【作者】 宋岩；

【导师】 赵冰；

【作者基本信息】 吉林大学 ， 物理化学， 2009， 博士

【摘要】 本文研究了近红外光谱分析技术在药物成分分析、中草药定性定量检测以及制药过程监控方面的应用。着重对盐酸左氧氟沙星、甲磺酸培氟沙星、盐酸西替利嗪、淫羊藿等4种药物进行了近红外光谱图信息提取和定量分析工作,探讨了不同预处理方法对模型的影响,介绍了近红外光谱技术在肺宁颗粒在线检测过程中的应用。对盐酸左氧氟沙星进行近红外光谱的检测,比较了PLS法和ANN法对盐酸左氧氟沙星针剂建立的定量模型。仿真实验表明,PLS法显示了更加优越的性能,应用此法可直接对大量未知样品进行测定。对甲磺酸培氟沙星进行近红外光谱的检测,利用PLS法对甲磺酸培氟沙星建立了定量模型。仿真实验表明,PLS法建立的模型预测结果准确。利用PCR法对甲磺酸培氟沙星进行了定量建模,比较了三种不同预处理方法,结果显示经过归一化处理后的模型效果最好。将近红外漫反射技术应用到盐酸西替利嗪的定量检测中,详细比较了PLS法和PCR法对模型的影响。经比较,利用PLS方法来处理盐酸西替利嗪片剂表现出更加优越的性能,样品的化学测定值和模型预测值的相关性更好。将近红外光谱技术与化学计量学方法联用,对淫羊藿进行了产地鉴别,同时实现对其主成分的定量检测。结果表明此方法可简化测试步骤,缩短检测时间,节约测试成本,为近红外光谱技术应用于淫羊藿等中草药的定性、定量检测提供了理论依据。介绍了近红外光谱技术在中药生产过程控制中的应用,概述了肺宁颗粒的制备过程和在线检测系统,分析了近红外光谱在线检测技术的前景及存在的问题。更多还原

