Human blood protein structure in-situ characterization method and application thereof in identification

Abstract

The invention relates to a human blood protein structure in-situ characterization method and application thereof in identification. The method comprises the following steps: firstly, carrying out in-situ characterization on a primary structure and an advanced structure of highly homologous protein by utilizing a polarization tuning spectrum and a spectrum dimension raising feature extraction method; the dimension of a common absorption spectrum is raised to a high-dimensional space, and the tiny change characteristics in the spectrum are enhanced. And carrying out a fusion method based on feature similarity on the high-dimensional spectrum, designing a feature screening method based on multi-similarity measurement criterion fusion, effectively screening typical features in the high-dimensional spectrum, and realizing in-situ characterization of a to-be-detected sample. On the basis of an in-situ characterization result, source verification can be carried out on a to-be-detected sample, in-situ characterization of first-level and high-level structures can be achieved while good specificity and accuracy are achieved, and the method has great significance in guaranteeing the safety, effectiveness and quality controllability of the quality of a human serum albumin product.

Description

Technical Field

[0001] The present invention belongs to the field of detection technology, and in particular relates to a method for in-situ characterization of human blood protein structure and its application in identification. Background Art

[0002] Human serum albumin (HSA) is used as a clinical drug, primarily for treating shock caused by hemorrhagic trauma and burns, alleviating cerebral edema or intracranial hypertension caused by injury, improving edema or ascites caused by cirrhosis and kidney disease, and preventing and treating hypoproteinemia. As a specialized drug required for clinical emergency care, human serum albumin is in short supply and expensive, making it a valuable drug, and counterfeiting is common in the market. Bovine serum albumin (BSA), a highly homologous protein in the pharmaceutical industry, has a similar molecular weight, similar physiological functions, and is very affordable. The two proteins are highly homologous, with a sequence similarity of 67%. While BSA plays an important role in biochemical research, genetic engineering, and pharmaceutical research, and is also used as a pharmaceutical health food and condiment, it cannot be used as a clinical drug. Therefore, the identification of highly homologous albumins is crucial for the safety, efficacy, and quality control of human serum albumin preparations. In situ characterization of the structure of human serum albumin in solution is crucial not only for the quality evaluation of HSA but also for the research of HSA-related drugs.

[0003] Currently, the main methods for identifying human albumin include microsequencing (N-terminal Edman degradation), mass spectrometry combined with other techniques, amino acidome analysis, immunodouble diffusion, and immunoelectrophoresis. Microsequencing offers strong specificity and high accuracy for identifying HSA and BSA, but is time-consuming and costly. Liquid chromatography-mass spectrometry (LC-MS / MS) is challenging for identifying highly homologous albumins and offers limited accuracy. Amino acid composition analysis may reveal amino acid variations due to insufficient acid hydrolysis or partial degradation, and requires confirmation in conjunction with other protein properties. The highly homologous properties of HSA and BSA are very similar, making accurate identification difficult. While infrared spectroscopy and Raman spectroscopy can rapidly distinguish human albumin from other proteins, they are insufficient for qualitative analysis of highly homologous HSA and BSA. Furthermore, immunodouble diffusion and immunoelectrophoresis can distinguish finished human albumin from other animal albumins, such as horse, bovine, and sheep blood, but they cannot characterize the structure of finished human albumin or further evaluate product quality.

[0004] The quality management of human serum albumin in China is mainly based on Chinese Pharmacopoeia. The production and testing of human serum albumin are clearly specified in Chinese Pharmacopoeia 2020. Among them, the production of human serum albumin is required to control the process, which is divided into three levels: whole process quality control, batch consistency control, and target component and non-target component control, that is, effective control of the whole production process of human serum albumin from raw materials, intermediate products, and products to ensure the quality of the final product. In the field of drug research and drug supervision, a specific identification and in-situ characterization method for human serum albumin is needed. SUMMARY

[0005] To solve the problems of complex sample preparation process, long analysis time, limited detection parameters, low sensitivity and other shortcomings of the commonly used instrument analysis methods such as mass spectrometry and chromatography; the traditional spectral method has good detection effect for molecular bond absorption, but the characterization ability for protein and other macromolecular structures is poor, the analysis method has low sensitivity and large interference, etc. The present application provides a human serum protein structure in-situ characterization method and its application in identification.

[0006] The technical scheme adopted by the present application is: a protein structure in-situ characterization method, detecting the polarization tuning spectrum of protein solution under the condition of temperature and concentration double disturbance, processing the obtained original spectrum by two-dimensional correlation analysis method, and using multi-dimensional slice spectrum to in-situ characterize the primary structure and / or high-order structure of protein solution.

[0007] Preferably, the specific steps are as follows:

[0008] Step one: under the condition of temperature and concentration double disturbance, the near-infrared spectrum is collected by a polarization tuning spectrum multi-modal spectrum detection device to obtain a data cube of albumin;

[0009] Step two: forming an original spectrum from the data cube; extracting features of the original spectrum at different concentrations at each temperature point by principal component analysis, extracting k mutually unrelated orthogonal components to form a principal component spectrum;

[0010] Step three: performing two-dimensional correlation analysis on the principal component spectrum to separate protein structure signals;

[0011] Step four: in-situ characterization of the primary structure and / or high-order structure of the protein solution by combining the synchronous spectrum and / or asynchronous spectrum obtained by two-dimensional correlation analysis.

[0012] Preferably, in step one, the polarization tuning spectrum multi-modal spectrum detection device comprises a light source, a polarizer, an analyzer, a beam splitter, a fiber coupling mirror, an optical switch device and a spectrometer, and further comprises a reference cell and a sample cell; the polarizer and the beam splitter are sequentially arranged downstream of the light source, the wide spectrum light emitted by the light source passes through the polarizer and is split into two paths under the action of the beam splitter, one of the two paths passes through the reference cell and the first fiber coupling mirror in sequence and is transmitted into the optical switch device by an optical fiber, and the other path passes through the sample cell, the analyzer and the second fiber coupling mirror in sequence and is transmitted into the optical switch device by an optical fiber; the two light signals are controlled by the optical switch and are detected in the spectrometer in time.

