【作者】 董文江；

【导师】 倪永年；

【作者基本信息】 南昌大学 ， 食品科学与工程， 2013， 博士

【摘要】 近年来,指纹图谱技术和化学计量学方法被广泛应用到食品和中药的质量控制中,这些技术能够较好的解决定性和定量控制食品和中药的内在质量,有效地保证食品的品质和中药的药效这一复杂问题。指纹图谱技术具有全面性、整体性、层次性、关联性和模糊性等特点,能够尽可能全面地获得食品和中药的化学组分信息,建立快速而有效的食品和中药安全质量控制体系,对国民经济的发展作出贡献。本论文以几种常见的食品和中药：山楂、薄荷、白前及其混伪品(白薇和徐长卿)和紫苏叶为研究对象,利用近红外漫反射光谱(NIRS)、高效液相色谱-二极管阵列检测(HPLC-DAD)、超高效液相色谱四级杆飞行时间串联质谱(UPLC-Q-TOF-MS)和顶空气相色谱质谱联用(HS-GC-MS)采集样品的波谱数据并建立可靠的指纹图谱,采用标准化学方法分别测定了山楂和薄荷的主要化学成分和抗氧化活性,最后结合多种化学计量学方法对所研究的体系进行评估和研究,旨在为上述试验材料的质量控制和评价提供一种比较新型的方法。主要研究内容及结果如下：1.建立了一种可定量测定中药及其混伪品含量的近红外光谱分析方法,采用的多元校正技术有偏最小二乘回归(PLSR)和径向基函数人工神经网络(RBF-ANN)。本研究采集了中药白前、白薇、徐长卿、白前中混有一种或两种伪品的近红外漫反射光谱数据,波长范围为800-2500nm,包括三个样品集,白前中混有白薇,白前中混有徐长卿,白前中同时混有白薇和徐长卿。PLSR和RBF-ANN模型均得到了令人满意的结果,即使掺假水平的质量比为5%时,但是RBF-ANN对上述三种样品预报的预报误差均方根(RMSEP)的结果更好。本研究表明利用近红外漫反射光谱结合多元数据分析方法(如PLSR或RBF-ANN)定量分析中药中混有一种或两种伪品含量是可行的。2.建立了近红外光谱校正模型用于区分来自三个不同产地的山楂,同时测定其中的总糖、总酸、总酚和总抗氧化活性。主成分分析(PCA)被用于不同地理来源水果的区分,三种模式识别方法：线性判别分析(LDA)、偏最小二乘判别分析(PLS-DA)和反传人工神经网络(BP-ANN)被用于分类并且比较了它们的预报结果。此外,基于一阶导数处理后的近红外光谱数据的三种多元校正模型：偏最小二乘回归(PLSR)、反传人工神经网络(BP-ANN)和最小二乘支持向量机(LS-SVM)被用于定量分析四个化学指标,即总糖、总酸、总酚和总抗氧化活性,同时利用预报集样品检验模型的可靠性。3.采用顶空气相色谱质谱联用(HS-GC-MS)、高效液相色谱-二极管阵列检测(HPLC-DAD)和超高效液相色谱四级杆飞行时间串联质谱(UPLC-Q-TOF-MS)结合多元数据分析研究了来自不同地理来源的薄荷样品。HS-GC-MS和HPLC-DAD数据集的主成分分析(PCA)和多准则决策法(MCDM) PROMETHEE#GAIA的得分投影图表明来自三个不同地区的样品形成明显的分界,反传人工神经网络(BP-ANN)和偏最小二乘判别分析(PLS-DA)模型的预报能力令人满意。PLS-DA的权重回归系数得到对于分类贡献较大的化合物主要有：chlorogenic acid, unknown3, kaempherol7-O-rutinoside, salvianolic acid L, hesperidin, diosmetin, unknown6和pebrellin。结果表明HS-GC-MS和UPLC-Q-TOF-MS结合多元数据分析可被用于植物性材料如薄荷的地理来源区分,同时鉴别对于分类起主要作用的化合物。4.建立了一种可同时分析薄荷中多种化学成分和生物活性的新型的近红外光谱方法,如总多糖、总黄酮、总多酚和总抗氧化活性。为了分辩上述组分的近红外光谱矩阵,最小二乘支持向量机(LS-SVM)被证明是最适合组分预测的化学计量学方法,尽管结果稍优于径向基函数偏最小二乘法(RBF-PLS),而偏最小二乘回归(PLSR)的结果不是很好。此外,主成分分析(PCA)和系统聚类分析(HCA)能够区分四个不同地区的样品,对三种分类方法：K-最近邻法(KNN)、线性判别分析(LDA)和偏最小二乘判别分析(PLS-DA)的预报结果进行了比较。总的来说,此方法不仅节省了分析时间而且减少了传统方法所需的试剂消耗,可进一步验证以取代传统的湿化学分析方法。5.利用高效液相色谱全轮廓图指纹图谱结合化学计量学建立了一种可用于区分不同栽培地区紫苏叶样品的方法,经过自适应迭代再加权最小二乘法(airPLS)和相关优化翘曲(COW)校正后,基线漂移和保留时间漂移现象得到了明显的改善,全轮廓色谱图数据为输入变量时,PCA结果表明不同来源的样品能够按特性各自聚为一类；分段间隔压缩变量后数据的PCA得到与全轮廓色谱图数据为输入变量时相当的结果,PLS-DA模型对于分类的识别能力和预报能力分别为92.8%和89.6%。更多还原

【Abstract】 Fingerprinting technology combined with chemometrics have been widely applied in the quality control of food and traditional Chinese medicine (TCM) products, which can effectively solve the problem of the intrisic quality control of food and TCM, and ensure the attributes of foods and pharmaceutical effects of TCM. Fingerprinting technology has the characterics of comprehensiveness, integrity, hierarchy, relevance, and fuzziness, the chemical information of food and TCM can be obtained comprehensively using this method. Therefore, it is important and necessary to construct a rapid and effective food and TCM quality control system for people health. In this study, some common food and TCM samples, such as Cynanchum stauntonii (CS) and its adulterants(Cyannchum atrati (CA) and Cynanchum paniculati (CP)); Chinese hawthorn fruit; Mint (Mentha haplocalyx Briq.); Perilla frutescens (L.) Britt, were used as samples in this study. The analytical methods included near-infrared spectroscopy (NIRS), high performance liquid chromatography-diode array detector (HPLC-DAD), ultra high performance liquid chromatography-quadrupole-time-of-flight-mass spectrometry (UPLC-Q-TOF-MS), and head space-gas chromatography-mass spectrometry (HS-GC-MS), several chemical components and antioxidant activity in Chinese hawthorn and Mint were determined using standard analytical methods, incorporating with several chemometric methods for data analysis, aiming at providing novel methods for the quality control and evaluation of the above-mentioned materials. The main conclusions were as follows:1. A rapid near-infrared spectroscopy (NIRS) analytical method which was supported by multi-variate calibration, e.g. partial least squares regression (PLSR) and radial basis function artificial neural networks (RBF-ANN) was developed, in order to quantify the TCM and the adulterants. In this work, Cynanchum stauntonii (CS), a commonly used TCM, in mixtures with one or two adulterants-two morphological types of TCM, Cynanchum atrati (CA) and Cynanchum paniculati (CP), were determined using NIR reflectance spectroscopy. The three sample sets, CS adulterated with CA or CP, and CS with both CA and CP, were measured in the range of800-2500nm. Both PLSR and RBF-ANN calibration models provided satisfactory results, even at an adulteration level of5%(w/w), but the RBF-ANN models with better root mean square error of prediction (RMSEP) values for CS, CA, and CP arguably performed better. Consequently, this work demonstrates that the NIR method of sampling complex mixtures of similar substances such as CS adulterated by CA and/or CP is capable of producing data suitable for the quantitative analysis of mixtures consisting of the original TCM adulterated by one or two similar substances, provided the spectral data are interrogated by multi-variate methods of data analysis such as PLSR or RBF-ANN.2. Near-infrared spectroscopy (NIRS) calibrations were developed for the discrimination of Chinese hawthorn(Crataegus pinnatifida Bge. var. major) frruit from three geographical regions as well as for the estimation of the total sugar, total acid, total phenolic content, and total antioxidant activity. Principal component analysis (PCA) was used for the discrimination of the fruit on the basis of their geographical origin. Three pattern recognition methods:linear discriminant analysis, partial least squares-discriminant analysis, and back-propagation artificial neural networks were applied to classify and compare these samples. Furthermore, three multivariate calibration models based on the first derivative NIR spectroscopy, partial least squares regression, back propagation artificial neural networks and least squares-support vector machines (LS-SVM), were constructed for quantitative analysis of the four analytes, total sugar, total acid, total phenolic content, and total antioxidant activity and validated by prediction data sets.3. Mint (Mentha haplocalyx Briq.) obtained from different geographical regions was characterized using head space-gas chromatography-mass spectrometry (HS-GC-MS) and ultra high performance liquid chromatography-quadrupole-time-of-flight-mass spectrometry (UPLC-Q-TOF-MS) and followed multivariate