Rapid quantitative analysis method for outer vesicle proteome

Abstract

The invention provides a rapid quantitative analysis method for an outer vesicle proteome, and belongs to the technical field of quantitative analysis of the outer vesicle proteome. According to the rapid quantitative analysis method for the outer vesicle proteome, firstly, a high-quality outer vesicle sample is obtained through low-speed centrifugation, ultra-speed centrifugation and membrane filtration; and obtaining proteome data by utilizing two-dimensional electrophoresis and mass spectrometry. Then, a mathematical model equation set is constructed for preliminary analysis, a protein interaction network is optimized through an embedded game optimization index model, and feature extraction and quantitative analysis are conducted through a convolution parameter adaptive function and a residual network deep learning model; finally, a quantitative analysis result containing the protein expression level and the function importance is generated in combination with the multi-dimensional protein expression contribution index, and the technical problem that in the prior art, in the rapid quantitative analysis process of the outer vesicle proteome, the protein component complexity is high, and consequently the quantitative precision is low is solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of exosome proteome quantitative analysis, and particularly relates to a rapid exosome proteome quantitative analysis method. BACKGROUND

[0002] Exosomes are extracellular vesicles secreted by cells with lipid bilayers, which play an important role in intercellular communication, disease diagnosis and treatment. Exosome proteome quantitative analysis is a key technology for studying the function and clinical application of exosomes. Traditional exosome proteome quantitative analysis methods mainly include mass spectrometry-based label quantification technology, two-dimensional electrophoresis analysis technology and immunoblotting technology. Among them, the mass spectrometry-based label quantification technology realizes the relative quantification of proteins in different samples through isotope labeling; the two-dimensional electrophoresis analysis technology realizes separation and quantification through the difference in protein isoelectric point and molecular weight; and the immunoblotting technology realizes semi-quantitative analysis by using specific antibodies to recognize target proteins. These technologies have been widely used in exosome proteomics research, providing an important tool for revealing the biological function of exosomes.

[0003] However, the traditional exosome proteome quantitative analysis method has significant defects. First, the composition of exosome proteome is complex, and there is interaction between different proteins, so it is difficult for the traditional method to accurately reflect the mutual influence between proteins, resulting in large deviation of quantitative results. Second, there is a large amount of background noise in mass spectrometry data, which reduces the signal quality and affects the quantitative accuracy. Third, the traditional data analysis method has limited feature extraction capability for protein expression patterns, and cannot fully mine the deep information in proteomics data. In addition, the existing method lacks a systematic evaluation of the importance of protein function, and it is difficult to understand the biological significance of exosome proteome from the overall perspective.

[0004] In the face of the high complexity of exosome proteome, the traditional quantitative analysis method is difficult to realize high-throughput and high-precision protein quantification analysis at the same time. Especially when dealing with large-scale proteome data, due to the lack of effective data integration and analysis strategy, it is difficult to accurately capture the dynamic change rule of protein expression. This limitation of quantitative precision seriously restricts the in-depth development of exosome proteomics research, and it is urgent to develop new quantitative analysis methods to improve the analysis accuracy and efficiency. That is, there is a technical problem of low quantitative precision caused by high complexity of protein components in the process of rapid exosome proteome quantitative analysis in the prior art. SUMMARY

[0005] Therefore, the present application provides a rapid exosome proteome quantitative analysis method, which can solve the technical problem of low quantitative precision caused by high complexity of protein components in the process of rapid exosome proteome quantitative analysis in the prior art.

[0006] The application is implemented in the following manner: the application provides a rapid quantitative analysis method for exosome protein groups, comprising the following steps: pretreating a sample to be tested, removing cell debris by low-speed centrifugation to obtain a pretreated sample; purifying exosomes in the pretreated sample by ultracentrifugation and membrane filtration to obtain purified exosomes; extracting proteins in the purified exosomes to obtain an exosome protein sample; obtaining electrophoretic mobility, electrophoretic spot intensity and mass spectrometry data by two-dimensional gel electrophoresis and mass spectrometry analysis; constructing a mathematical model equation group for data analysis and quantification, wherein the mathematical model equation group comprises a game optimization equation, a feature extraction equation, a quantitative analysis equation and a contribution evaluation equation, and is used for analyzing protein interaction, feature extraction, quantitative analysis and contribution evaluation; constructing an embedded game optimization index model, wherein the embedded game optimization index model comprises a parameter input layer, an interaction mapping layer, a complexity evaluation layer and an optimization iteration layer, and is used for optimizing the structure of a protein interaction network; and establishing a deep learning model based on the data to perform protein expression profile analysis and generate quantitative analysis data.

[0007] In the pretreatment step, the low-speed centrifugation condition is 300 to 800 revolutions per minute, and the centrifugation time is 10 to 15 minutes; in the step of purifying exosomes by ultracentrifugation, the centrifugation condition is 100,000 to 120,000 revolutions per minute, the centrifugation temperature is 4°C, and the centrifugation time is 70 to 90 minutes; and in the membrane filtration method, a polyether sulfone membrane with a pore size of 50 to 200 nanometers is used, and the filtration pressure is controlled at 0.1 to 0.5 megapascals.

[0008] In the protein extraction step, the concentration of urea in the lysis buffer is 8 moles per liter, and the concentration of dithiothreitol is 10 millimoles per liter; and in the two-dimensional gel electrophoresis, the isoelectric focusing voltage is 500 to 2,000 volts, and the electrophoresis time is 16 to 20 hours.

[0009] In the game optimization equation, the input parameters include protein concentration, exosome extraction batch, total protein amount, extraction efficiency and sample purity, and the output parameter is the complexity index of protein interaction; and in the feature extraction equation, the input parameters include peptide segment concentration, ionization intensity, mass spectrometry scan number, background noise level and sample purity, and the output parameter is the standardized feature vector after noise reduction.

