Recombinant hepatitis B vaccine stock solution identification method and process quality control method based on tunable Raman spectrum

Abstract

The invention relates to a tunable Raman spectrum-based recombinant hepatitis B vaccine stock solution identification method and a process quality control method. On the basis of a tunable Raman spectrum measurement system, multi-dimensional Raman spectrum information of a recombinant hepatitis B vaccine stock solution sample to be detected under different concentration disturbance conditions is tested, spectral characteristics are extracted from a high-dimensional spectrum, and characteristics of recombinant hepatitis B vaccine target protein HBsAg are represented through common peaks and specific peaks in two-dimensional correlation analysis. And the secondary structure of the target protein HBsAg of the recombinant hepatitis B vaccine can be identified or structural characterization and analysis can be further realized on the basis of spectral data. By means of the characterization method, the recombinant hepatitis B vaccine stock solutions of different production matrixes can be characterized and identified, and process quality control and evaluation of the recombinant hepatitis B vaccine stock solutions can be further achieved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of detection, and particularly relates to a method for identifying and a process quality control method for a recombinant hepatitis B vaccine stock solution based on tunable Raman spectroscopy. BACKGROUND

[0002] Recombinant hepatitis B vaccine is a biological product used to prevent hepatitis B. It is produced by expressing hepatitis B surface antigen (HBsAg) in yeast or Chinese hamster ovary (CHO) cells using genetic engineering techniques, thereby stimulating the human immune system to produce protective antibodies against hepatitis B virus. There are three types of recombinant hepatitis B vaccines in China, namely, recombinant hepatitis B vaccine (Saccharomyces cerevisiae), recombinant hepatitis B vaccine (Hansenula polymorpha), and recombinant hepatitis B vaccine (CHO, Chinese hamster ovary cells).

[0003] Quality assurance of recombinant hepatitis B vaccine is crucial for ensuring its safety, effectiveness, and preventing the widespread transmission of hepatitis B. High-quality vaccines can provide strong immune protection and reduce the risk of chronic infection and related complications, such as cirrhosis and liver cancer. Chinese vaccines need to go through strict regulatory approval processes before they are marketed, including clinical trials and quality assessments, to ensure vaccine quality. After the vaccine is marketed, it also needs to be continuously monitored and evaluated to ensure long-term safety and effectiveness.

[0004] Vaccine production processes need to strictly follow good manufacturing practices (GMP) to ensure the purity and safety of vaccines, including strict control of the production environment, screening of raw materials, and monitoring of each link in the production process. The Chinese Pharmacopoeia 2020 Edition Volume III proposes process quality control requirements for human vaccines, including whole-process quality control, batch consistency control, and control of target components and non-target components.

[0005] The control requirement of "batch consistency" is a control defect that is urgently needed but lacks methods. The quality control method of recombinant hepatitis B vaccine is mainly biological method, which is difficult to characterize and evaluate the batch consistency. The evaluation of "batch consistency" requires a fast and reliable analytical method that is compatible with QbD. SUMMARY

[0006] To solve the above technical problems, the application provides a method for identifying and a process quality control method for a recombinant hepatitis B vaccine stock solution based on tunable Raman spectroscopy.

[0007] The technical solution adopted by the application is: detecting the Raman spectrum of the recombinant hepatitis B vaccine stock solution under different laser wavelengths under concentration disturbance conditions, processing the obtained original spectrum by two-dimensional correlation analysis method, and using the two-dimensional correlation spectrum to in-situ characterize the primary structure and / or high-order structure of the recombinant hepatitis B vaccine target protein solution.

[0008] Preferably, the specific steps are as follows: The specific steps are as follows: Step 1: Under concentration perturbation conditions, Raman spectra under different wavelength excitations are detected using a tunable Raman spectroscopy measurement system to obtain the data cube of the recombinant hepatitis B vaccine stock solution; thus forming the original spectrum; Step 2: Perform principal component analysis on the original spectrum to generate principal component spectra, and use the principal component spectra to replace the original spectrum for two-dimensional correlation analysis to obtain synchronous and / or asynchronous spectra; Step 3: In situ characterize the primary and / or higher-order structures of the recombinant hepatitis B vaccine stock solution by combining synchronous and / or asynchronous spectra.

[0009] Preferably, the tunable Raman spectroscopy measurement system includes a tunable laser source, a beam splitter, a fiber optic coupler, a photodetector, a Raman spectrometer, a reference cell, and a sample cell. The tunable laser source emits lasers of different wavelengths, which are split into two optical paths by the beam splitter. In one optical path, the laser passes through the reference cell and the first fiber optic coupler and enters the photodetector. In the other optical path, the laser passes through the sample cell and the second fiber optic coupler and enters the Raman spectrometer.