[0013] Preferably, the concentration disturbance condition is that 3-6 concentration detection points with gradient increase are set in 3.5%-5.1%; and the temperature disturbance condition is that 3-6 temperature detection points with gradient increase are set in 57-61℃.

[0014] Preferably, in step two, the original data is subjected to second-order 15-point SG filtering denoising and baseline correction to form the original spectrum.

[0015] Preferably, in step three, the two-dimensional correlation analysis process is as follows:

[0016] Analysis of spectrum change function under the action of external disturbance variable , wherein t represents the external disturbance variable temperature (℃), is the wave number (cm -1 ). When t changes between and , the measured spectrum intensity y, i.e. the original spectrum set, is defined as :

[0017] (formula 1);

[0018] , wherein is the average spectrum:

[0019] (formula 2);

[0020] Two-dimensional correlation synchronous spectrum intensity is the vector product of the dynamic spectrum intensity y at different wave numbers :

[0021] (formula 3);

[0022] Two-dimensional correlation asynchronous spectrum intensity is the Hilbert-Noda matrix (H) vector product of the dynamic spectrum intensity y at different wave numbers :

[0023] ​(Formula 4);

[0024] wherein, As shown in Formula 5:

[0025] (Formula 5).

[0026] Preferably, the autocorrelation peak of the diagonal line in the synchronous spectrum is analyzed as a characteristic signal of the protein, for identification and structural characterization.

[0027] Preferably, in step one, the near-infrared spectrum and the Raman spectrum are collected by the polarization-tuned multi-modal spectral detection device under the condition of temperature-concentration double perturbation, to obtain the data cube of the albumin.

[0028] Preferably, the heterospectral two-dimensional correlation method is used to analyze the correlation between the sum spectrum region of the near-infrared spectrum and the characteristic spectrum region of the Raman spectrum, and the structural signal of the albumin is analyzed by the heterospectral two-dimensional correlation spectrum.

[0029] Preferably, the orthogonal cross-peak in the heterospectral two-dimensional correlation synchronous spectrum is used to calculate the correlation coefficient r of the orthogonal cross-peak (Formula 6), which represents the common response of the corresponding spectrum segments of the near-infrared spectrum and the Raman spectrum to the concentration perturbation and the temperature perturbation.

[0030] (Formula 6);

[0031] wherein is the peak value of the orthogonal cross-peak, max is the maximum value of all peak values of the orthogonal cross-peak, and min is the minimum value of all peak values of the orthogonal cross-peak; the larger the value of r is, the stronger the correlation between the near-infrared spectrum segment and the Raman spectrum segment is, and the greater the probability of coming from the same structure is;

[0032] Preferably, r<0.3 is defined as weak correlation, r<0.6 is defined as moderate correlation, and r≥0.6 is defined as strong correlation.

[0033] A human blood protein identification method, in which the protein structure in situ characterization method is used to characterize an unknown protein solution, and the obtained information is compared with the information of a standard human blood protein sample to identify the unknown protein solution.

[0034] Preferably, according to the two-dimensional correlation diagram, the structural information and the corresponding spectrum segment, the autocorrelation peak or the slice spectrum of the protein is extracted, a HSA qualitative discrimination model is established by using the partial least squares discrimination method (PLS-DA), and the unknown protein type is identified by the model.

[0035] The advantages and positive effects of the present invention are: organically combining principal component analysis, two-dimensional correlation spectroscopy, and partial least squares discriminant, comprehensively solving the difficulty of in situ characterization of human albumin structure (especially higher-order conformations) at physiological concentrations from the three aspects of structural signal extraction, structural signal separation, and structural signal analysis, and providing an effective technical means for studying the interaction between human albumin and drugs;

[0036] A feature similarity-based fusion method is used for high-dimensional spectra, and the fused features are used to establish a protein type identification model. While having good specificity and accuracy in identifying HSA, it can also achieve in situ characterization of the primary and higher-order structures of HSA solutions at physiological concentrations. This method can be used in the identification of human serum albumin, significantly shortening the identification cycle and enabling efficient, rapid and accurate identification, which is of great significance for ensuring the safety, effectiveness and quality controllability of human serum albumin products. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 Schematic diagram of the structure of a polarization-tuned multimodal spectroscopy detection device; wherein: 1. light source, 2. polarizer, 3. beam splitter, 4. sample cell, 5. analyzer, 6. fiber coupling mirror, 7. reference cell, 8. optical switch, 9. spectrometer;

[0038] Figure 2 Flowchart of the in situ characterization method for human blood protein structure;

[0039] Figure 3 Average raw near-infrared spectra of HSA solution and BSA solution under temperature perturbation; a: HSA solution; b: 4900-4700 cm -1 Band magnification; c: BSA solution; d: BSA solution 4900-4700 cm -1 Band zoom image;

[0040] Figure 4 The near-infrared principal component spectra (4900-4200 cm) of protein solutions of all concentrations were analyzed under temperature perturbation. -1 ) Spectra of the first principal component of principal component analysis; a: HSA solution; b: BSA solution;

[0041] Figure 5 Protein characteristic signal extraction under temperature perturbation 4800-4300 cm -1 Two-dimensional correlation analysis results; a: synchronous spectrum of HSA solution; b: asynchronous spectrum of HSA solution; c: synchronous spectrum of BSA solution; d: asynchronous spectrum of BSA solution;

[0042] Figure 6 Two-dimensional correlation analysis of the autocorrelation peak of the synchronization spectrum;

[0043] Figure 7 Results of near-infrared two-dimensional correlation analysis under polarization modulation spectroscopy perturbation; a: synchronous spectrum of HSA solution; b: synchronous spectrum of BSA solution;

[0044] Figure 8 Raman 2D correlation analysis results under a single temperature perturbation; a: 1700-1600 cm -1 Synchronous spectrum of HSA solution; b: 1500-1400 cm -1 Synchronous spectrum of HSA solution; c: 1700-1600 cm -1 Synchronous spectrum of BSA solution; d: 1500-1400 cm -1 Synchronous spectrum of BSA solution;

[0045] Figure 9 Raman-NIR spectroscopy heterogeneity correlation analysis results under temperature perturbation; ac: HSA solution; df: BSA solution;

[0046] Figure 10 Validation results of the HSA solution identification model. DETAILED DESCRIPTION

[0047] The embodiments of the present invention are described below with reference to the accompanying drawings.