data analyses. Principal component analysis (PCA), and the rank-ordering multi-criteria decision making (MCDM) PROMETHEE and GAIA score plots from HS-GC-MS and HPLC-DAD data sets showed a clear distinction among mint from three different regions in China. Classification results showed that satisfactory performance of the prediction ability for back propagation-artificial neural networks (BP-ANN) and partial least suqares-discriminant analysis (PLS-DA). The major compounds that contributed to the discrimination were chlorogenic acid, unknown3, kaempherol7-O-rutinoside, salvianolic acid L, hesperidin, diosmein, unknown6, and pebrellin in mint on the basis of regression coefficients of PLS-DA model. The results indicated that HS-GC-MS and UPLC-Q-TOF-MS fingerprints in combination with chemometric analyses can be used to distingusih the grographical origins of plants and identify compounds were responsible for discrimination.4. A novel near-infrared spectroscopy (NIRS) has been researched and developed for the simultaneous analyses of the chemical components and associated properties of Mint (Mentha haplocalyx Briq.) samples, which are commonly used for tea brewing, namely, total polysaccharide content (TPSC), total flavonoid content (TFC), total phenolic content (TPC), and total antioxidant activity (TAA). To resolve the NIRS data matrix for such analyses, LS-SVM was found to be the best chemometrics method for prediction although it was closely followed the RBF-PLS model; notably, the commonly used PLS was unsatisfactory in this case. Additionally, principal component analysis (PCA) and hierarchical cluster analysis (HCA), were able to distinguish the Mint samples according to their four provinces of origin; this was further facilitated with the use of the chemometrics classification methods-K-nearest neighbors (KNN), linear discriminant analysis (LDA), and partial least squares-discriminant analysis (PLS-DA). In general, given the potential savings on sampling and analysis time as well as on the costs of special analytical reagents required for the standard individual methods, NIRS offers a very attractive alternative for the simultaneous analysis of Mint samples.5. Perilla frutescens (L.) Britt samples from different geographical origins in China were distinguished using liquid chromatography incorporated with chemometrics method, a total of84samples were used for data analysis. The chromatographic fingerprints demonstrated the chemical compositons of the samples, enabling the discrimination of samples from different origins is possible. Prior to classification, the datasets were subjected to pretreatment including baseline correction and retention time alignment. Principal component analysis (PCA) was performed to the aligned and compressed data matrix, respectively, to investigate the data distribution and evaluate the data quality, seperation among the three sets of samples was observed in the PCA score plots. Partial least squares-discriminant analysis (PLS-DA) provided the recognition ability and prediction ability of92.8%and89.6%, respectively, and a relative satisfactory results were obtained in this study.更多还原