[0010] In the quantitative analysis equation, the input parameters include protein concentration, total protein amount, electrophoretic mobility, sample purity and exosome extraction batch, and the output parameter is the standardized protein expression level; and in the contribution evaluation equation, the input parameters include protein concentration, total protein amount, electrophoretic spot intensity, sample purity and exosome extraction batch, and the output parameter is the multidimensional protein expression contribution index.

[0011] The embedded game optimization index model quantifies the interaction intensity and contribution degree between proteins by constructing a multi-dimensional parameter space, and dynamically adjusts the model parameters by a stochastic gradient descent algorithm, and outputs an optimized protein interaction network structure.

[0012] The deep learning model is constructed based on a residual network, contains 4 to 6 residual modules, the module channel number is 64, 128, 256, 512 in turn, the convolution kernel size is 3*3, and a channel attention mechanism is introduced.

[0013] The convolution parameter adaptive function is further constructed to perform deep feature extraction and noise reduction processing on the standardized feature vector output by the feature extraction equation, and obtain feature data for training the deep learning model.

[0014] The quantitative analysis data table contains the expression level, relative abundance, functional importance score and biological annotation information of each protein, and the method repeatability and reproducibility are verified by calculating the coefficient of variation, and the coefficient of variation is less than 15%.

[0015] The mass spectrometry analysis adopts an electrospray ionization mode to obtain peptide segment concentration, ionization intensity, mass spectrometry scanning times and background noise level data.

[0016] Compared with the prior art, the present application provides a kind of outer vesicle proteome rapid quantitative analysis method, and the present application proposes a kind of outer vesicle proteome rapid quantitative analysis method combined with embedded game optimization index model and deep learning network.The method realizes the whole process standardization from sample preparation to data analysis by constructing a multi-level mathematical model system.Especially by introducing the embedded game optimization index model, a quantitative description framework of protein interaction is established, which effectively solves the problem that the traditional method cannot accurately reflect the interaction between proteins;At the same time, the convolution parameter adaptive function and the residual network structure are used for deep feature extraction of mass spectrometry data, which significantly improves the accuracy of signal processing.

[0017] The technical scheme of the present application solves the problems existing in the prior art through multiple innovations.Firstly, the embedded game optimization index model realizes the accurate quantification of protein network structure by constructing a multi-dimensional parameter space, which overcomes the limitation of the traditional method on the analysis ability of complex proteome.Secondly, the deep learning model based on residual network and the convolution parameter adaptive function are combined to realize intelligent noise reduction and feature extraction of mass spectrometry data, which improves the accuracy of data analysis.Thirdly, the introduction of multi-dimensional protein expression contribution index provides a systematic quantitative standard for protein functional importance evaluation.

[0018] Through these technical innovations, the present application successfully solves the technical problem of low quantitative precision caused by high protein component complexity in the rapid quantitative analysis process of the existing external vesicle proteome. The method not only accurately reflects the interaction relationship between proteins, but also effectively extracts and utilizes the deep feature information in the mass spectrometry data. Especially in the processing of complex external vesicle proteome samples, the method shows excellent analysis performance and provides reliable technical support for external vesicle proteomics research. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a flowchart of the method of the present application.

[0020] Figure 2 is a probability distribution graph of the external vesicle particle size in Example 2.

[0021] Figure 3 is a protein expression fold change graph with a confidence interval in Example 2. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical scheme and advantages of the embodiments of the present application more clear, the technical scheme in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0023] As Figure 1 shown, is a flowchart of a rapid quantitative analysis method of external vesicle proteome provided by the present application, and the method comprises the following steps:

[0024] S01, pretreating the sample to be measured, removing cell debris by low-speed centrifugation, the centrifugation condition is 300 to 800 revolutions per minute, the centrifugation time is 10 to 15 minutes, and a pretreated sample is obtained;

[0025] S02, purifying the external vesicle in the pretreated sample by ultracentrifugation, the centrifugation condition is 100000 to 120000 revolutions per minute, the centrifugation temperature is 4℃, the centrifugation time is 70 to 90 minutes, and a preliminary purified external vesicle is obtained;

[0026] S03, further purifying the preliminary purified external vesicle by membrane filtration, using a polyether sulfone membrane with a pore size of 50 to 200 nanometers, the filtration pressure is controlled at 0.1 to 0.5 megapascal, a purified external vesicle sample is obtained and the sample purity is determined;

[0027] S04, extracting proteins from the purified external vesicle sample using a lysis buffer, the urea concentration in the lysis buffer is 8 moles per liter, the dithiothreitol concentration is 10 millimoles per liter, an external vesicle protein sample is obtained and the total protein amount and protein concentration are determined;

[0028] S05, separating the outer vesicle protein sample by two-dimensional gel electrophoresis technology, the isoelectric focusing voltage is 500 to 2000 volts, the electrophoresis time is 16 to 20 hours, and the electrophoretic mobility and electrophoretic spot intensity data are obtained;

[0029] S06, identifying the outer vesicle protein sample by mass spectrometry based on isotope labeling, the mass spectrometry adopts electrospray ionization, and the peptide segment concentration, ionization intensity, mass spectrometry scanning times and background noise level data are obtained;