[0048] The present invention relates to a method for in situ characterization of human blood protein structure and its application in identification. First, the primary and higher-order structures of highly homologous human and bovine albumin are characterized in situ using polarization-tuned spectroscopy and a spectral dimension-upgrading feature extraction method. A multimodal spectral detection device based on near-infrared polarization-tuned spectroscopy is designed. The near-infrared polarization-tuned spectroscopy detection device is used to obtain absorption spectra of samples in different polarization states. Based on the multimodal spectroscopy, a spectral dimension-upgrading method is provided to upgrade ordinary absorption spectra to a high-dimensional space, enhancing the characteristics of subtle changes in the spectra. A feature similarity-based fusion method is then performed on the high-dimensional spectra, and a feature screening method is designed that fuses multiple similarity metrics to effectively screen typical features in the high-dimensional spectra, enabling in situ characterization of the sample to be tested. Based on the in situ characterization results, the sample to be tested can be rapidly identified. The fused features are used to establish a protein species identification model. The identification model uses a PLS-DA method based on a nonlinear kernel function to effectively extract nonlinear features in the high-dimensional spectra, thereby improving the model's ability to identify isomeric proteins. This method can be used to identify protein species or sources.

[0049] The structure of the polarization-tuned spectroscopy multimodal spectroscopy detection device is as follows: Figure 1As shown, the broad spectrum light emitted by the light source passes through a polarizer and is split into two paths by a beamsplitter. One path is the reference path, which passes through a reference cell, a fiber coupling mirror, and then an optical switch via an optical fiber. The other path is the sample path, which passes through a sample cell, an analyzer fiber coupling mirror, and then an optical switch via an optical fiber. The two light signals are controlled by the optical switch and enter the spectrometer in time-sharing mode for separate spectrum detection. By rotating the positions of the polarizer and analyzer, polarization-tuned spectra can be obtained. By leveraging the sensitivity of polarized light to molecular bond structure, polarization spectra of protein macromolecules in different polarization states can be obtained, providing rich information for protein structural analysis and classification. During detection, the reference spectrum is the unmodulated state. The polarizer and analyzer in the sample path are set to different angles, and N polarization states are selected for measurement. The angle difference a is set as one deflection unit, and the polarization states in states a, 2a, 3a, …, and Na are detected, resulting in multiple polarization-tuned spectra that are recorded separately.

[0050] Conventional absorption spectroscopy cannot effectively characterize highly homologous protein structures, especially higher-order conformations in solution. Polarization-tuned spectroscopy (PTS) utilizes the sensitivity of polarized light to molecular bond structure to obtain polarization spectra of protein macromolecules under different polarization states, providing rich information for protein structural analysis and classification.

[0051] Using a polarization-tuned multimodal spectroscopy device, near-infrared spectroscopic data of an albumin solution were collected under concentration and temperature perturbations, and a data cube was constructed. The obtained spectra were preprocessed. A second-order 15-point SG filter was applied to the raw spectra for denoising and baseline correction. Principal component analysis (PCA) was used to extract features from the raw spectra for each polarization state. K orthogonal (uncorrelated) principal component spectra were extracted, preserving the polarization information related to the protein structure and removing other interfering signals to obtain a raw spectrum containing a series of spectral signatures. Two-dimensional correlation was used to separate the albumin structural information, and the principal component spectra were used to replace the raw spectra for two-dimensional correlation analysis. The raw spectra were subjected to dimensionality-enhancing processing, and the resulting synchronous and asynchronous spectra were used to analyze the dynamics of albumin structural changes with temperature. The positive cross-peaks of the synchronous spectra represent the direction of the albumin dynamics, while the asynchronous spectra represent the rate of the albumin dynamics with temperature.

[0052] In certain embodiments of the present invention, when this method is used to distinguish between human serum albumin and bovine serum albumin, an autocorrelation peak of a synchronous spectrum is first established for a known type of albumin solution. This autocorrelation peak of the synchronous spectrum is then used as a reference standard spectrum to detect an unknown sample. The obtained autocorrelation peak of the synchronous spectrum is then compared with the reference standard spectrum to determine whether it is human serum albumin.

[0053] When it is necessary to distinguish between highly homologous human blood albumin and bovine blood albumin, first, a data cube of human blood albumin and bovine blood albumin is established by a polarization tuning spectrum multi-modal spectrum detection device. In some embodiments of the present application, a concentration disturbance condition can be set, such as a normal physiological environment human serum albumin concentration range of 35-51 g / L, i.e. 3.5%-5.1% (w / v, g / 100ml). In a pathological condition, five concentration points of 1%, 2%, 3%, 4%, and 5% in a concentration range of 1%-5% are set as the concentration disturbance condition. Spectra are collected by the polarization tuning spectrum multi-modal spectrum detection device under the concentration disturbance condition to obtain a data cube of albumin. Near-infrared spectrum determination conditions: wavelength range: 10000-4000 cm -1 , transmission sampling mode, scanning number: 32 (background / sample), resolution: 4 cm -1 .