[0010] Preferably, the excitation wavelength tuning range is 532-785 nm, and the Raman spectroscopy measurement range is 3083-75 cm⁻¹. -1 Light.

[0011] Preferably, in step two, the calculation method and formula are as follows: Analysis of the function of spectral change under the influence of external perturbation variables ,in The wavelength (°C) represents the excitation wavelength of the external perturbation variable. Wave number (cm) -1 When λ is in and When the values ​​vary between these values, the measured spectral intensity y, i.e., the original spectral set, is defined as follows: : (Formula 1); in, It is the average spectrum: (Formula 2); Two-dimensional correlation synchronous spectral intensity For different wave numbers Vector product of dynamic spectral intensities: (Formula 3); Two-dimensional correlated asynchronous spectral intensity For different wave numbers Lower dynamic spectral intensity Hilbert-Noda matrix ( ) vector product: (Formula 4); wherein, As shown in formula 5: (Formula 5).

[0012] Preferably, the characteristic peak of the diagonal line in the synchronous spectrum is analyzed as the characteristic signal of the target protein for identification and structural characterization; Preferably, the structural characteristics of the recombinant hepatitis B vaccine bulk are analyzed by analyzing the slice spectrum at a specific wave number in the heterogeneous spectrum.

[0013] A method for identifying a recombinant hepatitis B vaccine bulk, which is characterized by a recombinant hepatitis B vaccine structural characterization method based on tunable Raman spectroscopy, so as to determine whether the vaccine bulk to be tested is a recombinant hepatitis B vaccine bulk, or to determine what kind of recombinant hepatitis B vaccine bulk the vaccine bulk to be tested is.

[0014] Preferably, the spectrum information of the upgraded dimension is converted into a feature set G according to a hierarchical clustering algorithm, and a partial least squares is combined to establish a recombinant hepatitis B vaccine bulk identification analysis model; When establishing the recombinant hepatitis B vaccine identification model, multiple batches of recombinant hepatitis B vaccine bulks prepared by different processes and multiple types of other protein solutions are used for verification, and the recombinant hepatitis B vaccine bulk is used as positive verification, and the other types of protein solutions are used as reverse verification. When establishing the recombinant hepatitis B vaccine type identification model, multiple batches of recombinant hepatitis B vaccine bulks prepared by different processes are used for verification, and the recombinant hepatitis B vaccine bulk prepared by one of the processes is used as positive verification, and the recombinant hepatitis B vaccine bulks prepared by other processes are used as reverse verification, to obtain the recombinant hepatitis B vaccine bulk identification model prepared by the process.

[0015] A recombinant hepatitis B vaccine bulk preparation process control evaluation method, which takes the discrimination value output by the identification model in the method for identifying a recombinant hepatitis B vaccine bulk as the research object, and performs production process fluctuation characterization on samples of the same production process; and performs group process level characterization on samples produced by different processes.

[0016] The present application has the advantages and positive effects that: the tunable Raman spectroscopy is used to detect the sample to be tested, and a two-dimensional matrix feature set G is obtained after dimension upgrading, and the target protein characteristic spectrum is extracted as the basis for characterization and evaluation of the quality of the recombinant hepatitis B vaccine bulk; a new method for characterizing the quality of the recombinant hepatitis B vaccine bulk is provided. The identification model can be used for identifying the target protein, and can also identify the matrix and batch, and solve the problem of batch consistency evaluation difficulty. The quantitative output value of the identification model can be used as the data basis of statistical process control and population quality evaluation, and can be used for continuous quality monitoring of the product by the production unit and comprehensive control of the quality trend of all products of the domestic recombinant hepatitis B vaccine by the regulatory department, so that the production of the recombinant hepatitis B vaccine based on QbD can be realized, and the product quality can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 Structure diagram of a tunable Raman spectrum measurement system; wherein, 1, a tunable light source, 2, a beam splitter, 3, a sample cell, 4, a fiber coupling mirror, 5, a Raman spectrometer, 6, a reference cell, 7, a photodetector; Figure 2 Flow chart of the identification model of the recombinant hepatitis B vaccine bulk solution; Figure 3 Relationship between the first principal component score and the concentration; Figure 4 Generalized two-dimensional correlation analysis synchronous spectrum; a is a manufacturer 1; b is a manufacturer; c is a manufacturer 3; d is a manufacturer 4; Figure 5 Distribution diagram of the identification value of the PLS-DA model based on the two-dimensional correlation analysis synchronous spectrum for the solution type; (a): training set (b): verification set; Figure 6 Quality control chart of the recombinant hepatitis B vaccine bulk solution; (a): mean value control chart of manufacturer 1; (b): mean value control chart of manufacturer 3; (c): range control chart of manufacturer 1; (d): range control chart of manufacturer 3. DETAILED DESCRIPTION

[0018] The embodiments of the present application will be described below in combination with the drawings.