[0030] S07, constructing a mathematical model equation set for data analysis and quantification, the mathematical model equation set includes game optimization equation, feature extraction equation, quantitative analysis equation and contribution evaluation equation, wherein the game optimization equation inputs protein concentration, outer vesicle extraction batch, total protein amount, extraction efficiency and sample purity, and outputs protein interaction complexity index, the feature extraction equation inputs peptide segment concentration, ionization intensity, mass spectrometry scanning times, background noise level and sample purity, and outputs denoised standardized feature vector, the quantitative analysis equation inputs protein concentration, total protein amount, electrophoretic mobility, sample purity and outer vesicle extraction batch, and outputs standardized protein expression level, and the contribution evaluation equation inputs protein concentration, total protein amount, electrophoretic spot intensity, sample purity and outer vesicle extraction batch, and outputs multidimensional protein expression contribution index;

[0031] S08, constructing an embedded game optimization index model, the embedded game optimization index model includes four levels, the first level is a parameter input layer, the protein interaction complexity index of the game optimization equation is input, the second level is an interaction mapping layer, a protein network interaction matrix is constructed, the third level is a complexity evaluation layer, a network complexity index is calculated, and the fourth level is an optimization iteration layer, model parameters are dynamically adjusted by a stochastic gradient descent algorithm, and an optimized protein interaction network structure is output;

[0032] S09, constructing a convolution parameter adaptive function to perform deep feature extraction and denoising processing on the standardized feature vector output by the feature extraction equation, and obtaining feature data for training a deep learning model;

[0033] S10, constructing a deep learning model based on a residual network to perform nonlinear conversion on the standardized protein expression level data output by the quantitative analysis equation, the conversion mode is to take the logarithm of the protein concentration plus 1, 4 to 6 residual modules are designed, the module channel numbers are 64, 128, 256 and 512 in turn, the convolution kernel size is 3 by 3, a channel attention mechanism is introduced, the channel attention mechanism calculates weights through a multilayer perceptron and global pooling and is processed through a sigmoid function, and spatial attention mapping is realized;

[0034] S11, establish an outer vesicle protein group parameter and a neural network parameter mapping function, the neural network parameter mapping function converts the feature data into a high-dimensional feature space, trains the deep learning model based on the residual network through a stochastic gradient descent optimization algorithm and a cross-entropy loss function, and obtains a trained deep learning model;

[0035] S12, input the feature data of the outer vesicle protein sample into the trained deep learning model for protein expression profile analysis, combine the multi-dimensional protein expression contribution index output by the contribution evaluation equation, and generate a quantitative analysis data table containing protein expression level and functional importance;

[0036] S13, according to the optimization result of the embedded game optimization index model, the quantitative analysis data table and the multi-dimensional protein expression contribution index, establish a standardized process for outer vesicle protein group quantitative analysis, including result statistical verification and bioinformatics annotation;

[0037] S14, perform method repeatability and reproducibility verification, calculate the coefficient of variation, and output the final quantitative analysis data table when the coefficient of variation is less than 15%, the quantitative analysis data table contains the expression level, relative abundance, functional importance score and biological annotation information of each protein.

[0038] Wherein, the embedded game optimization index model quantifies the interaction intensity and contribution degree between proteins by constructing a multi-dimensional parameter space; the convolution parameter adaptive function automatically learns and adjusts the protein spectrum feature extraction parameters by using deep learning technology; the multi-dimensional protein expression contribution index evaluates the functional importance of proteins by calculating the protein concentration, protein total amount, electrophoretic mobility and sample purity parameters; the neural network parameter mapping function converts the original feature vector of the outer vesicle protein group into a high-dimensional feature space through the construction of a nonlinear mapping to realize accurate quantitative analysis; the multilayer perceptron is a basic calculation unit in neural network.

[0039] The construction process of the deep learning model based on the residual network specifically includes, first, performing nonlinear conversion on the single exosome protein group concentration data, the conversion function of which is defined as f(x) = log(protein concentration + 1), and designing a neural network with 4-6 residual modules, the channel numbers of which are 64, 128, 256, and 512 in turn, and the convolution kernel size is 3*3; at the same time, a channel attention mechanism is introduced, the weight calculation formula of which is w = sigmoid(MLP(global_pool(x))), and a spatial attention mapping is realized, a mapping function Φ(P) = Neural_Network(P) of the exosome protein group parameters and the neural network parameters is established, the model is trained through a stochastic gradient descent optimization algorithm and a cross-entropy loss function, finally the model performance index is calculated, and a model explainability module is developed to analyze the feature extraction process of the neural network on the exosome protein group, wherein the parameter mapping of the neural network refers to converting the original feature vector of the single exosome protein group into a more representative high-dimensional feature space through constructing a complex nonlinear mapping function, so as to realize the accurate quantitative analysis of protein expression.

[0040] The specific implementation of the above step S01 is to remove cell debris by low-speed centrifugation. The main purpose of this step is to pretreat the sample to be tested to improve the efficiency and quality of subsequent exosome purification. First, the sample to be tested is centrifuged at room temperature at a speed of 500 revolutions per minute for 12 minutes. The large cell debris is precipitated at the bottom of the centrifuge tube by the action of centrifugal force. Then the supernatant is carefully aspirated with a sterile pipette to avoid disturbing the precipitate, and the supernatant is transferred to a new sterile centrifuge tube. Finally, the supernatant is visually inspected to ensure that there are no visible cell debris, and the pretreated sample is obtained. This step uses the principle of centrifugal sedimentation to achieve the preliminary separation of the sample by taking advantage of the difference in sedimentation rate of different particles in the gravitational field.

[0041] The specific implementation of step S02 is to purify the exosomes in the pretreated sample by ultracentrifugation. The main purpose of this step is to separate the exosomes from the pretreated sample by ultrahigh centrifugal force. First, the pretreated sample is transferred to an ultracentrifuge tube, and the centrifugation temperature is set to 4°C, the centrifugation speed is set to 110,000 revolutions per minute, and the centrifugation time is set to 80 minutes. After centrifugation, the supernatant is discarded, and a white or light yellow precipitate is visible at the bottom of the centrifuge tube, which is the preliminary purified exosomes. Then the precipitate is gently resuspended with pre-cooled phosphate buffer to avoid air bubbles. This step is based on the theory of particle sedimentation, which uses ultrahigh centrifugal force to make the exosomes with higher density settle at the bottom of the centrifuge tube, thereby enriching and purifying the exosomes.