【Abstract】 The active ingredient (AI) of a drug is not only its essential component, but also the most important determinant of the drug potency. The quality and quantity of the AI in drugs are directly related to their medicinal properties. Therefore, the quality and quantity of AI in drugs are very important markers of their effectiveness, which must be put under strictly control during the manufacture and sale process of drugs. Instead of analyzing the final drug products statically, the analysts must also see to the detailed qualities of the raw material, the intermediate products, the semi-finished products and the final products, in order to realize a real-time, on-site and in-line quality control. In this study, we developed a new method for quick, simple and accurate analysis of some drugs.Near infrared (NIR) normally refer to the electromagnetic wave which has a wavelength between 780nm to 2526nm. Generally, the near infrared region can be further sub-divided into two regions: the 780nm– 1100nm region, which is called short-wavelength near infrared, and the 1100nm– 2526nm region, which is called long-wavelength near infrared. Spectroscopic signals in near infrared region are generated by the single, double and triple bond signals of the absorption frequencies of O-H, C-H and N-H chemical bonds. Since this region is crowded with the double and triple bonds signals, and with severe overlap of signals, it is generally to be considered as a weak informative region, which means it can not be analyzed in a routine manner. Modern near infrared spectroscopic technology however, combines the power of chemometrics and infrared spectroscopy. It makes use of the power of quick and simple data collection of IR spectroscopy and the information extraction and processing power of chemometrics. The key technical point of recent near infrared technology is to set up the multi-factor correction model of infrared data. With the development of the computer science and applied mathematics and their applications in analytical chemistry, NIR spectroscopy and chemometrics are getting more and more attention since 1980s. It has been introduced into variable research fields and has been developed into an independent analytical technique, and changing the old idea that the theoretical study are lagging behind the practical experience in analytical chemistry. In present study, we used modern NIR spectroscopy to analyze the active ingredient of drugs. Modern NIR spectroscopy uses the transmitted and reflective spectra to analyze the chemical contents and the molecular structures of certain materials. It has several advantages, such as no contamination to the samples, no contact detection, non-destructive, easy to handle and easy to be integrated into the production line. Any substance would generate absorption signal in NIR region, which contains plenty of structural and molecular information. Thus the NIR spectroscopy can be used into multiple fields such as agriculture, medicine and health science. For instance, in the field of medicinal analytical study, NIR spectroscopy has been applied to analyze not only different dosage forms such as raw materials, tablet, capsule and liquid, but also different types of drugs, such as, protein, herbal medicine and antibiotics. NIR is even more useful to analyze the purity of the raw drug materials, packaging materials, and to monitor the production technique; different optical fiber probes can be utilized to analyze and control the production process. The key point of the modern NIR spectroscopy is how to extract useful information from the spectra.In present study we emphasize the method of extracting useful information from the NIR spectra during the analysis of drugs. We use the spectra of several synthetic and herbal medicines as examples to investigate the data processing of NIR spectra using wavelet transform, partial least square method (PLS), artificial neural network (ANN) and other chemometric methods.We use NIR spectroscopic methods to perform quantitative analysis for several synthetic and herbal medicines and compare the results with conventional methods accordingly, thus to prove the effectiveness and accuracy of the NIR spectroscopic analysis method.1.We used Levofloxacin Hemihydrate as the first target of our study. We use NIR combined with PLS or ANN methods to rapidly determine its amount. We set up the quantitative analysis models for PLS or ANN methods, compared the modeling process, the optimization of the parameters and the accuracy of the results for these two methods, and we discussed the optimized methods for setting up the important parameters in these two models in detail. The R2 and RMSEP for PLS model were 0.964 and 0.2428, respectively. Whereas for the ANN model, the R2 and RMSEP after wavelet transform was 0.944 and 0.5722, respectively. Virtual experiment suggested that, both of these two methods result in accurate models with can give good prediction. However, the PLS model is a little bit better than ANN model, indicated by its less error and deviation. The prediction of PLS model also has better correlation with the results of chemical assay.2.NIR spectroscopy was also used to analyze the pefloxacin mesilate (PM) injection solution. The recorded NIR spectra were analyzed and the quantitative model was generated by partial least square method and principle component regression (PCR) method. The R2 value, RMSEP and RPD of the PLS model, generated from several different batches of PM injection solution, were 0.959, 0.00087 and 4.95. The virtual experiment based on the different models suggested that the model generated using partial least square method gave us accurate prediction. The performance of this model was good, which showed small mean square deviation and mean error. For the model generated by PCR method, three different preliminary data processing methods, e.g. smoothing, normalization and multivariate scatter-correction, were used and the results were compared in detail. The normalization gave us the best results.3.Similar approach was used to perform the quantitative analysis for Cetirizine Hydrochloride samples. PLS and PCR methods were used to generate the quantitative models. The modeling process, optimization of the parameters and the accuracy of the results using these two methods were compared. PLS method with cross-validation, first-order derivative and vector normalization data processing, was used to generate the quantitative analysis model. The resulted model had aR2 value of 0.878, the RMSECV value was 5.49E-5 and RPD value was 2.55. For the model generated by PCR method, different data processing methods, i.e. smoothing, normalization, were used and compared, multivariate scatter-correction, noise suppression, derivative, baseline correction and selection of variables, were used and compared. It was found that the normalization method gave us the best results, which has a correlation coefficient of 0.81, root mean square deviation of 0.000451 and mean residual error of -1.32E-05. Compare the result of these two methods, both PLS model and PCR model shows good generalization capability and accuracy of prediction. However, the PLS methods shows a little bit better performance while dealing with the Cetirizine Hydrochloride tablet samples ,with considerable less relative error and root mean square deviation than PCR methods, its prediction also has a better correlation coefficient with the experimental assay results. These proved that the PLS could be used to handle large amount of samples. 4.NIR spectroscopy and chemometrics were also used to identify the origins of Aceranthus sagittatus samples and to quantitatively analyze their active ingredient content. Nine different samples collected from different areas were analyzed, using vector normalization and/or first–order derivative processing methods. The results suggest that the vector normalization data processing gave us better qualitative identification model, the Cluster analysis gave us satisfactory results. The classification error rate was 0. It can effectively identify the different origins of the samples thus this resulted qualitative analysis model can be used to quickly identify the origin of unknown samples. This analysis method is fast, simple and low-cost, which provide us a novel way for the qualitative identification of herbal medicine. The active ingredient of these Aceranthus sagittatus samples were determined using HPLC method, the results were used as the training data to generate the NIR quantitative analysis model using PLS methods. After multiple optimized preliminary data processing procedure, the wavelength range for the further study was chosen as 7502.1～4597.7 cm-1. 42 different Aceranthus sagittatus samples were used as the calibration data set. After multivariate scatter-correction, the model gave us satisfactory results. The optimized model was used to predict the active ingredient of 9 random Aceranthus sagittatus samples. The results indicated that the standard deviation of the prediction was 0.0206, which means that it has very good accuracy of prediction and very effective.5.In present study, we also discussed some of the application of modern NIR spectroscopic technology in the in-line control of herbal medicine production. We reviewed the manufacture process of the product“fei ning ke li”and its in-line production control system, especially the effect of raw material pretreatment system, the structure and analysis model of the NIR spectrometer in this system and how this system has been established. We also discussed the perspective and potential problems of the application of NIR technique in in-line diction system.This study provided a deep insight into the information extraction methods and quantitative analysis of NIR spectra, and compared the effects of different chemometric methods for the data processing. It also acquired valuable spectroscopic character of these several drugs in near infrared region. , which will help to establish the spectroscopic database for the analysis of these drugs, and provided a solid foundation for the further development of the in-line production control system. The experimental data also indicated that this proposed NIR data analysis method increased the accuracy of the analysis, which has a high practical value. This proposed quantitative analysis method was quick and simple, low-cost, without the requirement of complicated sample pretreatment. It can be easily applied to analyze large amount of sample. This study provided an effective method for pharmaceutical analysis, and also provided an convincing evidence to support the application of NIR spectroscopy in the analysis of drugs.更多还原