[0019] The present application relates to a kind of recombinant hepatitis b vaccine bulk based on tunable raman spectrum identification method and process quality control method.First, a kind of tunable raman spectrum measurement system is constructed, the laser of different wavelengths is generated in the system through tunable laser, as Raman excitation light source, measure the Raman spectrum of sample under different wavelengths;The coupling information of Raman spectrum of sample under different wavelengths is processed as multidimensional intensity distribution, so as to provide the basis for target protein structure resolution and discrimination.Based on the tunable raman spectrum measurement system, the multidimensional Raman spectrum information of the sample of the recombinant hepatitis b vaccine bulk under different concentration disturbance conditions is tested, the spectral features are extracted from high-dimensional spectrum, the common peaks and specific peaks of the target protein of the recombinant hepatitis b vaccine are characterized by two-dimensional correlation analysis, and the production matrix recognition based on the secondary structure of the target protein can be further realized.The batch difference characterization and analysis.

[0020] The tunable raman spectrum measurement system includes two light paths, as Figure 1 Different wavelength lasers are emitted in tunable laser light source 1, and the lasers are divided into two light paths through beam splitter 2, the laser in one light path passes through reference cell 6 and fiber coupling mirror 4 in turn, enters photodetector 7, and reference cell 6 can be used to deduct the spectral background;The laser in the other light path passes through sample cell 3 and fiber coupling mirror 4 in turn, and enters Raman spectrometer 5.For detecting sample, the Raman spectrum of the same sample excited under different wavelength laser conditions is detected, and high-dimensional spectral data information of target protein is established.

[0021] As shown in Figure 2 Based on the tunable raman spectrum measurement system, the Raman spectrum data of the recombinant hepatitis b vaccine bulk under concentration disturbance conditions is collected, and data cube is established.The obtained spectrum is pretreated, principal component spectrum is used to replace original spectrum for two-dimensional correlation analysis, the original spectrum is processed by dimension increasing, target protein structure information is separated by two-dimensional correlation, and the obtained synchronous spectrum and asynchronous spectrum are used to analyze the characteristics of the structure of target protein (such as HBsAg) with concentration change kinetics.The orthogonal cross peak of synchronous spectrum represents the direction of target protein kinetic change, and the asynchronous spectrum represents the rate of target protein kinetic change with temperature.

[0022] Whether the target protein is the recombinant hepatitis b vaccine can be identified by the method, further, the same type of protein product prepared by different preparation processes (matrix) can be identified, the process type (matrix) can be identified and judged, and the batch difference can be further characterized and analyzed.In some embodiments of the present application, the method is used to identify the recombinant hepatitis b vaccine bulk prepared by different production matrices, which can be used to identify whether it is the recombinant hepatitis b vaccine bulk, and can be used to identify whether the recombinant hepatitis b vaccine bulk is prepared by different processes (matrix) such as saccharomyces cerevisiae, CHO cell or hansenula.

[0023] The specific steps are as follows: Step 1: Data collection of recombinant hepatitis B vaccine stock solution; detect the vaccine stock solution by a tunable Raman spectrum measurement system, detect the Raman spectrum of the vaccine stock solution under different excitation light wavelength detection conditions under different concentration disturbance conditions; when the method is used for identification of the vaccine stock solution, three concentration disturbance conditions of 120%, 100% and 80% can be used; the tunable Raman spectrum acquisition parameters are as follows: excitation light wavelength tuning range 532-785 nm, Raman spectrum measurement range 3083-75 cm -1 -3 cm -1 , laser power 200 mW, integration time 50 s, and accumulation number 10.

[0024] The obtained Raman spectrum data are pretreated by SG function 2-order 15-point filtering denoising.

[0025] Step 2: Separate target protein structure signals by two-dimensional correlation spectrum method; in order to improve the spectral resolution, especially the difference in Raman spectra obtained by different excitation wavelengths in tunable Raman spectrum, the Raman spectra of the sample under different wavelengths are processed in a higher dimension, so as to improve the spectral resolution. The Raman spectrum characteristics excited by different wavelengths are extracted, then a plurality of pairs of characteristics are screened for similarity measurement, and the positions of the characteristics in the matrix coordinates are determined according to the similarity between the characteristics. After the positions of each characteristic in the characteristic matrix are determined, the expression value or characteristic value is mapped. This will generate a unique image for each sample.