[0042] The specific implementation of step S03 is to further purify the preliminary purified exosome using membrane filtration method. The main purpose of this step is to improve the purity of the exosome sample through membrane filtration technology. First, a polyether sulfone membrane with a pore size of 100 nanometers is selected. The filtration device is connected to a vacuum pump, and the filtration pressure is controlled at 0.3 megapascal. Then, the resuspended exosome suspension is slowly injected into the filtration device, and the filtrate is collected. Monitor the pressure change during filtration and adjust in time to maintain a stable filtration rate. Finally, use a spectrophotometer to measure the purity of the sample, and the purity threshold should be greater than 95%. This step uses the size screening principle of membrane separation technology to remove impurity particles and improve the purity of the exosome sample.

[0043] The specific implementation of step S04 is to extract exosome proteins using lysis buffer. The main purpose of this step is to fully release the proteins in the exosome. First, prepare the lysis buffer, with urea concentration of 8 moles per liter and dithiothreitol concentration of 10 millimoles per liter. Then mix the purified exosome sample with the lysis buffer at a volume ratio of 1:3, incubate on ice for 30 minutes, and mix gently every 5 minutes during the incubation. Subsequently, use the Bradford method to determine the protein concentration, and the total protein amount threshold should be greater than 500 micrograms, and the protein concentration threshold should be greater than 1 milligram per milliliter. This step uses the synergistic effect of denaturant and reducing agent to destroy the spatial structure of proteins, achieving full dissolution and extraction of proteins.

[0044] The specific implementation of step S05 is to separate the exosome protein sample using two-dimensional gel electrophoresis technology. The main purpose of this step is to achieve high-resolution separation of proteins. First, perform isoelectric focusing by loading the protein sample onto a polyacrylamide gel strip, set the voltage to 1500 volts, and electrophorese for 18 hours. Then perform the second sodium dodecyl sulfate polyacrylamide gel electrophoresis, set the voltage to 120 volts, and electrophorese for 4 hours. Finally, use Coomassie brilliant blue staining, scan the gel to obtain the electrophoretogram, and analyze the electrophoretic mobility and spot intensity. This step is based on the differences in protein isoelectric point and molecular weight to achieve two-dimensional separation of proteins.

[0045] The specific implementation of step S06 is to identify the exosome protein sample using isotope-labeled mass spectrometry analysis method. The main purpose of this step is to obtain accurate protein qualitative and quantitative information. First, perform enzyme digestion of the protein sample by adding trypsin, with an enzyme digestion time of 16 hours and a temperature of 37°C. Then use a liquid chromatography system to separate the peptides, set the flow rate to 300 nanoliters per minute, and the gradient elution time to 90 minutes. Finally, perform mass spectrometry analysis using electrospray ionization, with a scan range of 300 to 1500 mass-to-charge ratio, to obtain mass spectrometry data. This step uses the high sensitivity and high resolution characteristics of mass spectrometry technology to achieve accurate identification of proteins.

[0046] The specific implementation of step S07 is to construct a mathematical model equation set for data analysis and quantification. The main purpose of this step is to establish a complete data processing framework. The game optimization equation uses the Nash equilibrium theory to calculate the complexity index of protein-protein interaction. The feature extraction equation uses the wavelet transform algorithm to denoise and standardize the mass spectrometry data. The quantitative analysis equation uses the multivariate linear regression model to calculate the standardized protein expression level. The contribution evaluation equation uses the principal component analysis method to evaluate the influence weight of different parameters on protein expression. Through the synergistic effect of multiple mathematical models, this step realizes the comprehensive analysis and processing of data.

[0047] The specific implementation of step S08 is to construct an embedded game optimization index model. The main purpose of this step is to optimize the protein-protein interaction network structure. The parameter input layer receives the output results of the game optimization equation, the interaction mapping layer uses the graph neural network to construct the protein interaction matrix, the complexity evaluation layer uses the entropy method to calculate the network complexity index, and the optimization iteration layer dynamically adjusts the model parameters through the stochastic gradient descent algorithm. The model training uses the mini-batch stochastic gradient descent algorithm with a batch size of 64, the learning rate is set to 0.001, and the training rounds are 100. Through multi-level optimization structure, this step realizes the dynamic optimization of protein network.

[0048] The specific implementation of step S09 is to construct a convolutional parameter adaptive function. The main purpose of this step is to realize the deep feature extraction and denoising. First, design a multi-layer convolutional neural network structure, including 3 convolutional layers and 2 pooling layers, the convolution kernel size is 3x3, and the step is 1. Then use batch normalization technology to process feature data, add funnel activation function to improve the nonlinear expression ability of the model. Finally, optimize the network parameters through the back propagation algorithm, the loss function uses mean square error, and the optimizer selects the adaptive moment estimation algorithm. Through deep learning technology, this step realizes the automatic extraction and optimization of features.

[0049] The specific implementation of step S10 is to construct a deep learning model based on residual network. The main purpose of this step is to establish an efficient protein expression quantification model. First, perform nonlinear transformation on the data, and use the logarithmic function to process the protein concentration. Then construct a neural network containing 5 residual modules, each module contains two convolutional layers and a short circuit connection. Finally, introduce the channel attention mechanism, calculate the weight through the multilayer perceptron and global pooling, and use the sigmoid function for activation. Through residual learning and attention mechanism, this step improves the expression ability and generalization performance of the model.