[0054] Second-order 15-point SG filtering denoising and baseline correction are performed on the original data to realize spectrum pretreatment and form an original spectrum; a human blood albumin and bovine blood albumin characteristic spectral segment can be selected as 4900-4200 cm -1 . Principal component analysis is used to extract features from the original spectrum of each temperature point at five concentrations to extract k orthogonal (mutually unrelated) principal component spectra, and a series of spectral characteristic signals are obtained by removing other interference signals while retaining albumin information. The principal component number k is determined according to the cumulative representation rate and the principal component score. The principal component spectrum can represent the main structural information of albumin and is the basis for identifying albumin species.

[0055] Further, a temperature disturbance condition can be added to supplement the data cube of albumin for separating albumin structural information. The temperature disturbance condition: the denaturation temperature of albumin is usually about 65°C, and a temperature range of 57-61°C is set with a change gradient of 1°C. The temperature change process includes a warming process and a cooling process, and each temperature point is balanced for 5 minutes. Spectra are collected by the polarization tuning spectrum multi-modal spectrum detection device under the double disturbance condition to obtain a data cube of albumin. Near-infrared spectrum determination conditions: wavelength range: 10000-4000 cm -1 , transmission sampling mode, scanning number: 32 (background / sample), resolution: 4 cm -1 .

[0056] Based on the data cube, the albumin structural signal is separated by two-dimensional correlation spectroscopy. Two-dimensional correlation is used to separate albumin structural information, and principal component spectra are used instead of original spectra for two-dimensional correlation analysis. The calculation method and formula are as follows:

[0057] Analysis of the function of spectral changes under the action of an external disturbance variable , wherein t represents the external disturbance variable temperature (°C), is the wave number (cm-1 ). When t and When the measured spectral intensity y, that is, the original spectrum set, is defined as :

[0058] (Formula 1);

[0059] in, is the average spectrum:

[0060] (Formula 2);

[0061] Two-dimensional correlation synchronization spectral intensity For different wave numbers The vector product of the dynamic spectral intensity y is:

[0062] (Formula 3);

[0063] Two-dimensional correlated asynchronous spectral intensity For different wave numbers The Hilbert-Noda matrix of the dynamic spectral intensity y is ( ) Vector product:

[0064] (Formula 4);

[0065] in, As shown in Formula 5:

[0066] (Formula 5);

[0067] The synchronous and asynchronous spectra in two-dimensional correlation were used to analyze the characteristics of the dynamic process of albumin structure changes with temperature, which was used to distinguish the highly homologous HSA and BSA. The positive cross-peaks of the synchronous spectrum represent the direction of albumin dynamic changes, while the asynchronous spectrum represents the rate of albumin dynamic changes with temperature.

[0068] The dual-perturbation 2D correlation spectroscopy analysis method described above solved the problem of distinguishing the highly homologous HSA and BSA. 2D correlation spectroscopy typically uses a single perturbation condition. Single-temperature perturbation 2D correlation spectra contain little information and are ineffective in characterizing highly homologous protein structures, particularly higher-order conformations in solution. However, adding concentration perturbation significantly increases the amount of information about albumin structure in the data cube, ultimately enabling the differentiation of the highly homologous HSA and BSA based on this wealth of available information.

[0069] Further, the method can be further used for albumin structure signal analysis. Under the conditions of temperature and concentration perturbation, the molecular dynamics of HSA and BSA are obtained by combining near-infrared and Raman spectroscopy methods, and the correlation analysis of the near-infrared combination spectrum region and the characteristic spectrum region (amide I, II and III bands (1700-1200 cm -1 )) of Raman spectrum is performed by using the heterogeneous spectrum two-dimensional correlation method, so as to analyze the spectrum band difficult to be attributed to the near-infrared combination region and the primary structure or high-level conformation of the separated characteristic signal.

[0070] Raman spectrum determination conditions: excitation light source 785 nm, fiber probe sampling mode, long focal length (5 mm), spectral measurement range 3083-75 cm -1 , exposure time 1 min, accumulation number 4 times, resolution 4 cm -1 . Spectrum acquisition: the near-infrared spectrum and the Raman spectrum of each concentration point in the concentration perturbation condition are collected respectively under the temperature perturbation condition, and the optical path is 1 mm. The spectral intensity, signal-to-noise ratio, maximum and minimum noise in a certain spectral range are checked to ensure the accuracy and reliability of the spectrum for identification and structure characterization.

[0071] The original spectrum is pretreated, the characteristic extraction of the original spectrum is performed by using principal component analysis, the series of spectral characteristic signals are obtained by removing other interference signals by retaining the albumin information, and then the albumin structure signal is separated by using two-dimensional correlation spectroscopy; the two-dimensional correlation is used for albumin structure information separation, the principal component spectrum is used to replace the original spectrum for two-dimensional correlation analysis, the calculation method is the same as that in the foregoing steps, and the synchronous spectrum and the asynchronous spectrum are obtained.

[0072] The albumin structure signal is analyzed by using the heterogeneous spectrum two-dimensional correlation method, the spectrum band difficult to be attributed to the near-infrared combination region and the primary structure or high-level conformation of the characteristic signal are analyzed, the corresponding relationship between the structure information and the corresponding spectrum segment is obtained. The orthogonal cross peak in the synchronous spectrum of the heterogeneous spectrum two-dimensional correlation represents the common response of the corresponding spectrum segments of the near-infrared spectrum and the Raman spectrum to the concentration perturbation and the temperature perturbation, and the correlation coefficient r (formula 6) of the orthogonal cross peak is calculated.

[0073] (Formula 6);

[0074] In the formula, the peak value of the orthogonal cross peak, max is the maximum value of all peak values of the orthogonal cross peaks, and min is the minimum value of all peak values of the orthogonal cross peaks. The larger the r value is, the stronger the correlation between the near-infrared spectrum segment and the Raman spectrum segment is, and the greater the probability of coming from the same structure is. The r value less than 0.3 is defined as weak correlation, the r value less than 0.6 is defined as medium correlation, and the r value greater than or equal to 0.6 is defined as strong correlation.