[0026] The principal component spectrum obtained at each excitation wavelength is used to replace the original spectrum, and two-dimensional correlation analysis is performed under the excitation wavelength as a disturbance condition. The tunable Raman spectrum at different wavelengths can be expressed as , wherein represents different excitation wavelengths, is the wavenumber (cm -1 ). According to extract spectral characteristics, including; (formula 1); wherein, is the minimum wavelength, is the maximum wavelength is the average spectrum: (formula 2); correlation spectrum characteristics: (formula 3); wherein, m is the total number of wavelength points m= - , For the i-th wavelength point, the index of the wavelength point in the i-th wavelength range, used for traversing all measured wavelength points; Spectral vector features: (Equation 4); Wherein, The i-th wavelength point, The k-th wavelength; Hilbert-Noda matrix (Equation 5); Wherein, k represents the index of another wavelength point, similar to i, used for traversing the wavelength points in the Hilbert-Noda matrix ; j represents, together with k, the index of the matrix element in the definition of the Hilbert-Noda matrix .

[0027] Two sets of two-dimensional correlation spectra are obtained through the above process, which are defined as synchronous spectrum and asynchronous spectrum respectively, and can be further used for characterization and analysis of the structure of the target protein of the recombinant hepatitis B vaccine.

[0028] Step 3: Analysis of the correlation between the structure of the target protein of the recombinant hepatitis B vaccine and the production matrix; by analyzing the synchronous spectrum of the two-dimensional correlation analysis of the product obtained from different production matrices, the common spectrum area and the specific spectrum area of the recombinant hepatitis B vaccine bulk expressed by different production matrices are compared; according to the information of the two-dimensional correlation spectrum, the commonness of the target protein of the recombinant hepatitis B vaccine and the difference caused by different production matrices are explored.

[0029] The asynchronous spectrum of the two-dimensional correlation analysis can also be further analyzed, and the slice spectrum at a specific wave number in the asynchronous spectrum is analyzed, so as to speculate the secondary structure characteristics of the protein.

[0030] In some embodiments of the present application, unknown samples can be identified and detected according to the characteristics of the two-dimensional correlation spectrum. First, a standard spectrum is established for a known sample type, and then for an unknown sample, the Raman spectrum data set under different laser wavelength conditions at different concentrations is detected, and after dimensionality processing, the autocorrelation peak of the unknown sample is obtained according to the synchronous spectrum information, and the peak information embodied by the two-dimensional correlation spectrum is compared with the common peak and / or specific peak in the standard spectrum, so as to judge the type of the unknown sample or the manufacturing process used. The gap between different preparation processes can also be measured according to the number and distribution of common peaks and specific peaks.

[0031] Based on the above vaccine solution characterization method and structural characteristics, a vaccine solution identification model can be established to directly determine the type and process type of the target solution to be tested; an identification model can be established for each of the solution type and process type, and when used to identify the solution type, it can include step 4-1, which is as follows: Step 4-1: Establishment of identification model; using multiple batches of recombinant hepatitis B vaccine bulk solution prepared by different processes as samples, adding human serum albumin solution, bovine serum albumin lyophilized powder, placenta polypeptide injection, recombinant human interleukin for injection, interferon eye drops, compound amino acid and immunoglobulin injection and other types of solution samples; obtaining the dimensionality data of two-dimensional correlation analysis of the above samples according to step 1-3; converting the dimensionality of the spectral information into a feature set G according to the hierarchical clustering algorithm, and combining the partial least squares to establish a recombinant hepatitis B vaccine bulk solution identification analysis model. The identification model uses different batches of recombinant hepatitis B vaccine bulk solution prepared by different production substrates as positive verification, and other protein biological products as reverse verification.

[0032] When used to identify the vaccine solution preparation process, it can include step 4-2, which is as follows: Step 4-2: Establishment of identification model; selecting vaccine solutions prepared by different preparation processes, and establishing an identification model for each type of vaccine; obtaining the dimensionality data of two-dimensional correlation analysis of the selected solution samples according to step 1-3; converting the dimensionality of the spectral information into a feature set G according to the hierarchical clustering algorithm, and combining the partial least squares to establish a recombinant hepatitis B vaccine bulk solution identification analysis model; when establishing an identification model for a type of preparation process, using the solution of this type as positive verification and other types as reverse verification to obtain the identification model of this type. Then, an identification model is established for each process type according to the same method.

[0033] In the process of establishing the model, first, the dimensionality of the spectral information is converted into a feature set G according to the hierarchical clustering algorithm. The hierarchical clustering of the dimensionality of the spectral information is performed to determine the distance between different features. In the hierarchical clustering algorithm, the distance between data clusters is a key parameter for accurate and reasonable clustering. Designing a suitable distance can achieve good classification performance and fast calculation speed. It mainly includes bottom-up condensation clustering and top-down splitting clustering. The condensation clustering hierarchical clustering method is selected, and the minimum distance is used as the clustering criterion.