[0050] The specific implementation of step S11 is to establish an outer vesicle protein group parameter and neural network parameter mapping function, and the main purpose of this step is to realize effective conversion of the feature space. First, a nonlinear mapping function is constructed to map the original feature vector to a high-dimensional feature space. Then, the model parameters are optimized using the stochastic gradient descent algorithm, and the learning rate adopts the cosine annealing strategy with an initial learning rate of 0.01. Finally, the model performance is evaluated by the cross-entropy loss function, and the early stopping strategy is set to prevent overfitting. This step improves the expression ability of the model through feature space conversion.

[0051] The specific implementation of step S12 is to perform protein expression profile analysis, and the main purpose of this step is to generate comprehensive quantitative analysis results. First, the feature data is input into the trained deep learning model to obtain protein expression prediction values. Then, the output results of the contribution evaluation equation are combined to calculate the functional importance scores of each protein. Finally, a quantitative analysis data table is generated, containing information such as expression level, relative abundance, and functional score. This step realizes comprehensive analysis of protein expression through model prediction and multi-dimensional evaluation.

[0052] The specific implementation of step S13 is to establish a standardization process, and the main purpose of this step is to ensure the reliability and repeatability of the analysis results. First, data processing standards are developed based on the results of the optimized model, including data preprocessing, quality control, and result verification. Then, statistical analysis is performed using paired sample tests and variance analysis to evaluate the significance of the results. Finally, bioinformatics annotation is performed using the Gene Ontology database to classify and annotate protein functions. This step ensures the scientificity of the analysis results through standardized processes.

[0053] The specific implementation of step S14 is to perform method validation and data output, and the main purpose of this step is to ensure the reliability of the method and generate final results. First, repeatability experiments are performed to calculate the coefficients of variation for technical and biological repeats, with a coefficient of variation threshold of 15%. Then, protein expression data is normalized to eliminate the effects of batch effects. Finally, a quantitative analysis data table containing complete information is generated to ensure data traceability and repeatability. This step ensures the accuracy and reliability of the analysis results through strict quality control.

[0054] The game optimization equation is specifically represented as follows:

[0055]

[0056] In the formula, CI is the protein interaction complexity index, with a value range of 0 to 1; P ij is the concentration product of the i-th and j-th proteins; α ij is the extraction batch correlation coefficient, ranging from 0.8 to 1.2; β is the distance attenuation coefficient, with a default value of 0.1; dij is the protein spatial distance; E ij is the extraction efficiency, ranging from 0.6 to 0.9; γ is the sample purity coefficient, ranging from 0.85 to 0.98; S is the system error term, ranging from 0.01 to 0.05.

[0057] The feature extraction equation is specifically represented as follows:

[0058]

[0059] In the formula, F is the normalized feature vector; W is the weight matrix; Φ is the nonlinear mapping function; H is the peptide segment concentration matrix; I is the ionization intensity matrix; N is the scanning number vector; R is the noise reduction coefficient matrix; λ is the regularization parameter, with a default value of 0.01; w k is the kth feature weight; η is a random noise vector, following a Gaussian distribution.

[0060] The quantitative analysis equation is specifically represented as follows:

[0061]

[0062] In the formula, L is the normalized protein expression level; C i is the ith protein concentration; θ i is the concentration weight coefficient; T i is the total amount coefficient; φ i is the total amount weight; M i is the mobility coefficient; ω is the purity influence factor, ranging from 0.8 to 1; S i is the batch correction factor; δ is the system error, ranging from 0.02 to 0.08.

[0063] The contribution evaluation equation is specifically represented as follows:

[0064]

[0065] In the formula, D is the multidimensional protein expression contribution index; C i is the ith protein concentration; α i is the concentration contribution weight; T i is the total amount value; β i is the total amount contribution weight; I i is the electrophoretic spot intensity; γ i is the intensity contribution weight; P i is the purity correction coefficient; ε is the error term, ranging from 0.01 to 0.05.

[0066] The interaction mapping matrix is specifically represented as follows:

[0067]

[0068] where m ij represents the interaction strength between the i-th and j-th proteins, with a value ranging from 0 to 1; when i equals j, m ij is 0.

[0069] The channel attention weight calculation is specifically represented as follows:

[0070] w = σ(MLP(G(x))) = σ(W2ReLU(W1G(x)));

[0071] where w is the attention weight vector; σ is the sigmoid activation function; MLP is the multi-layer perceptron; G is the global pooling operation; x is the input feature; W1, W2 are the weight matrices; ReLU is the rectified linear unit activation function.

[0072] The convolution parameter adaptive function is specifically represented as follows:

[0073]

[0074] where O is the output feature map; F is the activation function; K is the convolution kernel parameter matrix; X is the input feature; represents the convolution operation; b is the bias term; A is the adaptive weight matrix; ξ is the noise term, ranging from 0.001 to 0.01.

[0075] The neural network parameter mapping function is specifically represented as follows:

[0076]

[0077] where Φ(P) is the mapped high-dimensional feature; P is the original feature vector; W i is the i-th layer weight matrix; V i is the linear transformation matrix; B i is the bias vector; σ is the nonlinear activation function; l is the network layer number; ∈ is the random disturbance term, following the standard normal distribution.

[0078] The network complexity index calculation is specifically represented as follows:

[0079]

[0080] where C is the network complexity index; p ij is the normalized interaction strength; n is the number of proteins; λ is the balance coefficient, with a default value of 0.5; H is the network entropy value.

