红外光谱在药物分析中的应用研究

【作者】 刘小丽；

【导师】 李华；

【作者基本信息】 西北大学 ， 中药学， 2013， 博士

【摘要】 本论文的研究主题为用近红外(NIR)光谱法结合化学计量学方法对药物组成成分进行定量或定性分析；以及中红外光谱结合二阶导数、自退卷积和曲线拟合方法进行药物对蛋白二级结构影响的研究,并取得令人满意的研究结果。其具体研究内容如下：1.近红外光谱法用于分析三种对映异构体的组成借助牛血清白蛋白(BSA)与一对对映异构体的特异结合作用,采用近红外光谱法结合偏最小二乘回归(PLS)算法分别对组氨酸、薄荷醇和色氨酸对映异构体混合液中D-异构体的组成含量进行定量分析,并与主成分回归模型和多元线性回归模型进行了比较。对数据预处理方法、建模波段范围和模型参数进行了优化,结果表明近红外光谱结合PLS可成功的用于三种对映异构体中D-异构体的含量预测；但三种对映异构体的最佳模型的准确度并不相同,其中对组氨酸的预测结果最好,依次为薄荷醇与色氨酸,这可能与BSA与对映异构体的特异结合程度有关。2.近红外光谱法同时分析槐米中总黄酮和多糖含量采用近红外光谱法结合支持向量回归(SVR)算法对槐米中总黄酮和多糖的同时测定进行研究,并与PLS模型进行比较。对建模波段和数据预处理方法进行了优化,采用径向基函数为核函数,利用网格寻优算法计算支持向量回归模型的最优参数。结果表明SVR模型对槐米中总黄酮含量预测性能明显好于PLS模型；对多糖含量的预测,SVR模型性能也略好于PLS模型。对两大类物质,两种模型对总黄酮的预测结果都好于对多糖的预测结果。3.近红外光谱技术定量分析尼可地尔粉末样品采用近红外光谱法结合SVR算法对尼可地尔粉末样品中尼可地尔含量测定进行研究。利用主成分分析和独立成分分析两种信息压缩算法来改善SVR模型的预测性能,结果表明两种算法对SVR的预测性能均有改进,并且建模速度明显加快；其中独立成分分析对SVR性能改善效果较明显,并使其预测性能优于PLS回归模型。4.近红外光谱技术用于板蓝根颗粒厂家鉴别的研究利用聚类分析和支持向量分类(SVC)两种方法结合近红外光谱数据,对三个不同生产厂家的板蓝根颗粒进行分类鉴别。两种方法采用适当的数据预处理和参数设置,在合适的波段范围均能实现板蓝根颗粒的厂家鉴别,但SVC模型的效果更佳,在多个波段范围都可实现100%的分类准确率。较高的分类准确率同时也表明三个厂家的生产工对比,发现后者采集的数据更易于获得满意的分类结果。5.光谱法研究反式白藜芦醇和虎杖苷与牛血清白蛋白的相互作用主要采用荧光技术分别对反式白藜芦醇及虎杖苷与牛血清白蛋白的相互作用进行了研究,推测了结合机理。利用中红外光谱信息结合二阶导数和自退卷积技术对酰胺Ⅰ带光谱进行曲线拟合,分析药物与BSA结合后对BSA二级结构的影响,辅助荧光实验进行相互作用机理研究。分析结果表明两种化合物与BSA的相互作用是静态猝灭机理,结合过程自发进行,均使BSA的二级结构组成发生了变化,但反式白藜芦醇与BSA的结合常数大于虎杖苷与BSA的结合常数,且主要作用力类型有差异。更多还原

【Abstract】 In this paper, near infrared (NIR) spectrometry combined with chemometrics has been applied for quantitative and qualitative analysis of the components in drugs; mid-infrared spectroscopy combined with second derivative, self-deconvolution and curve fitting also has been used for study the changes of the secondary conformation of bovine serum albumin (BSA) when binding with drug molecules, and satisfying results have been obtained. The main contents of the paper are described as follows:1. Determination of three enantiomeric compositions of ehiral compounds by NIR spectrometry and chemometric analysisThe NIR spectrometry combined with partial least squares regression (PLS) was used to determine the D-isomer compositions of ehiral compounds, such as histidine, menthol and tryptophane, discriminated by BSA. The PLS model was compared with principal component regression model and multiple linear regression model. To optimize models, the range of wavenumber and the preprocessing methods were screened, and the parameters were optimized. The results show that NIR spectrometry combined with PLS can correctly predict the D-isomer compositions in the mixture. The accuracy of the model for D-histidine is higher than models for D-menthol and D-tryptophane, this difference may relate to the specific recognition of BSA to ehiral compounds.2. Simultaneous determination of general flavone and polysaccharide in Sophora japonica by NIR spectrometryThe support vector regression (SVR) was applied to construct the mathematic model to correlate near infrared spectral features with the composition of general flavone and polysaccharide, and the results were compared with PLS. The range of wavenumber and the preprocessing methods were screened. The radial basis function was used as kernel function for SVR, the parameters of SVR were optimized by grid search technique. The results show that SVR model performs significantly better than PLS for general flavone; and for polysaccharide determination, the performance of SVR is slightly better than PLS. Compare general flavone and polysaccharide, two kinds of model both show better performance for general flavone.3. Quantitative analysis of nicorandil powder via NIR spectrometryThe NIR spectrometry combined with SVR was used to determine the concentration of nicorandil. Principal component analysis (PCA) and independent component analysis (ICA) were used to compress information. The results show that two methods can both improve the performance of SVR model and can accelerate the speed of modeling. In particular, ICA can improved the performance of SVR model significantly, and made the performance of SVR better than that of PLS.4. Identification of Banlangen Granules from different manufacturers via NIR spectrometryBased on near infrared data of Banlangen Granules, two methods, i.e. clustering analysis and support vector classification (SVC), were used to identification of Banlangen Granules from three different manufacturers. With appropriate preprocessing methods and parameters setting, two methods both can get satisfactory results in proper wavenumber ranges. SVC show better performance than clustering analysis for the accuracy rates of it can get to100%in several wavenumber ranges. The high accuracy also shows that the manufacturing process and product quality of these three manufacturers are stable. Comparing two data collection methods, when the sample slowly rotated in watch glass, the near infrared data is easier to get satisfactory results.5. Study on the interaction of tran-resveratrol and polydatin with BSA by spectroscopyThe fluorescent spectrometry was used to study the interaction of BSA with trans-resveratrol and polydatin respectively and the bonding mechanisms were speculated. The mid-infrared spectra in amide I were processed with second derivative, self-deconvolution and curve fitting methods, the results can used to analyze the changes of protein major secondary structures after interact with trans-resveratrol and polydatin. The results of mid-infrared analysis can assist fluorescent spectrometry to study the interaction mechanisms. Trans-resveratrol and polydatin are all interact with BSA through static quenching procedures, and the procedures happen spontaneously, the interaction made the protein major secondary structures changes. However, the binding constant of trans-resveratrol with BSA is greater than that of polydatin with BSA, and the main interaction forces are different.更多还原