【作者】 姜玮；

【导师】 臧恒昌； 刘爱华；

【作者基本信息】 山东大学 ， 制药工程（专业学位）， 2015， 硕士

【摘要】 人凝血因子Ⅷ (coagulation factor Ⅷ, FⅧ)是治疗甲型血友病和获得性凝血因子Ⅷ缺乏而致的出血症状等的不可或缺的药品,但我国当前整个血液制品行业长期处于供不应足的状态,因此提高产品的活性收率以及比活性是解决现状的一个必要举措。国内FⅧ的生产过程多采用离子交换层析法,从人血浆冷沉淀中分离纯化FⅧ,包括酸沉淀、A50吸附、S/D灭活、层析、超滤、冻干等主要环节,由于缺乏很好的过程控制方法,导致产品收率低、比活低。因此,有必要针对生产现状,将现代过程分析与控制技术引入到FⅧ生产过程中的关键环节当中,增加对生产工艺过程的了解,有效控制工艺过程,提升FⅧ的收率及比活性。近红外光谱分析技术(Near Infrared Spectroscopy, NIRS)是当前最重要的一种过程分析技术(Process Analytical Technology, PAT),具有快速准确无损等优势,在制药行业中已有较为广泛的应用,在血液制品中的应用也有研究。本课题选取FⅧ生产过程中的酸沉淀、层析洗脱和冻干环节为研究对象,采用NIRS结合不同的化学计量学方法,尝试建立了针对FⅧ比活性、效价以及水分的定量分析模型,为其在线应用奠定了基础。具体研究内容如下：1.FⅧ酸沉淀过程分析和近红外定量模型的建立本实验考察了人凝血因子Ⅷ生产过程中酸沉淀终点控制方法的合理性,然后用近红外光谱分析技术建立了FⅧ比活性的偏最小二乘(partial least squares,PLS)定量分析模型,选取了测定过一级数据的108个样品光谱参与建模,其中随机选取6次试验的64个样品划入校正集,另外4次试验的44个样品划入验证集,通过对校正集样品和验证集样品的含量分布以及主成分得分图进行分析,证明分类结果比较理想,满足样品集的划分要求。通过对校正集样品的统计分析,将杠杆值或学生化残差值过高的样品进行判断,若确为异常样品则予以剔除。在光谱测量过程中不可避免地会引入噪声和产生基线漂移,在建立校正模型前有必要对光谱进行适当的预处理,微分可以有效消除基线的漂移,放大信号差异,平滑多与微分联用,可有效提高信噪比,标准化(Autoscale)可以赋予光谱中所有波长的变量相同的权重,凸显样品光谱的差异,最终选择这3种预处理方法—一阶微分+SG13点平滑十标准化,采用处理后的光谱数据建立PLS校正模型,模型的各项参数较为理想,Rc=0.9649, Rp=0.9229, RMSEC=0.0576, RMSEP=0.0493。对建立好的模型进行预测精密度的考察,在验证集样品中随机选取3个样品,包含不同的FⅧ比活性值,重复测量6次光谱,然后采用建立的模型通过采集光谱得到样品的预测值,经计算预测值的相对标准偏差在5%左右,且随着比活的升高而减小,证明此分析方法具有较好的重复性,可用于未知样品的预测。2.人凝血因子Ⅷ层析过程中洗脱液效价的近红外定量分析研究本实验建立了FⅧ层析过程中洗脱液FⅧ效价的定量模型。层析洗脱过程获得了68个样品,FⅧ效价分布比较均匀,符合样品集的要求,且不存在显著异常点。比较不同方法的PLS建模结果可知,选择中心化(Meancenter)+基因算法(Genetic algorithm, GA)所建立的PLS模型R2为0.9010、RMSEP为0.4090,既有较高的线性,也降低了预测误差,它是所使用的最佳PLS建模方法。而采用ANN、SVM方法建立模型,其结果表明采用sigmoid传输函数的BP-ANN建模方法的R2=0.942, RMSEC=0.3754, RMSEP=0.4272,是所有方法中最佳的建模方式。这可能是因为FⅧ的效价是物质的活性信息,较为复杂,与物质的浓度不成定量线性关系,所以采用非线性的回归方法能够得到更好的建模结果。应用近红外光谱分析技术和此定量模型,可以在FⅧ层析洗脱过程中针对FⅧ效价进行特异性监测,从而更准确的控制收集洗脱液的起点和终点。3.人凝血因子Ⅷ冷冻干燥过程中水分含量的近红外定量分析研究本实验利用近红外光谱和相关的化学计量学建立了FⅧ冻干过程中水分含量的定量检测模型,通过考察不同的预处理方法和光谱区间选择方法,最终确定了正交信号校正(OSC)结合二阶导数加SG平滑的预处理方法,以及GA作为谱区选择方法。所建立模型的主要参数为R2=0.918, RMSEC=0.2141, RMSECV=0.4102, RMSEP=0.3848,均处于可以接受的范围内。通过验证,模型的线性、分辨能力、准确度和精密度等均符合要求,证明了该模型的实用性与有效性。该模型的建立,是实现在线监测冻干过程中FⅧ制剂水分含量变化的坚实基础,一旦在线监测系统得以实现,冻干时间将得到极大的优化,避免因冻干时间过长造成的FⅧ活性损失,而利用光纤进行多点监控,可以保证整批产品的水分含量都达到标准,使不合格产品的数量达到最低。本论文的创新点有：本研究首次将近红外光谱分析技术应用于人凝血因子Ⅷ的生产过程研究,分别建立了酸沉淀过程中FⅧ比活性的PLS定量模型、层析过程中洗脱液FⅧ效价的定量分析模型和冻干过程中水分含量变化的定量分析模型。更多还原