[0075] Based on the above content, a human albumin qualitative discrimination model can be established according to the features in the human albumin spectrum information to determine the qualitative properties of unknown samples.

[0076] Based on the two-dimensional correlation maps and the correspondence between albumin structure and spectral segments, autocorrelation peaks or slice spectra that can distinguish highly homologous proteins were extracted. A qualitative discriminant model for HSA was established using the partial least squares discriminant analysis (PLS-DA). The model was validated using forward and reverse validation.

[0077] The positive validation uses the finished product of human albumin injection that has passed the statutory inspection, and the reverse validation uses highly homologous bovine albumin solution, other protein products and compound amino acids.

[0078] Using the true positive rate ( )(Formula 17), True Negative Rate ( ) (Formula 18) evaluates the model sensitivity and specificity, and the model accuracy ( )See Formula 19.

[0079] (Formula 17);

[0080] (Formula 18);

[0081] (Formula 19);

[0082] Where, is the number of true positives, is the number of true negatives, The total number of positive The total number of negative is the number of true negative samples, is the number of false negative samples.

[0083] The aforementioned characterization and identification methods address the high cost, cumbersome operation, and inability to characterize structures of existing protein-based biomacromolecule analysis and identification technologies. By establishing a discrimination model, rapid, batch, and accurate identification and classification of two proteins are possible. Using tuned polarization spectroscopy, spectra in different polarization states are acquired, and spectral dimension-upgrading methods are used to correlate absorption spectra in different polarization states. This allows for the analysis of the primary structure or higher-order conformation of difficult-to-attribute spectral bands and characteristic signals in the near-infrared sum frequency region.

[0084] By organically combining polarization-tuned spectroscopy, principal component analysis, spectral dimension upgrading, and feature fusion and discrimination, this method comprehensively addresses the challenges of in situ characterization of human albumin structure (especially higher-order conformations) at physiological concentrations, focusing on structural signal extraction, separation, and interpretation. This method provides an effective technical approach for studying drug interactions between human albumin and other drugs. The PLS-DA qualitative identification model, developed based on these effective characterization results, exhibits excellent specificity, sensitivity, and accuracy, and is of significant significance for the quality assessment of human albumin. In certain embodiments of the present invention, this characterization and identification method can also be extended to the characterization and analysis of other highly homologous protein structures.

[0085] The present invention is described below with reference to the accompanying drawings. Experimental methods without specific operating steps are carried out in accordance with the corresponding product specifications. Unless otherwise specified, the instruments, reagents, and consumables used in the examples can be purchased from commercial companies.

[0086] Example 1: Construction of a polarization-tuned spectroscopy measurement device

[0087] like Figure 1 As shown, a polarization-tuned spectroscopy measurement device is constructed. It includes a light source 1, a polarizer 2, an analyzer 5, a beam splitter 3, a fiber coupler 6, an optical switch 8, and a spectrometer 9. A reference cell 7 and a sample cell 4 are provided. The sample cell 4 is used to store samples to be measured, while the reference cell 7 can store samples in different states to improve spectral measurement accuracy. Polarizer 2 and beam splitter 3 are sequentially provided downstream of light source 1. Broadband light emitted by light source 1 passes through polarizer 2 and is split into two paths by beam splitter 3. One path passes through reference cell 7 and a first fiber coupler, and is then transmitted to optical switch 8 via an optical fiber. The other path passes through sample cell 4, analyzer 5, and a second fiber coupler, and is then transmitted to optical switch 8 via an optical fiber. The two optical signals are controlled by optical switch 8 and enter spectrometer 9 in a time-sharing manner for spectral detection.

[0088] The sample used is a normal physiological human serum albumin solution with a concentration range of 35-51 g / L (i.e., 3.5%-5.1% (w / v, g / 100 ml)). The reference spectrum is for the non-polarization modulation condition, and the polarization state is adjusted by the angle difference between the polarizer and analyzer. Polarization tuning spectra are measured under different polarization states in the wavelength range of 10,000-4,000 cm -1 , resolution 4cm -1 .

[0089] Comparative Example 1:

[0090] The near infrared spectra of human serum albumin HSA and bovine serum albumin BSA solution samples were detected under different temperature conditions. The results are as follows Figure 3As shown in Figures 3a and 3c, after simple spectral preprocessing of the original spectra, HSA and BSA have similar peak shapes and cannot be distinguished.

[0091] The near-infrared spectrum of HSA and BSA in the range of 4900-4700 cm -1 Spectrum amplification, such as Figure 3 b and Figure 3 As shown in d, the temperature variation of the two samples is consistent. As the temperature rises, the wavelength of 4900-4700 cm -1 The absorbance of the spectrum in the band decreases.

[0092] Example 2: Characterization of human and bovine serum albumin by near-infrared two-dimensional correlation spectroscopy

[0093] 2.1 Construction of human serum albumin and bovine serum albumin data cubes

[0094] The polarization-tuned spectroscopy measurement device constructed in Example 1 was used to detect samples of human serum albumin and bovine serum albumin solutions. Double perturbation conditions were set to perform concentration perturbation and temperature perturbation respectively. Concentration perturbation conditions: the concentration range of human serum albumin in a normal physiological environment is 35~51g / L, i.e. 3.5%~5.1% (w / v, g / 100ml). In the case of pathological conditions, five concentration points of 1%, 2%, 3%, 4%, and 5% in the concentration range of 1%~5% were set as concentration perturbation conditions. Temperature perturbation conditions: the denaturation temperature of albumin is usually around 65°C. The temperature range is set to 57~61°C, with a gradient of 1°C. The temperature change process includes a heating process and a cooling process, and each temperature point is balanced for 5 minutes. Spectra were collected under double perturbation conditions to obtain the data cube of albumin.

[0095] Near infrared spectroscopy measurement conditions: Wavelength range: 10000-4000cm -1 , transmission sampling mode, 32 scans (background / sample), resolution 4cm -1 .

