The Control Study Of Critical Production Process Of Human Albumin By Using Near Infrared Spectroscopy

Human albumin (HA) is an important and frequently-used blood product in clinical practice, while it is quite expensive to afford for ordinary people. At present, in Chinese enterprises, HA is mainly obtained from the separation of raw plasma by cold ethanol fractionation method, which is restricted by the plasma volume donated by volunteers. In the present production process, the critical process parameters (CPPs) are unaltered even different volunteers’plasma have different content of protein. It is obviously inappropriate. Therefore, there has a research significance to improve the yield of HA by accurately monitoring and controlling the critical production processes using process analytical technology (PAT). As a widely used PAT tool, near infrared spectroscopy (NIRS), due to the advantages of rapid, non-destruction and environment friendly, has potential applications in online monitoring and quality analysis of manufacturing process in the field of pharmaceutical industry.In this research, NIRS was applied in ethanol precipitation and acetate buffer precipitation these two critical production processes of HA production to realize the process monitoring and the determining of key quality parameters. Moreover, the modeling approaches were investigated detailedly to improve the validity of NIRS model.The details are listed as below:(1) Research on FI+II+III supernatant ethanol precipitation process monitoring by using NIRSIn this study, ethanol precipitation process of HA was selected as one of research object, and 8 normal batches and 3 abnormal batches of ethanol precipitation process were carried out in laboratory. Before the establishment of qualitative monitoring model,6 normal batches samples were selected as calibration set to build principal component analysis (PCA) model, at the same time 2 normal batches and 3 abnormal batches samples were selected as validation set to investigate the model ability of process monitoring and fault diagnosis. The research results showed the PCA model could be successfully used in process monitoring and fault diagnosis. Before the establishment of quantitative analysis model, HA and total protein contents were selected as modeling parameters. Two partial least squares regression (PLSR) models for HA and total protein contents rapid determination were built by using calibration set of 6 batches samples and validation set of 2 batches samples. And determination coefficient of prediction (Rp2), root mean square error of cross validation (RMSECV), root mean square error of prediction (RMSEP) and ratio of performance deviation (RPD) were 0.969,0.530 g/L,0.341 g/L,5.47 for HA, respectively, and 0.975,0.501 g/L,0.465 g/L,5.57 for total protein, respectively.(2) Research on HA content determination during acetate buffer precipitation process of FIV supernatant by using NIRSIn this study, HA content was determined by NIRS to realize the HA content process monitoring during acetate buffer precipitation process. Firstly,8 batches of acidic precipitation process was simulated in laboratory, and HA contents were determined by bromocresol green (BCG) method. Then,5 batches samples were selected as calibration set and 3 batches samples were selected as validation set to build PLSR model for HA content determination. Before the establishment of PLSR model, different variable selection methods were compared detailedly, and 35 variables were selected to build model. The determination coefficient of calibration (R2), Rp2, RMSECV and RMSEP were 0.977,0.978,0.7038 g/L and 0.5893 g/L, respectively.(3) Research on pH values determination during acetate buffer precipitation process of FIV supernatant by using NIRSIn this study, pH values during acetate buffer precipitation process were determined by NIRS to realize endpoint determination of acidic precipitation process. Firstly,6 batches acidic precipitation process were carried out in laboratory. Ethanol contents were determined by gas chromatography (GC) and pH values were obtained by acidity meter. Then, PLSR model for pH value determination was built by with 74 samples. Before the establishment of model, several influence factors, including sample set selection methods, spectral pre-processing methods and background spectra were investigated in detail. Finally, a best model for pH values determination was obtained, and the values of Rc2, Rp2, RMSEC, RMSECV and RMSEP were 0.968, 0.956,0.0512,0.0875 and 0.0594, respectively. The results showed that PLSR model had excellent predictive ability and repeatability and could determine the pH value and acidic precipitation endpoint accurately.The innovation points are listed as below:(1) This research brought NIRS to HA production firstly, and achieved monitoring the ethanol precipitation process qualitatively and quantitatively for the first time.(2) This research established PLSR model for the determination of HA content and pH value during acetate buffer precipitation process by using NIRS for the first time.(3) This paper investigated various model optimizing methods and provided beneficial reference for the model establishment in complex system.