[0081] Each equation is constructed based on the following principles:

[0082] 1. The game optimization equation uses an exponential decay term to describe the influence of spatial distance on interaction, and introduces batch coefficients and purity correction to realize data standardization;

[0083] 2. The feature extraction equation combines L1 regularization to avoid overfitting and improves feature expression ability through nonlinear mapping;

[0084] 3. The quantitative analysis equation considers the comprehensive influence of protein physicochemical properties and introduces batch correction to eliminate system error;

[0085] 4. The contribution evaluation equation uses normalization processing to ensure the comparability of different dimension indicators;

[0086] 5. The diagonal elements of the interaction mapping matrix are set to 0 to avoid autocorrelation and ensure data rationality;

[0087] 6. The channel attention weight calculation uses nonlinear activation to improve model expression ability, and global pooling is used to capture the dependence between channels.

[0088] The derivation and establishment process of the game optimization equation is as follows: based on the Nash equilibrium theory, the objective function is constructed: f(i, j) = P ij U ij , where U ij is the utility function; introduce spatial distance influence: U ij = α ij exp(-βd ij ), use exponential decay to describe distance effect; consider extraction efficiency and purity influence: E ij S term; integrate to get the final equation.

[0089] The structure of the protein concentration matrix P ij is:

[0090]

[0091] Where p ij represents the concentration product of the i th and j th protein, with units of milligrams per milliliter.

[0092] The derivation and establishment process of the feature extraction equation is as follows: construct the basic feature vector: v = [H, I, N] T ; enhance feature expression through nonlinear mapping: Φ(v); add L1 regularization term: to prevent overfitting; introduce noise reduction processing and random noise.

[0093] The structure of the peptide concentration matrix H is:

[0094]

[0095] Where h ijThis represents the concentration of the j-th peptide of the i-th protein, in moles per liter.

[0096] The derivation and establishment process of the quantitative analysis equation is as follows: Establishing a linear combination: C i θ i +T i φ i Describe the basic expression level; introduce mobility correction: M i Batch effect correction included: S i Considering systematic error: δ. The structure of the mobility coefficient vector M is: M = [m1, m2, ..., m n ] T ; where m i Let be the normalized migration rate of the i-th protein.

[0097] The derivation process of the contribution assessment equation is as follows: A multi-dimensional evaluation system is constructed, comprising three dimensions: concentration, total amount, and intensity; a weighting coefficient, α, is introduced. i ,β i γ i Purity correction was added; normalization was performed. The contribution weight vector structure is: W = [α1, α2, ..., α...]. n ] T .

[0098] Specifically, the principle of this invention is as follows: The core technical principle of this invention lies in solving the accuracy problem of quantitative analysis of extravesicular proteome by constructing a multi-level mathematical model system. First, at the sample processing level, a multi-step purification strategy is adopted to ensure the acquisition of high-quality extravesicular protein samples. Cell debris is removed by low-speed centrifugation, extravesicles are enriched by ultracentrifugation, and further purified by membrane filtration. This systematic purification process lays the foundation for subsequent quantitative analysis.

[0099] At the data analysis level, this invention constructs a mathematical model system consisting of game optimization equations, feature extraction equations, quantitative analysis equations, and contribution evaluation equations. Specifically, the game optimization equations, by establishing a multi-dimensional parameter space, transform protein-protein interactions into a quantifiable mathematical description; the feature extraction equations address noise issues in mass spectrometry data, achieving effective signal denoising and feature reconstruction; and the quantitative analysis equations and contribution evaluation equations respectively solve the problems of accurate quantification of protein expression levels and assessment of functional importance.

[0100] At the machine learning level, the application designs a deep learning model based on a residual network, cooperates with a convolution parameter self-adaptive function, and realizes deep feature extraction of proteome data. The model improves the accuracy of feature extraction through nonlinear conversion and multiple residual structures; meanwhile, a channel attention mechanism is introduced to enhance the model's ability to identify key features. This deep learning architecture can effectively process the complex data structure of exosome proteome, providing technical support for accurate quantification.

[0101] A specific embodiment 1 of the application is provided below, and the specific implementation of each step in embodiment 1 is described in detail as follows.

[0102] The specific implementation of step S01 is to first place the sample to be tested in a centrifugal tube, set the centrifugal speed to 500 revolutions per minute, and centrifuge for 12 minutes. The macromolecular impurities and cell debris are settled at the bottom of the centrifugal tube by the action of centrifugal force. This process is based on Stokes' sedimentation law, and the sedimentation speed is positively correlated with particle size, density and centrifugal force. After centrifugation, the supernatant is carefully aspirated to avoid disturbing the precipitate, and is transferred to a new sterile centrifugal tube. Visual inspection is performed to ensure that there are no visible cell debris, and a pretreated sample is obtained. The turbidity of the pretreated sample should be less than 0.1, and the pH value should be between 7.2 and 7.4. The sedimentation process can be described by the modified Stokes equation: where v is the sedimentation speed, d is the particle diameter, ρ p is the particle density, ρ f is the medium density, and η is the medium viscosity. ω is the angular velocity, and r is the radius of rotation. This step realizes the preliminary purification of the sample by low-speed centrifugation, providing a quality-qualified pretreated sample for subsequent exosome separation.

[0103] The specific implementation of step S02 is to purify the pretreated sample using an ultracentrifuge. The sample is transferred to an ultracentrifugal tube, the centrifugation temperature is set to 4°C, the speed is set to 110,000 revolutions per minute, and the centrifugation time is 80 minutes. During the ultracentrifugation process, the exosomes gradually settle under the action of centrifugal force, and the sedimentation process follows the modified Lamm equation: where F c is the centrifugal force, and m is the particle mass. After centrifugation, the supernatant is discarded, and white or light yellow precipitate is visible at the bottom of the centrifugal tube. The precipitate is resuspended using pre-cooled phosphate buffer, the buffer pH is 7.4, and the ionic strength is 150 millimoles per liter. Bubbles should be avoided during the resuspension process to prevent damage to the exosomes. The resuspension volume of the precipitate should be controlled to be 1 / 10 of the original volume to achieve the concentration of the exosomes. This step realizes the enrichment and preliminary purification of the exosomes through ultracentrifugation.