[0096] The angles of the sample path polarizer and analyzer are set to differ by 15° as a polarization state. In this embodiment, polarization states of 15°, 30°, 45°, 60°, 75°, 90°, 105°, 120°, 135°, 150°, 165°, and 180° are selected respectively. The polarization tuning spectra are recorded as S n (λ), wavelength range: 10000-4000cm -1 , resolution 4cm -1 .

[0097] 2.2 Extraction of albumin characteristic structure information from the data cube

[0098] The original spectrum was subjected to second-order 15-point SG filtering for denoising and baseline correction, with the wavelength of 4900-4200 cm -1 Albumin features were extracted from the spectral segments. Principal component analysis (PCA) was used to extract features from the original spectra at each temperature point at five concentrations. K orthogonal (uncorrelated) principal component spectra were extracted. A series of spectral signatures was obtained by retaining albumin information and removing other interfering signals. The number of principal components, k, was determined based on the cumulative representativeness and principal component score.

[0099] The first principal component spectrum was used to replace the original near-infrared spectrum to perform a two-dimensional correlation analysis of the temperature disturbance. The results are shown in Figure 2. Figure 4 As shown, the spectral resolution is Figure 3 The information content of albumin is much higher, and the autocorrelation peaks and cross peaks of HSA and BSA are significantly different, which indicates that the kinetic processes of structural changes of highly homologous HSA and BSA with increasing temperature are different, which can be used as the main basis for distinguishing the two.

[0100] 2.3 Separation of albumin structural signals by two-dimensional correlation spectroscopy

[0101] Two-dimensional correlation is used to separate the albumin structural information. The main component spectrum obtained replaces the original spectrum for spectrum dimension-upgrading. The main component spectra of different polarization states are cross-correlated with the reference spectrum. The calculation method and formula for dimension-upgrading are as follows:

[0102] Function of spectral change , where t represents the spectral polarization state, is the wave number (cm -1 ). When t and When the measured spectral intensity y, i.e. the principal component spectrum set, changes between :

[0103] (Formula 1);

[0104] in, is the average spectrum:

[0105] (Formula 2);

[0106] Two-dimensional correlation synchronization spectral intensity For different wave numbers The vector product of the dynamic spectral intensity y is:

[0107] (Formula 3);

[0108] Two-dimensional correlated asynchronous spectral intensity For different wave numbers The Hilbert-Noda matrix of the dynamic spectral intensity y is ( ) Vector product:

[0109] (Formula 4);

[0110] in, As shown in Formula 5:

[0111] (Formula 5);

[0112] The aforementioned dimensionality increase method transforms the original one-dimensional spectrum into two two-dimensional spectra, defined as synchronous and asynchronous spectra, respectively. These spectra are then used to analyze the dynamics of albumin structure changes with temperature and to distinguish between highly homologous HSA and BSA. The positive cross-peaks of the synchronous spectrum represent the direction of albumin dynamics, while the asynchronous spectrum represents the rate of albumin dynamics change with temperature.

[0113] Will Figure 4 After the content is upgraded, the Figure 5 The synchronous spectrum and asynchronous spectrum are shown in Figure 2. Figure 5 As shown in the synchronous spectra of (a) and (c), the positive cross peak positions of HSA and BSA are significantly different: the positive cross peak of human serum albumin is at (4320 cm -1 , 4436 cm -1 ), (4512 cm -1 , 4752 cm -1 ), while the positive cross peak of bovine serum albumin is at (4344 cm -1 , 4592 cm -1 ), (4344 cm -1 , 4436 cm -1 ), (4436 cm -1 , 4592 cm -1 ) place. Figure 5 As shown in the asynchronous spectra (b) and (d), human albumin has a -1 , 4436 cm -1 )、(4356 cm -1 , 4512 cm -1 ) and a negative cross peak at (4436 cm -1 , 4512 cm -1 )、(4512cm -1 , 4752 cm -1 ) and a positive cross peak appeared at (4392 cm -1 , 4592 cm -1 )、(4436 cm -1 , 4592 cm-1 )、(4548 cm -1 , 4592 cm -1 ) and a negative cross peak at (4592 cm -1 , 4776 cm -1 ), indicating that the kinetic change rates of HSA and BSA are also different.

[0114] extract Figure 5 The autocorrelation peak of the diagonal of the mid-sync spectrum is used as a characteristic signal to distinguish highly homologous albumin proteins for identification and structural characterization. Figure 6 As shown in the figure, HSA and BSA are significantly different. The characteristic signal of HSA is 4752 cm -1 、4708 cm -1 、4512 cm -1 4436 cm -1 4356 cm -1 、4320 cm -1 The characteristic signal of BSA is 4776 cm -1 、4592 cm -1 4548cm -1 4436 cm -1 、4392 cm -1 4344 cm -1 .

[0115] Comparative Example 2:

[0116] Human serum albumin and bovine serum albumin samples were tested according to the method of Example 2, and near-infrared two-dimensional correlation spectra were obtained. The difference was that in step 2.1, only single polarization tuning spectroscopy was used to obtain near-infrared two-dimensional correlation spectra. The results were as follows: Figure 7 As shown. In addition, only a single temperature perturbation is used to perform Raman 2D correlation analysis. The results are shown in Figure 8 shown.

[0117] Depend on Figure 7 and Figure 8 It can be seen that due to the wide spectral peak broadening and low resolution, it cannot be used for identification and structural characterization. This spectral band is covered by the strong absorption of water and cannot be used directly.

[0118] Example 3: Analysis of albumin structural signals by heterogeneous mass spectrometry two-dimensional correlation method

[0119] The molecular dynamics of HSA and BSA were simultaneously obtained by combining near-infrared and Raman spectroscopy under the conditions of temperature and concentration dual perturbation. The near-infrared composite spectrum region was compared with the Raman spectral characteristic spectrum region (the bands of amide I, II, and III (1700-1200 cm) by using the heterogeneous mass spectrometry two-dimensional correlation method.-1 )) perform correlation analysis to resolve the primary structure or higher-level conformation of the spectral bands that are difficult to attribute in the near-infrared composite frequency region and the separated characteristic signals.

