Research On Development Of Near-infrared Models In Heparin Sodium Refining Process

Near-infrared (NIR) spectroscopy, providing direct chemical information related to substance, is generally chosen as a powerful tool of Process Analysis Techology (PAT) for quantitative and qualitative analysis in agriculture, food, wine, animal feed, tobacco, chemical and pharmaceutical industries in combination with chemometrics. NIR is quick, non-destructive, requiring no sample preparation or hazardous chemicals. And now NIR has been applied to the various operations in pharmaceutical process:rapid identification of raw materials, process monitoring and understanding, and final product release testing. However, the intensities of NIR absorption bands are weak and the bands are usually broad, overlapping from multiple components and/or functional groups. Therefore, to explore the method to extract the feature information applied to effectively establish NIR analysis models from complex background has been the common issues for the application of NIR in pharmaceutical field. It is also the bottleneck restricting the further development of NIR technology.Heparin, highly sulfated glycosaminoglycans (GAGs), is used as anticoagulant drugs with high economic efficiency and complex background in manufacturing process. Heparin was used as a model drug, and the aim of this study is to apply of NIR spectroscopy in critical quality control points of heparin refining process including enzymolysis, ethanol precipitation and drying, and to establish the qualitative and quantitative NIR analysis models under complex background, which will lay the foundation for the further development and application of NIR spectroscopy in pharmaceutical field. The results and conclusion of the research about heparin process based on NIR spectroscopy are listed belows:1. Application of NIR spectroscopy in determination of heparin potency in crude heparin enzymatic-removing-protein processIn this study, to rapid and efficiently measure heparin potency in crude heparin enzymatic-removing-protein process, three different linear and nonlinear regressions tools (i.e. partial least squares (PLS), back propagation artificial neural network (BP-ANN), and support vector regression (SVR) including epsilon-SVR and nu-SVR), were systemically studied and compared in developing the quantitative analysis model of72samples from6batches. The number of principal component (PCs) was optimized by venetian blinds cross-validation method, and its performance was tested by the coefficient of determination of calibration set (R2c), coefficient of determination of validation set (R2p) and the root-mean-squares error of validation (RMSEP). The results showed that the performance of models developted by nonlinear regressions tools were better than PLS, and the optimum result of the epsilon-SVR model was achieved as follow:R2c=0.9952, R2p=0.7661and RMSEP=3.5449IU/mg. The overall results sufficiently demonstrated that the epsilon-SVR regression tool has the potential to effectively determination the potency of heparin in crude heparin enzymatic-removing-protein process.2. Application of NIR spectroscopy in understanding and monitoring ethanol precipitation of the mixture of glycosaminoglycansIn this study, mixture with different percentages of heparin and chondroitin sulfate were prepared and the NIR spectra of16batches were achieved from ethanol precipitation of the mixtures. Firstly, spactra of8batches were analysed by principal component analysis (PCA) and the score control charts were established to identify the characteristic trajectory. And the all8batches had a similar trajectory, and the control chart of PC1scores revealed the increasing amount of ethanol, and the control chart of PC2scores reflected the stability of the ethanol precipitation, while the control chart of PC3scores mainly reflected the multi-stage variation of precipitation process. Then, the process was divided into three phases and partial least regression discriminant analysis (PLS-DA) monitoring model was developed. And the predictive ability of the PLS-DA model was reliable. It was suggested that PCA combined with PLS-DA method was a rapid and efficient tool to monitor the multi-stage variation of precipitation process. What’s more, this study offers a valid way to rapid monitoring of heparin ethanol precipitation to obtain high purity single heparin.3. Application of NIR spectroscopy in monitoring heparin drying processIn this study, NIR spectroscopy was adopted to demonstrate the feasibility of analyzing the heparin drying process qualitatively and quantitatively with chemometrics. A total of210samples were acquired from15different batches (12normal batches and3abnormal batches) during the heparin drying process. Firstly, a quantitative PLS model with168samples from12normal batches was established, and the values of R2P and RMSEP were0.9560and0.9030%, respectively. What’s more, the ratio of standard deviation of the validation set to standard error of prediction (RPD) of the PLS model was4.46, indicating that the predictive results were acceptable. Then, the qualitative models were respectively built based on PC1, Hotelling’s T2and standard error of prediction (SEP) with samples from15batches. The statistical control charts could successfully distinguish the normal and abnormal batches, suggesting that NIR spectroscopy had the capability to realize the quality control and the fault diagnosis of drying process. Both the qualitative and quantitative results above demonstrated that the PLS and MSPC is a effective method to monitor the drying process of heparin and could provide a convenient method for analyzing and understanding the process.Major innovative achievements: 1. Established a NIR method to monitor the heparin potency in the crude heparin enzymatic-removing-protein process.2. Established a NIR method to monitor the multi-stage variation of ethanol precipitation of the mixtures of heparin and chondroitin sulfate.3. Established a NIR method to qualitatively and quantitatively analysis of heparin drying process.