【Abstract】 Coagulation factor Ⅷ is an indispensable medicine for the treatment of Hemophilia A and the bleeding caused by acquired deficiency of FⅧ, but China’s current total blood products industry can’tmeet the need for a long period of state, so how to increase the activity yield and specific activity of the product is very important. DomesticFⅧproduction process use more of ion exchange chromatography, which isolated and purified FⅧ from human plasma cryoprecipitate, including acid precipitation, A50 adsorption, S/D inactivation, chromatography, ultrafiltration, freeze-dried, and other major steps. Due to the lack of a good process control technology, resulting in low production yield and low specific activity. Therefore, it is necessary to introduce the modern process analysis and control technology into the key steps of the FⅧ production process, in order to increase the understanding of the production process, and effectively control of the process, finally enhance the yield and specific activity of the product. Near infrared spectroscopy, the most important one of Process Analytical Technology with the advantages of fast, accurate and non-destructive, has been used widely inthe pharmaceutical industry, and also in blood product.This thesis selected 3 processes--acid precipitation, elution process in chromatography and freeze-drying as research object, and used NIRS combined with different Chemometrics methods, finally established quantitative analysis models in connection with FⅧ specific activity, potency and moisture, thus laid the foundation for its online application.The details are listed below:1. The acid precipitation process analysis and the establish of the quantitative analysis model.This study investigated the rationality of the endpoint control method of FⅧ acid precipitation process, and then build a partial least squares quantitative model for FⅧ specific activity by NIRS. The research selected 108 samples which have been determined the reference data, and randomly classified six trials including 64 samples as calibration set, four trials including 44 samples as validation set. Through the analysis of the distribution of the calibration set and validation set samples content as well as the principal component score plot, we found that the classification results are ideal and meet the requirements of the sample collection division. Through statistical analysis of calibration set samples, judged the samples with high leverage value and studentized residual value, and excluded abnormal samples. The spectrummeasurement process will inevitably introduce noise and generate baseline drift, so it is necessary for proper pretreatment of the spectrum before the establishment of the calibration model. Derivative methods can effectively eliminate the baseline drift and amplify the signal difference, Smooth combined with Derivative can effectively improve the signal to noise ratio, Autoscale can give the same right variable weight to spectrum at all wavelengths and highlight the differences between the sample spectra, so eventually we selected these 3 methods to pretreatment the spectrum. Then build the PLS calibration model, and the parameters of the model are as follows:Rc=0.9649, Rp=0.9229, RMSEC=0.0576, RMSEP=0.0493. We executed the predicting precision investigation of the model, and the relative standard deviation of the predicted value was about 5%, proving that the analysis method has a good reproducibility and can be used to predict unknown samples.2. Research of NIRS quantitative analysis for the FⅧ potency in the elution process of chromatography.This study established a quantitative model for the determination of FⅧ potency in the chromatography elute process. We got 68 samples from the elute process, which had a uniform potency distribution to meet the sample set requirement. Compared different PLS modeling results with kinds of methods, we knew that Meancenter and Genetic Algorithm were the best methods combination with the result as follows: R2=0.9010, RMSEP=0.4090, which has a high linear and a low prediction error. However, the BP-ANN modeling method with the transfer function sigmoid, has a better result than PLS and other methods, which has parameters as follows:R2=0.942, RMSEC=0.3754, RMSEP=0.4272. The FⅧ potency is the activity information of the substance, which is very complex and the concentration of the substance is not a linear relationship between the quantitative, so the nonlinear regression modeling results can be better. By using NIRS and this quantitative model, we can monitor the F Ⅷ potency specifically in the chromatographic elution process, and control the start and end points more accurately.3. Research of NIRS quantitative analysis for the product moisture in the FⅧ freeze-dried process.Utilized NIRS and Chemometrics, this study established a quantitative model for the determination of moisture in the product during the freeze-dried process. After investigated different pretreatment and spectral region selection methods, we eventually chose OSC combined with second derivative plus SG smooth as pretreatment methods and GA as spectrum region selection method. The model parameters are as follows:R2=0.918, RMSEC=0.2141, RMSECV=0.4102, RMSEP=0.3848. Through validation, the model linear, resolution, accuracy and precision are adhere to the requirements, which proved the practicality and effectiveness of the model. The establishment of the model is the foundation for on-line monitoring FⅧ moisture content changes in lyophilized preparation process, which can optimize drying time and avoid the activity loss caused by too long dying time. Multi-point monitoring using fiber can ensure that the moisture content of bulk products reach the standard.The innovation points are listed below:This research applied NIRS in the FⅧ production process for the first time, respectively established PLS quantitative model for the FⅧ specific activity analysis in the acid precipitation process, quantitative model for the FⅧ potency analysis in the chromatography elution process and PLS quantitative model for the moisture analysis in the freeze-dried process.更多还原