[0104] The specific implementation of step S03 is to further purify the exosome by membrane filtration method. A polyether sulfone membrane with a pore size of 100 nanometers is selected. The filtration device is connected to a vacuum pump, and the filtration pressure is controlled at 0.3 megapascal. The membrane filtration process can be described by Darcy's law: where Q is the filtration rate, K is the membrane permeability coefficient, A is the membrane area, ΔP is the pressure difference, and L is the membrane thickness. The resuspended exosome suspension is slowly injected into the filtration device, and the filtrate is collected. The pressure change during filtration is monitored, and timely adjustment is made to maintain a stable filtration rate. The purity of the sample is measured at a wavelength of 280 nanometers using a spectrophotometer, and the purity index is calculated: where A 280 and A 260 are the absorbance values at 280 nanometers and 260 nanometers, respectively, and the purity index should be greater than 1.8. This step removes impurity particles by membrane separation technology to improve the purity of the exosome sample.

[0105] The specific implementation of step S04 is to extract exosome proteins using a lysis buffer. The urea concentration in the lysis buffer is 8 moles per liter, and the dithiothreitol concentration is 10 millimoles per liter. The purified exosome sample is mixed with the lysis buffer at a volume ratio of 1:3 and incubated on ice for 30 minutes, with gentle mixing every 5 minutes during the incubation period. The protein denaturation process follows the thermodynamic equation: ΔG = ΔH - TΔS, where ΔG is the change in Gibbs free energy, ΔH is the enthalpy change, T is the temperature, and ΔS is the entropy change. The protein concentration is determined using the Bradford method, and a standard curve is established: C = k(A 595 -b), where C is the protein concentration, k is the proportionality coefficient, A 595 is the absorbance value at 595 nanometers, and b is the intercept. The total amount of protein should be greater than 500 micrograms, and the concentration should be greater than 1 milligram per milliliter. This step achieves the complete dissolution and extraction of proteins through the synergistic effect of denaturants and reducing agents.

[0106] The specific implementation of step S05 is to separate the exosome protein sample using two-dimensional gel electrophoresis technology. First, isoelectric focusing is performed, and the protein sample is loaded onto a polyacrylamide gel strip. The voltage is set to 1500 volts, and the electrophoresis time is 18 hours. The isoelectric focusing process follows the electrophoretic migration equation: where μ is the electrophoretic mobility, q is the protein charge, and r is the protein hydration radius. Then, the second sodium dodecyl sulfate polyacrylamide gel electrophoresis is performed, and the voltage is set to 120 volts. The electrophoresis time is 4 hours. The protein migration speed during electrophoresis can be described by the modified electrophoresis equation: where v is the migration speed, E is the electric field strength, and f is the friction coefficient. Coomassie brilliant blue staining is used, and the electropherogram is obtained by scanning the gel. The electrophoretic mobility and spot intensity are analyzed. The electrophoretic mobility calculation formula is: where d pDistance of protein migration, d f Distance of dye front migration. This step achieves high-resolution separation by the difference of protein isoelectric point and molecular weight.

[0107] The specific implementation of step S06 is to identify the outer vesicle protein sample by using isotope-labeled mass spectrometry analysis method. The protein sample is subjected to enzymolysis, trypsin is added, and the enzymolysis time is 16 hours and the temperature is 37°C. The enzymolysis reaction follows the Michaelis equation: Where v is the reaction rate, V max is the maximum reaction rate, [S] is the substrate concentration, and K m is the Michaelis constant. The peptide segments are separated using a liquid chromatography system, and the flow rate is set to 300 nanoliters per minute, and the gradient elution time is 90 minutes. The chromatographic separation process can be described by the Van Deemter equation: Where H is the theoretical plate height, u is the linear flow rate, and A, B, and C are constants. Mass spectrometry analysis is performed, and the scanning range is 300 to 1500 mass-to-charge ratio. The formation of charged droplets during electrospray follows the Rayleigh limit equation: Where q is the charge, γ is the surface tension, ε0 is the vacuum permittivity, and r is the droplet radius. This step achieves accurate identification of proteins through mass spectrometry technology.

[0108] The specific implementation of step S07 is to construct a system of mathematical model equations for data analysis and quantification. The game optimization equation expression is: For analyzing the interaction between proteins. The feature extraction equation expression is: To realize noise reduction and standardization of mass spectrometry data. The quantitative analysis equation expression is: To calculate the standardized protein expression level. The contribution evaluation equation expression is: To evaluate the influence weight of different parameters on protein expression. The parameters of each equation are obtained by fitting experimental data, and the least squares method is used to optimize the objective function: Where y k is the experimental observation value, f(x k , β) is the model prediction value, and β is the parameter to be optimized. This step realizes comprehensive analysis and processing of data through the synergistic effect of multiple mathematical models.

[0109] The specific implementation of step S08 is to construct an embedded game optimization index model, which includes four levels, namely the parameter input layer, the interaction mapping layer, the complexity evaluation layer, and the optimization iteration layer. The parameter input layer receives the output result of the game optimization equation, which is the protein interaction complexity index. The interaction mapping layer describes the protein network structure through the interaction mapping matrix: The complexity evaluation layer calculates the network complexity index using the entropy method: The optimization iteration layer dynamically adjusts the model parameters by the stochastic gradient descent algorithm, and the update rule is: where θ is the model parameter, α is the learning rate, and L is the loss function. The model training uses a small batch stochastic gradient descent algorithm with a batch size of 64, a learning rate of 0.001, and a training round of 100 rounds. Through multi-level optimization structure, the dynamic optimization of the protein network is realized.