[0120] Raman spectroscopy conditions: excitation light source 785 nm, fiber optic probe sampling mode, long focal length (5 mm), spectral measurement range 3083-75 cm -1 , exposure time 1 min, accumulation times 4 times, resolution 4 cm -1 Spectral Acquisition: Near-infrared and Raman spectra were collected for each sample at each concentration point under the concentration perturbation conditions, using a 1mm optical pathlength. Spectral quality checks were performed using parameters such as spectral intensity, signal-to-noise ratio, and maximum and minimum noise within a specified spectral range to ensure the accuracy and reliability of the spectra for identification and structural characterization.

[0121] Raman-NIR heterogeneity spectrometry Figure 9 The positive cross peaks in the two-dimensional correlation synchronization spectrum of the heterogeneous spectrum are used to represent the common response of the corresponding spectral bands of the near-infrared spectrum and the Raman spectrum to concentration perturbations and temperature perturbations, and the correlation coefficient r of the positive cross peaks is calculated (Formula 6).

[0122] (Formula 6);

[0123] In the formula is the peak value of the positive cross peak, max is the maximum value among all positive cross peaks, and min is the minimum value among all positive cross peaks. The larger the r value, the stronger the correlation between the near-infrared and Raman spectra, and the greater the probability that they come from the same structure. r < 0.3 is defined as weak correlation, r < 0.6 as moderate correlation, and r ≥ 0.6 as strong correlation.

[0124] The structural analysis results and peak correlation coefficients of the positive cross peaks are shown in Tables 1 and 2. The experimental results show that although the distribution of the cross peaks of HSA and BSA heterogeneous spectra partially overlap, the structural analysis results are significantly different. - The spiral signal accounts for a high proportion, and the near-infrared characteristic signal is at 4356 cm -1 4664 cm -1 There is a strong correlation between -Fold and - The corner signal accounts for a small proportion, at 4512 cm -1 、4708 cm -1 The correlations are moderate; -The helical signal also accounts for a high proportion, at 4548 cm -1 4664 cm -1 There is a moderate correlation. -Folding signal accounts for a small proportion, at 4392 cm -1 There is a moderate correlation at 4548 cm -1 There is weak correlation.

[0125] Table 1 Structural and conformational attribution of near-infrared characteristic spectral regions of human albumin solution

[0126]

[0127] Table 2 Structural and conformational attribution of the near-infrared characteristic spectrum of bovine albumin solution

[0128]

[0129] Example 4: HSA qualitative discrimination model

[0130] 4.1 Establishment of HSA qualitative discrimination model

[0131] according to Figure 5 、 Figure 6 and Figure 9 The two-dimensional correlation diagram shown in Figure 1, as well as the structural information and corresponding spectral segments in Table 1, were used to extract autocorrelation peaks or slice spectra that can distinguish highly homologous proteins. A qualitative discriminant model for HSA was established using the partial least squares discriminant method (PLS-DA).

[0132] To extract high-dimensional spectral features from synchronous and asynchronous spectra, high-dimensional spectral feature parameters are calculated separately. Homogeneity reflects the proximity of the grayscale value to the diagonal; energy reflects the uniformity of the image grayscale distribution and the coarseness of the texture; contrast reflects the clarity of the image and the depth of the texture; and correlation reflects the similarity of the image in rows or columns. The specific calculation formula is as follows:

[0133] (Formula 7);

[0134] (Formula 8);

[0135] (Formula 9);

[0136] (Formula 10);

[0137] in, , , , Represents the probability statistics of pixels with gray levels i and j appearing simultaneously in the corresponding texture image.

[0138] In order to further screen effective features, six similarity measurement criteria, including Euclidean distance, Mahalanobis distance, cosine distance, spectral gradient angle, Pearson correlation coefficient, and spectral information divergence, are selected from different angles such as distance, angle, shape, and information entropy for weighted fusion. The fusion formula is shown in Formula 11.

[0139] (Formula 11);

[0140] Where, is the fusion similarity value, They represent the similarity metrics of Euclidean distance, Mahalanobis distance, cosine distance, spectral gradient angle, Pearson correlation coefficient, and spectral information divergence, respectively. It is the fusion weighting coefficient of each similarity measurement value, which is used to fuse and weight the similarity measurement methods from different angles such as distance, angle, shape, and information entropy, so as to more comprehensively measure the correlation between the characteristics of the data sample to be tested and the historical data sample, screen out effective feature variables, and improve the accuracy of discrimination.

[0141] Establish a qualitative discrimination model for human albumin. According to the characteristics of high-dimensional spectrum, a fusion similarity weighted kernel partial least squares discrimination method is designed. First, the input data is transformed into Mapping to high-dimensional feature space , and then get the kernel feature matrix Then, the corresponding sample weights of the test sample and each training sample in the high-dimensional feature space are measured based on the fusion similarity , the weights corresponding to each training sample constitute a weight matrix After weighting the samples in the high-dimensional feature space through the weight matrix, the PLS-DA regression algorithm is used to establish a discriminant model to estimate the type of the current sample to be tested. .

[0142] According to the PLS algorithm of nonlinear kernel function, a local weighted regression model is established for the test samples in the following steps:

[0143] make , ,initialization , the number of principal components , convergence conditions and the maximum number of iterations (You can set Equal to the output variable , use the concentration value as the initial value for both);

[0144] calculate The score vector , and normalize:

[0145] (Formula 12);

[0146] If the convergence condition is met or the maximum number of iterations is reached , then calculate the residual, let , And calculate the residual:

[0147] (Formula 13);

[0148] (Formula 14);

[0149] judge Is it equal to If so, stop the iteration, the algorithm ends, and the input The principal component score matrix , then the regression coefficient matrix is:

[0150] (Formula 15);

[0151] The predicted output value of the sample to be tested is calculated as shown in Formula 16:

[0152] (Formula 16).