[0034] The number of initial data clusters is set to The algorithm complexity of the traditional hierarchical clustering method is When the amount of original data is large, the calculation time will increase rapidly. In some embodiments of the present application, the method is improved, and the steps of the improved hierarchical clustering method are as follows: Step (1): Calculate the feature distance between each two objects in the data setD and stored in an array d middle; As a widely used similarity metric, distance dynamics time warping distance, based on the idea of ​​dynamic programming, not only avoids the one-to-one matching problem of Euclidean distance, but also achieves a "one-to-many" mapping between time series by warping sequences. Given two spectral sequences... and Numerical information distance (set to) The calculation formula is: (Formula 6); In the formula, ΔΕ, ΔΜ, Δk They are respectively X, Y The difference in the mean, the difference in the extreme values, and the difference in the slope of the sequences; a, b, c These are the weights corresponding to the differences, and a + b + c =1.

[0035] Construct a n × m matrix network A In matrix networks A The matrix elements are represented as... x i and y j The distance between two points. (The matrix is ​​then used for this purpose.) A The set of each pair of adjacent elements in the equation is called a curved path, denoted as P = { p 1 ,p 2 ,…,p s ,…,p k Find an optimal curved path such that the sequence X and Y The total cost of bending is minimized, that is: (Formula 7); Therefore, the feature distance between data clusters is defined. D : D = aD 1 +βD DTW (Formula 8); In the formula, α , β These are the weights for the corresponding distances, and α + β =1.

[0036] Step (2): For the arrayd ascending order of distance values; Step (3): judging the current element in the array d whether the two objects corresponding to the current element are in the same class, if not, merging the two objects into one class; if one of the objects has been assigned to a class, merging the other object into the class; if they both have their own class, merging the classes into one class; Step (4): taking the next element in the array, repeating step (3) until all data in the array is processed.

[0037] According to the feature clustering results, the feature set G can be obtained, wherein each row is the same class feature, and each column is the class feature randomly arranged.

[0038] The models constructed by steps 4-1 and 4-2 can be verified; the true positive rate (TPR) (formula 9), the true negative rate (TNR) (formula 10) are used to evaluate the sensitivity and specificity of the model, and the model accuracy (ACC) (formula 11).

[0039] TPR = TP / P = TP / (TP + FN) (formula 9); TNR = TN / P = TN / (TP + FN) (formula 10); ACC = (TP + TN) / (P + N) (formula 11); wherein, TP TP is the number of true positives, TN TN is the number of true negatives, P TP is the total number of positives, N TN is the total number of negatives, TN TN is the number of true negative samples, FN FN is the number of false negative samples.

[0040] By combining the structural characterization information with the pattern recognition method to establish a model, the rapid identification of the sample to be tested can be realized; the quantitative output value of the model can be used as the data basis for statistical process control and population quality evaluation, and process quality monitoring and batch consistency evaluation can also be realized. For example, the above characterization method is used for the characterization of recombinant hepatitis B vaccine bulk, which can analyze the structural information of the target protein in the recombinant hepatitis B vaccine bulk, establish a discrimination model for the recombinant hepatitis B vaccine bulk, realize the rapid identification of the recombinant hepatitis B vaccine bulk, and to a certain extent, can be used for quality monitoring and batch consistency evaluation.

[0041] The recombinant hepatitis B vaccine is identified from the intermediate product of the key process step of vaccine production, i.e. the recombinant hepatitis B vaccine bulk. The extracted characteristic spectrum of the target protein contains common and characteristic information of three kinds of recombinant hepatitis B vaccine production substrates (Saccharomyces cerevisiae, Hansenula, CHO) bulk, and common and changing characteristic information of different batches of bulk of the same substrate, which can characterize and evaluate the quality of the recombinant hepatitis B vaccine bulk. The main component of the recombinant hepatitis B vaccine bulk is the virus-related target protein, and the structure of the target protein is taken as the basis for the quality evaluation of the vaccine bulk, so as to realize the process quality control of the final product and the consistency evaluation between batches. The tunable Raman spectrum is used, combined with the spectrum dimensionality method, to obtain the characteristic set G, and the model is established by combining the pattern recognition method, so as to realize the rapid identification. Through the identification method, each batch of product can be identified, and the consistency evaluation between batches can be met.

[0042] The process control method is specifically shown in step 5.

[0043] Step 5: Process control method Step 5-1 Process control method: The state of the production process is monitored by using a statistical method, which is used for process quality control, to determine that the production process is in a controlled state, so as to reduce the variation of product quality. The production process fluctuation is characterized by different batch means ( ) and batch range ( ), wherein the mean control chart takes as the lower control limit, takes as the upper control limit, and takes as the center line, which is calculated as shown in formulas 12 and 13, the range control chart takes as the lower control limit, takes as the upper control limit, and takes as the center line, which is calculated as shown in formulas 14 and 15, and the batch within the control limit meets the quality standard.

[0044] (Formula 12);

[0045] (Formula 13);

[0046] (Formula 14);

[0047] (Formula 15); wherein is the sampling interval, n is the sampling capacity, is obtained according to the number of tests, is the mean of the range of different batches, and SE is the standard error (Standard Error).

