both of the two calibration models had the potential to predict lycopene and total acid, sugar, phenolics and antioxidant activity in the dehydrated tomatoes, and RBF-NN method was satisfactory for the prediction of quality parameters and was better than the more simple calibration models, such as PLSR.2. A rapid and non-destructive Near-infrared(NIR) spectroscopy technique combined with chemometric was built to discriminated powdered, pure purple sweet potato samples, pure white sweet potato samples and adulterated sweet potato samples. In this research, NIR spectra of purple sweet potato powder with different adulteration concentrations(0%, 20%, 40%, 60%, 80% and 100%) were collected, and then NIR spectra were pretreated by standard normal variate(SNV). Take the characteristic wavelength variables selected by GA-PLS and successive projections algorithm(SPA) as input data to build k?nearest neighbors(KNN) and linear discriminant analysis(LDA) models for prediction the adulteration concentrations of purple sweet potato samples respectively, the results demonstrated that SPA was more effective for selecting characteristic wavelength variables. And NIR combined with chemometric could be used to differentiate purple sweet potato samples with different adulteration concentrations, and to quantitatively predict their adulteration concentrations. Establish radial basis function?partial least squares(RBF?PLS) methods for quantitative analysis of the total anthocyanin(TA) and total antioxidant activity(TAA) obtained from the purple sweet potato samples. RBF?PLS method could be accurately and effectively used for predicting the TA and TAA of the purple sweet potato samples.3. Two kinds of fingerprints, full spectrum chromatography fingerprint and two-dimensional HPLC- GC-MS fingerprint were built and combined with chemometric to differentiate different Flos chrysanthemum cultivars and predict the unknown samples. In the first fingerprinting technique, correlation optimized warping(COW) and adaptive iteratively reweighted penalized least squares(airPLS) methods were applied to correct retention time shifts and baseline drifts of the HPLC fingerprint. PCA was used to discriminate the Flos chrysanthemum cultivars, then partial least squares(PLS), continuum power regression(CPR), and K-nearest neighbors(KNN) were used to predict the unknown samples. The elliptic joint confidence region(EJCR) method was used to evaluate the prediction ability of the models. In the second fingerprinting technique, a two-dimensional chromatographic fingerprint dataset of the four kinds of Flos chrysanthemum cultivar was built based on HPLC and GC-MS common peaks. PCA was used to extracte characteristic variables from the two-dimensional fingerprints, then take the characteristic variables as input data to do PCA and kernel principal component analysis(KPCA) analysis, both of the two PCA methods could be used to discriminate and classify different Flos chrysanthemum cultivars. Finaly, four supervised pattern recognition approaches: RBF-NN, least square support vector machines(LS-SVM), linear discriminant analysis(LDA) and KNN were used to predict the unknown samples.