[0153] 4.2 Verification of the HSA qualitative discrimination model

[0154] Using the true positive rate ( )(Formula 17), True Negative Rate ( ) (Formula 18) evaluates the model sensitivity and specificity, and the model accuracy ( )See Formula 19.

[0155] (Formula 17);

[0156] (Formula 18);

[0157] (Formula 19);

[0158] Where, is the number of true positives, is the number of true negatives, The total number of positive The total number of negative is the number of true negative samples, is the number of false negative samples.

[0159] The model was validated using forward validation and reverse validation. Forward validation used a finished product of human albumin injection that passed statutory inspection, and reverse validation used highly homologous bovine albumin solution, other protein products, and compound amino acids, specifically ovalbumin solution, amino acid solution, placenta polypeptide solution, interleukin injection, interferon spray, and immunoglobulin injection. The data cube was measured and its category was determined according to the steps of Example 2 and Example 3, respectively, to detect the sensitivity and specificity of the HSA qualitative discrimination model. The model validation results are shown in Figure 2. Figure 10 As shown in Table 3, it can be seen that human albumin solution samples can be accurately classified using this model, which fully demonstrates the sensitivity, specificity and accuracy of this identification method.

[0160] Table 3 Analysis results of HSA solution identification model

[0161]

[0162] Example 5: Application of in situ characterization of human blood protein structure in source identification

[0163] The unknown solution sample was characterized according to the method of Example 2 and Example 3, and the near infrared spectra under different temperature and different polarization state perturbation conditions were measured. The principal component analysis of the spectrum was performed to extract the signal of albumin in the solution. Figure 6 The autocorrelation peaks of the two-dimensional correlation analysis synchronization spectrum shown in the figure are compared, and the HSA qualitative discrimination model is used to determine whether the unknown solution sample is human serum albumin.

[0164] The embodiments of the present invention are described in detail above, but the contents described are only preferred embodiments of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A method for in situ characterization of protein structure, characterized by: The polarization-tuned spectrum of the protein solution under the dual perturbation conditions of temperature and concentration is detected, the original spectrum obtained is upgraded by two-dimensional correlation analysis method, and the primary structure and / or higher-order structure of the protein solution are characterized in situ using multi-dimensional slice spectrum.

2. The method for in situ characterization of protein structure according to claim 1, characterized in that: The specific steps are as follows: Step 1: Under the conditions of temperature and concentration dual perturbations, a near-infrared spectrum is collected using a polarization-tuned spectroscopy multimodal spectral detection device to obtain a data cube for albumin; Step 2: Cube the data to form the original spectrum; use principal component analysis to extract features from the original spectrum of different concentrations at each temperature point, and extract k mutually unrelated orthogonal principal component spectra; Step 3: Perform two-dimensional correlation analysis using principal component spectra to separate protein structural signals; Step 4: In situ characterization of the primary structure and / or higher-order structure of the protein solution using synchronous and / or asynchronous spectra obtained by two-dimensional correlation analysis.

3. The method for in situ characterization of protein structure according to claim 2, characterized in that: In step one, the polarization-tuned spectroscopy multimodal spectroscopy detection device includes a light source, a polarizer, an analyzer, a spectroscope, a fiber coupling mirror, an optical switch device and a spectrometer, and also includes a reference cell and a sample cell; a polarizer and a spectroscope are sequentially arranged downstream of the light source, and the wide-spectrum light emitted by the light source passes through the polarizer and is divided into two paths under the action of the spectroscope, one of which passes through the reference cell and the first fiber coupling mirror in sequence, and is transmitted into the optical switch device by the optical fiber; the other passes through the sample cell, the analyzer, the second fiber coupling mirror in sequence, and is transmitted into the optical switch device by the optical fiber; the two optical signals are controlled by the optical switch and enter the spectrometer in time-sharing to detect the spectrum.

4. The method for in situ characterization of protein structure according to claim 2, wherein: The concentration perturbation condition is to set 3-6 concentration detection points with increasing gradients between 3.5% and 5.1%; the temperature perturbation condition is to set 3-6 temperature detection points with increasing gradients between 57 and 61°C.

5. The method for in situ characterization of protein structure according to claim 2, characterized in that: In step 2, the raw data were subjected to second-order 15-point SG filtering for denoising and baseline correction to form the raw spectrum.

6. The method for in situ characterization of protein structure according to claim 2, wherein: The diagonal autocorrelation peak in the synchronous spectrum was analyzed as the characteristic signal of the protein.

7. The method for in situ characterization of protein structure according to any one of claims 2 to 6, characterized in that: In step 1, near-infrared spectra and Raman spectra are collected separately using a polarization-tuned spectroscopy multimodal spectroscopy detection device under the conditions of temperature and concentration dual perturbations to obtain the data cube of albumin.

8. The method for in situ characterization of protein structure according to claim 7, characterized in that: The heterogeneous mass spectrometry two-dimensional correlation method was used to perform correlation analysis between the near-infrared composite spectrum region and the Raman spectrum characteristic spectrum region, and the albumin structure signal was analyzed through the heterogeneous mass spectrometry two-dimensional correlation spectrum.

9. A method for identifying human blood proteins, characterized in that: The unknown protein solution is characterized by the in situ protein structure characterization method described in any one of claims 1 to 8, and the obtained information is compared with the standard human blood protein sample information to achieve identification of the unknown protein solution.

10. The identification method according to claim 9, characterized in that: Based on the two-dimensional correlation diagram, structural information and corresponding spectral segments, the autocorrelation peaks or slice spectra of the protein are extracted, and the HSA qualitative discrimination model is established using the partial least squares discriminant method to identify the type of unknown protein.

Images