[0048] The quantified output value of the model is taken as the data basis of statistical process control to realize the evaluation of process quality monitoring. The method can be used by production units to continuously monitor the quality of their own products, to obtain process control data and quality annual data, to better understand the quality of their own products, to realize QbD-based production of recombinant hepatitis B vaccine, to realize batch consistency evaluation, and to improve product quality. It can also be used by regulatory departments to comprehensively control the quality trend of all domestic recombinant hepatitis B vaccines. QbD-based production of recombinant hepatitis B vaccine can be realized, and product quality can be improved.

[0049] Further, SOPs for sample processing, spectrum acquisition processing and analysis and the like can be established to ensure the operability and reliability of the method in actual production or supervision, so that the method is adapted to QbD and applied to actual production or supervision.

[0050] The experimental method of the operation steps is not specifically described below, and is performed according to the corresponding product instruction. The instruments, reagents and consumables used in the examples can be purchased from commercial companies unless otherwise specified.

[0051] Example 1: Construction of a tunable Raman spectrum measurement system

[0052] The tunable Raman spectrum measurement system includes a tunable laser light source 1, a beam splitter 2, a fiber coupling mirror 4, a photodetector 7 and a Raman spectrometer 5, and further includes a reference cell 6 and a sample cell 3. The system includes two light paths, one of which passes through the reference cell 6, and the other of which passes through the sample cell 3.

[0053] The tunable laser light source emits laser light of different wavelengths, which is split into two light paths by the beam splitter 2. The laser light in one light path passes through the reference cell 6 and the first fiber coupling mirror in turn, and enters the photodetector 7. The laser light in the other light path passes through the sample cell 3 and the second fiber coupling mirror in turn, and enters the Raman spectrometer 5.

[0054] The tunable Raman spectrum spectrum acquisition parameters are as follows: excitation light wavelength tuning range 532-785 nm, Raman spectrum measurement range 3083-75 cm -1 The transmission light intensity of the empty sample cell is taken as the background.

[0055] Example 2: Characterization of the structure of the HBsAg target protein

[0056] 2.1 Collection of Raman spectrum and extraction of HBsAg target protein structure information Four different production matrix (Saccharomyces cerevisiae, CHO cell, Hansenula) of recombinant hepatitis B vaccine bulk produced by four different manufacturers were collected, each manufacturer product group included different concentrations of recombinant hepatitis B vaccine bulk (120%, 100%, 80%). The Raman spectrum information of each sample was collected by the system of example 1, the tunable Raman spectrum spectrum acquisition parameters were adjusted, the excitation light wavelength tuning range was 532-785 nm (20-50 nm step adjustable), the Raman spectrum measurement range was 3083-75 cm -1 The transmission light intensity of the empty sample pool was taken as the background.

[0057] The spectrum was collected according to the spectrum collection SOP, and the Raman spectrum data set of different manufacturers' products under different concentrations and different excitation light wavelengths was obtained.

[0058] The original spectrum data was preprocessed and principal component analysis was performed, as shown in the following table: Figure 3 The first principal component score decreases with the increase of protein concentration, so the first principal component loading matrix is used as the principal component spectrum to represent the target protein information in the hepatitis B vaccine bulk.

[0059] 2.2 Two-dimensional correlation analysis of HBsAg target protein structure The two-dimensional correlation was used to separate the HBsAg target protein structure information, and the obtained principal component spectrum was used instead of the original spectrum for spectrum dimensionality processing. The tunable Raman spectrum at different wavelengths can be expressed as , wherein represents different excitation wavelengths, is the wave number (cm -1 ). According to , the spectral features are extracted, including the following steps: (Formula 1); , wherein is the average spectrum: (Formula 2); Correlation spectral features: (Formula 3); Spectral vector features: (Formula 4) , wherein (Formula 5); Through the above dimensionality processing, the original one-dimensional spectrum can be transformed into two two-dimensional spectra, which are synchronous spectrum and asynchronous spectrum.

[0060] 2.3 Characterization of HBsAg target protein structure characteristic peaks The synchronous spectrum obtained after dimensionality processing of the Raman spectrum original detection data of samples from different manufacturers is as follows:Figure 4 As shown in the figure, the synchronous spectrum was extracted, and the common and specific spectral regions of recombinant hepatitis B vaccine stock solutions expressed on different production matrices were compared. The figure shows that the common spectral region is used for the identification of the HBsAg target protein; due to the different production matrices, the specific spectral regions of samples from different manufacturers show significant differences, and the characteristics of the specific spectral regions can be used to distinguish recombinant hepatitis B vaccine stock solutions from different production matrices.

