【作者】 刘颖；

【导师】 倪永年；

【作者基本信息】 南昌大学 ， 应用化学， 2011， 硕士

【摘要】 作为中药质量控制的一种手段,中药指纹图谱旨在从整体上对中药进行研究,可以全面反映中药中所含的复杂成分的关系。通过将各种分析仪器和化学计量学相结合应用于中药指纹图谱研究中,为中药的质量控制提供了有力的评价手段和广阔的发展前景。本论文中利用高效液相色谱法和近红外光谱法这两种检测方法,分别建立了中成药感冒咳嗽颗粒、中药饮片白术和五味子的指纹图谱,并利用化学计量学方法对得到的指纹图谱进行分析。第一章：主要是结合文献对中药质量控制的意义和中药指纹图谱的研究现状及其发展前景进行了综述,详细叙述了中药产业发展的现状和存在的问题,对建立中药指纹图谱的分析手段进行了详细评述,并介绍了化学计量学方法在指纹图谱数据分析中的应用情况。第二章：采用高效液相色谱法对来自三个不同厂家的感冒咳嗽颗粒进行分析,建立其液相色谱指纹图谱,并结合组成感冒咳嗽颗粒的六味主要原药材的色谱图,利用相似度分析和主成分分析对样品的质量进行质量控制。结果发现来自不同厂家的样品其质量存在一定的差异,各图谱存在差别,通过主成分分析和相似度计算可以对不同厂家的样品进行区分。同时,结合原药材的图谱可以进一步地与成药样品谱图进行对比,通过参考相关文献从具体成分上的差异对成药样品进行研究。利用该方法可以对不同厂家的感冒咳嗽颗粒样品进行区分,为中成药样品的质量控制提供参考手段。第三章：以三类不同炮制方法的白术样品和白术的替代品苍术样品为研究对象,利用高效液相色谱法与二极管阵列检测器和荧光检测器联用的方法建立了中药白术的高效液相色谱二维指纹图谱。通过无监督模式识别方法主成分分析和系统聚类分析的结果发现,利用二维指纹图谱数据得到的分类效果要比单个检测器得到的指纹图谱数据更佳,可以获取更多的信息。利用PROMETHEE和GAIA多准则决策方法对样品进行排序和分类,发现不同类的样品之间存在差异,且各类白术样品按照炮制方法的程度有规则地进行排列。同样,利用有监督模式识别方法对不同数据建立模型进行预报发现,二维指纹图谱数据的识别率和预报率最佳,其中人工神经网络方法可以对训练样品和未知样品进行准确预报。再结合椭圆置信区间方法比较了两种有监督模式识别方法的模型预报结果,发现这两个模型结果均可靠满意。第四章：使用高效液相色谱-二极管阵列检测器建立了中药五味子的高效液相色谱三维指纹图谱。收集的五味子样品包括南五味子和北五味子样品,这两者外形极其相似但成分和药效存在差异。通过使用平行因子分析方法对五味子样品的三维指纹图谱数据进行解析可以得到分别与样品中主要组分浓度分布,组分的色谱流出曲线和组分的光谱图相关的三组数据。从浓度分布图及其基于组分浓度的主成分分析结果中可以看到两类五味子样品的成分间存在差异。使用三种有监督的模式识别方法建立五味子样品预报模型,可以将南五味子和北五味子样品进行正确分类。第五章：采用近红外光谱法对南五味子和北五味子样品进行分析,建立了五味子样品的近红外光谱指纹图谱。使用不同的前处理方法对原始光谱数据进行处理,并分别采用两种不同的变量选择方法用于对各种前处理数据进行有效波长选择。通过建立有监督模式识别模型对数据进行分析,结果发现使用前处理后的数据预报结果比原始数据更加可靠；使用有效波长选择得到的数据建立的模型预报结果优于全波长数据。该方法可为中药五味子质量控制提供一种快速的鉴别手段。更多还原

【Abstract】 As a means of quality control of traditional Chinese medicinc (TCM), TCM fingcrprinting aims to conduct a comprehensive research on TCM to fully reflect the complex relationship of the complicated components in medicines. High performance liquid chromatography (HPLC) and near infrared spectroscopy (NIR) combined with chemometrics methods are applied in this article to establish the TCM fingerprints.In the first chapter, the significance of quality control of TCMs and the development of TCMs fingerprinting are reviewed. The developing situation and existing problems of the TCM industry are described. Besides, the methods for establishing TCMs fingerprints and the applications of chemometrics in fingerprints are introduced.In the second chapter, HPLC was used to establish the fingerprints of a Chines patent medicine Ganmao Kesou Granules (GKG).11 GKG samples from three manufactures and six crude materials were analyzed by HPLC. Similarity analysis and principal component analysis (PCA) were conducted to evaluate the quality of GKG samples. It was found that there were differences in the quality of samples from difference manufacturers by comparing their chromatograms. By using PCA and similarity analysis it is able to discriminate them. Compared the crude materials’chromatograms with the patent medicines’, the differences of specified constituent could be obtained combining the reference literatures. The results showed that the proposed method could be used to discriminate GKG samples from difference manufacturers and be set as a reference method for the quality control of Chinese patent medicies.In the third chapter, three kinds of processed Rhizoma atractylodis macrocephalae (RAM) samples and RAM’s substitute Rhizhoma atractylodis (RA) were studied with HPLC-DAD-FLD. Two-dimensional fingerprints were esbalished by combing the UV fingerprints and fluorescence fingerprints, Unsupervised pattern recognition methods PCA and cluster analysis results showed that two-dimensional fingerprints provided better classification ability as compared with one dimensional fingerprints method. Multiple criteria decision making methods PROMETHEE and GAIA ranked and classified the samples, and the results reflected that the RAM samples ranked in order according to the processing degree. Supervised pattern recognition methods, K-nearest neighbors (KNN), partial least squares (PLS) and artifical neural networks (ANN), were used to build the prediction models with two dimensional and one dimensional fingerprints data. It is proved that two dimensional data sets exhibit better recognition and prediction abilities.In the fourth chapter, HPLC-DAD was used to establish the three-dimensional fingerprints of Wuweizi. Nan-wuweizi and Bei-wuweizi samples were collected, which have similar appearances but different in the constituents. Parallel factor analysis (PARAFAC) was used to resolve the three-way fingerprints, and three data sets that were related to the constituent content, the chromatographic elution profiles and spectral profiles of the constituents were obtained.K-nearest neighbors (KNN), partial least squares (PLS) and least squares - support vector machine (LS-SVM), were used to build the training and prediction models for Wuweizi samples, and performed well in the classification results.In the fifth chapter, NIR fingerprints of Nan-wuweizi and Bei-wuweizi samples were established. Different pre-processing methods were applied to pre-treat the original spectrums and effective wavelengths were selected with the use of variable selecting methods. The prediction results of training models showed that the pre-processed spectrums proved to be more reliable. And the models based on effective wavelengths performed better results than that of the whole spectrum.更多还原