[0061] The survey found that manufacturers 1 and 2 both used *Saccharomyces cerevisiae* as the production substrate, manufacturer 3 used CHO cells, and manufacturer 4 used *Hansenula polymorpha* cells. Table 1 summarizes the common peaks of the recombinant hepatitis B vaccine stock solution and the specific peaks of different production substrates. The common peak of the recombinant hepatitis B vaccine stock solution is 2644 cm⁻¹. -1 CHO cells and Hansenula polymorpha exhibited specific peaks, at 2440 cm⁻¹, respectively. -1 2372 cm -1 The characteristic peaks of the diagonal in the synchronous spectrum can be analyzed as characteristic signals of proteins and used to identify the type of recombinant hepatitis B vaccine stock solution.

[0062] Table 1. Comparison of Raman spectra of recombinant hepatitis B vaccine bulk matrix (process)

[0063] Example 3: Establishment of a recombinant hepatitis B vaccine identification model Fourteen batches of recombinant hepatitis B vaccine stock solutions (Saccharomyces cerevisiae, CHO, and Hansenula polymorpha) were selected. In addition, human serum albumin solution, bovine serum albumin lyophilized powder, placental polypeptide injection, recombinant human interleukin for injection, interferon eye drops, compound amino acids, and immunoglobulin injection were selected. Tunable Raman spectra were acquired according to the method in Example 2, and feature set G was obtained based on hierarchical clustering algorithm.

[0064] Calculate the feature distance between every two objects in the dataset. D and stored in an array d middle; Given two spectral sequences and Numerical information distance (set as...) D 1) The calculation formula is: (Formula 6); In the formula, They are respectively X, Y The difference in the mean, the difference in the extreme values, and the difference in the slope of the sequences; a、 b, c These are the weights corresponding to the differences, and a + b + c =1.

[0065] Construct a n × m matrix network A , where the matrix element represents the distance between two points. A Each set of adjacent elements in the matrix is called a bending path, denoted as . Find an optimal bending path that minimizes the total bending cost of the sequence A and , i.e. X Y (Formula 7). Thus, the feature distance between data clusters is defined as : D D = aD 1 +βD DTW (Formula 8). where α , β are the weights of the corresponding distances, and α + β =1.

[0066] Sort the array d in ascending order of distance values; Determine the current element in the array d . First, determine whether the two objects corresponding to the element are in the same class. If not, merge the two objects into one class. If one of the objects is already assigned to a class, merge the other object into the class. If both objects have their own classes, merge the classes into one class. Take the next element in the array and repeat step (3) until all data in the array have been processed.

[0067] According to the feature clustering results, the feature set G can be obtained, where each row represents the same class of features, and each column represents the random arrangement of class features.

[0068] Combine the feature set G with the partial least squares to establish a recombinant hepatitis B vaccine bulk identification analysis model. The identification model uses different batches and different production substrates of recombinant hepatitis B vaccine bulk as positive verification, and other protein biological products as negative verification, to obtain an identification model for identifying whether an unknown sample is a recombinant hepatitis B vaccine.

[0069] The recombinant hepatitis B vaccine bulk identification model is as follows: Figure 5 ​​As shown, based on the distribution of discriminant values ​​between the positive and negative validation samples, 0.5 is determined as the qualitative threshold. Therefore, samples with a discriminant value greater than 0.5 belong to the "target category," and samples with a discriminant value less than 0.5 belong to the "non-target category." The results of the solution type discrimination model show that, when discriminating recombinant hepatitis B vaccine stock solution in both the training and validation sets, 0 samples were misclassified based on the qualitative threshold of 0.5. This method exhibits good sensitivity, specificity, accuracy, and precision, and can correctly identify recombinant hepatitis B vaccine stock solution.

[0070] Example 4: Process Control Method 4.1 Process Control Using recombinant hepatitis B vaccine bulk solutions from manufacturers 1 and 3 as research subjects, multiple batches of products were tested separately, and discriminant values ​​were calculated using the recombinant hepatitis B vaccine production matrix identification model disclosed in Example 3. The process control of the product was evaluated by using the statistical values ​​of the discriminant values.

[0071] Statistical process control (SPC) methods are used to monitor the discriminant status of the production process for process quality control, determining whether the production process is under control to reduce product quality variation. Production process variability is characterized by the average discriminant value of different batches (…). ) and batch discrimination range ( ) to achieve, where the mean control chart is based on As the lower control limit, with As the upper control limit, The midline is calculated as shown in formulas 12 and 13. The range control chart uses... As the lower control limit, with As the upper control limit, The midline is calculated as shown in Formulas 14 and 15. Batches within the control limits meet the quality standards.

[0072] (Formula 12);

[0073] (Formula 13);

[0074] (Formula 14);

[0075] (Formula 15); in Where n is the sampling interval and n is the sampling capacity. According to the table obtained from the number of trials, This represents the mean of the ranges across different batches.

[0076] like Figure 6The quality control chart of two manufacturers of reorganization of hepatitis B vaccine bulk was shown in Table 2. The two manufacturers were in the control limit in the monitoring batch; the variability of the product of manufacturer 3 was smaller than that of the product of manufacturer 1; one batch of manufacturer 1 was located in the control limit, indicating that it had a certain potential risk, which should be investigated and confirmed in time, and laboratory test was necessary. The range control chart could better characterize the batch-to-batch variability and abnormal batch.

[0077] Table 2 Statistical process parameters

[0078] The above detailed description of the embodiments of the present application, but the content is only the preferred embodiments of the present application, can not be considered for limiting the scope of the present application. Any equivalent changes and improvements made in the scope of the present application should still belong to the scope of the present application.

Claims

1. A method for structural characterization of target protein of recombinant hepatitis B vaccine based on tunable Raman spectroscopy, characterized in that: The Raman spectrum of the recombinant hepatitis B vaccine original solution under the concentration disturbance condition is detected under the excitation of different laser wavelengths, the original spectrum is processed by two-dimensional correlation analysis method, and the two-dimensional correlation spectrum is used for in-situ characterization of the primary structure and / or high-order structure of the recombinant hepatitis B vaccine target protein solution.

2. The method according to claim 1, wherein the method is characterized in that: The specific steps are as follows: Step one: under the concentration disturbance condition, the Raman spectrum under the excitation of different wavelengths is detected by the tunable Raman spectrum measurement system, and the data cube of the recombinant hepatitis B vaccine original solution is obtained; the original spectrum is formed; Step two: the principal component analysis is performed on the original spectrum to form the principal component spectrum, and the two-dimensional correlation analysis is performed on the principal component spectrum instead of the original spectrum to obtain the synchronous spectrum and / or asynchronous spectrum; Step three: the primary structure and / or high-order structure of the recombinant hepatitis B vaccine original solution is characterized in-situ by combining the synchronous spectrum and / or asynchronous spectrum.

3. The method according to claim 2, wherein the method is characterized by: The tunable Raman spectrum measurement system comprises a tunable laser light source, a spectroscope, a fiber coupling mirror, a photoelectric detector, a Raman spectrometer, a reference cell and a sample cell; the tunable laser light source emits laser beams of different wavelengths through the spectroscope to form two light paths, in one light path, the laser passes through the reference cell and the first fiber coupling mirror into the photoelectric detector; in the other light path, the laser passes through the sample cell and the second fiber coupling mirror into the Raman spectrometer.

4. The method according to claim 2, wherein the method is characterized by: Excitation light wavelength tuning range 532-785 nm, Raman spectrum measurement range 3083-75 cm -1 Light.

5. The method according to claim 2, wherein the method is characterized by: The characteristic peak of the diagonal line in the synchronous spectrum is taken as the characteristic signal of the target protein.

6. A method for identifying the raw solution of recombinant hepatitis B vaccine, characterized in that: The recombinant hepatitis B vaccine target protein structure characterization method based on the tunable Raman spectrum in any one of claims 1-5 is used for information characterization of the vaccine original solution to be tested, so as to determine whether the vaccine original solution to be tested is the recombinant hepatitis B vaccine original solution; Or determine what kind of recombinant hepatitis B vaccine original solution the vaccine original solution to be tested is.

7. The method for identifying the raw recombinant Hepatitis B vaccine according to claim 6, characterized in that: According to the hierarchical clustering algorithm, the spectrum information after dimensionality reduction is converted into a feature set G, and a partial least squares is combined to establish a discrimination analysis model of the recombinant hepatitis B vaccine original solution.

8. The method for identifying the raw recombinant Hepatitis B vaccine according to claim 7, characterized in that: When the discrimination model of the recombinant hepatitis B vaccine is established, multiple batches of recombinant hepatitis B vaccine original solutions prepared by different processes and multiple types of other protein solutions are used for verification, the recombinant hepatitis B vaccine original solution is used as positive verification, and the other types of protein solutions are used as negative verification; When the type discrimination model of the recombinant hepatitis B vaccine is established, multiple batches of recombinant hepatitis B vaccine original solutions prepared by different processes are used for verification, one type of recombinant hepatitis B vaccine original solution prepared by a process is used as positive verification, and the other types of recombinant hepatitis B vaccine original solutions prepared by different processes are used as negative verification, to obtain the discrimination model of the recombinant hepatitis B vaccine original solution prepared by the process.

9. A method for evaluating the process control of a recombinant hepatitis B vaccine bulk preparation process, characterized by: In the method for discriminating the recombinant hepatitis B vaccine original solution according to claim 7, the discrimination value output by the discrimination model is taken as the research object, and the production process fluctuation is characterized for samples of the same production process; The group process level is characterized for samples produced by different processes.

Images