[0110] The specific implementation of step S09 is to construct a convolution parameter adaptive function for deep feature extraction and noise reduction, and the expression is: First, a multi-layer convolutional neural network structure is designed, which includes 3 convolutional layers and 2 pooling layers, and the convolution kernel size is 3x3 and the step is 1. The convolution operation can be represented as: where O ij is the element of the output feature map, K mn is the convolution kernel parameter, and X i+m，j+n is the input feature. Then use the batch normalization technique to process the feature data: where μ is the mean, σ is the standard deviation, γ and β are learnable parameters. The funnel activation function is added to improve the nonlinear expression ability of the model: f(x) = max(0.01x, x). Finally, the network parameters are optimized by the back propagation algorithm, and the loss function uses the mean square error: The optimizer uses the adaptive moment estimation algorithm. Through deep learning technology, automatic feature extraction and optimization are realized.

[0111] The specific implementation of step S10 is to construct a deep learning model based on a residual network. First, the data is nonlinearly converted: f(x) = log(x + 1). A neural network containing 5 residual modules is constructed, each module containing two convolutional layers and a short circuit connection. The output of the residual module can be represented as: y = F(x, {W i}) + x, where F is the residual function, and W i is the weight parameter. The channel attention mechanism is introduced, and the weight calculation formula is: w = σ(W2ReLU(W1G(x))), where G is the global pooling operation, ReLU is the rectified linear unit activation function, and σ is the sigmoid function. The role of the attention mechanism is to learn the importance of different channels, enhance useful features, and suppress useless features. Spatial attention mapping is achieved by calculating the spatial correlation of the feature map: M s = σ(f 7×7 ([AvgPool(F); MaxPool(F)])), where f 7×7For 7x7 convolution operation, AvgPool and MaxPool are average pooling and maximum pooling operation, respectively. This step improves the expression ability and generalization performance of the model through residual learning and attention mechanism.

[0112] The specific implementation of step S11 is to establish an outer vesicle protein group parameter and neural network parameter mapping function, and its expression is: First, a nonlinear mapping function is constructed to map the original feature vector to a high-dimensional feature space. The feature vector contains information such as protein concentration, total amount, and electrophoretic mobility, which can be expressed as: P = [p1, p2, …, p n ] T Then, the random gradient descent algorithm is used to optimize the model parameters, and the learning rate adopts the cosine annealing strategy: Where η t is the current learning rate, η min and η max are the minimum and maximum learning rates, respectively, t is the current iteration number, and T is the total iteration number. Finally, the model performance is evaluated by the cross-entropy loss function: Where y ij is the true label, is the predicted value, N is the number of samples, and C is the number of classes. The early stopping strategy is set to prevent overfitting, and the training is stopped if the validation set loss does not decrease for 5 consecutive epochs. This step improves the expression ability of the model through feature space conversion.

[0113] The specific implementation of step S12 is to perform protein expression profile analysis. First, the feature data is input into the trained deep learning model to obtain the protein expression prediction value. The prediction process can be expressed as: Where f is the model function, and θ is the model parameter. Then, the output result of the contribution evaluation equation is combined to calculate the functional importance score of each protein: S i = D i · w i , where D i is the multi-dimensional protein expression contribution index, and w i is the weight coefficient. Finally, a quantitative analysis data table is generated, and the data standardization adopts the minimum and maximum normalization: This step realizes the comprehensive analysis of protein expression through model prediction and multi-dimensional evaluation.

[0114] The specific implementation of step S13 is to establish a standardization process. First, data processing standards are developed based on the results of the optimized model. Data preprocessing includes outlier detection: Where Z i is the standard score, μ is the mean, and σ is the standard deviation. Then, statistical analysis is performed using the paired sample t-test: Where is the difference mean, s d is the difference standard deviation, n is the sample size. The variance analysis uses F test: where MS B is the between-group mean square, MS W is the within-group mean square. Finally, bioinformatics annotation is performed, and the Gene Ontology database is used to classify protein functions. This step ensures the scientific nature of the analysis results through a standardized process.

[0115] The specific implementation of step S14 is method validation and data output. First, perform a repeatability experiment and calculate the coefficient of variation: The coefficient of variation threshold is set to 15%. The correlation analysis between technical repeats and biological repeats uses the Pearson correlation coefficient: Batch effect correction uses a combined linear regression model: y ij = α + β j + ∈ ij , where α is the overall mean, β j is the batch effect, and ∈ ij is the random error. Finally, generate a quantitative analysis data table containing complete information to ensure data traceability and repeatability. This step ensures the accuracy and reliability of the analysis results through strict quality control.

[0116] To better understand and implement the present application, the following provides an embodiment 2 of a specific application scenario of the present application: taking a breast cancer exosome proteome analysis conducted by researchers as an example, the specific implementation process of the exosome proteome rapid quantitative analysis method is described in detail. The research team extracted exosomes from the serum of 50 breast cancer patients and 50 healthy controls, with each sample volume being 5 milliliters.

[0117] First, the serum samples were pretreated, with the centrifugation condition set to 500 revolutions per minute and the centrifugation time set to 12 minutes to obtain the pretreated samples. The turbidity of the pretreated samples was 0.08, and the pH value was 7.3. The ultracentrifugation method was used for purification, with the centrifugation temperature set to 4°C, the rotation speed set to 110,000 revolutions per minute, and the centrifugation time set to 80 minutes. A polyether sulfone membrane with a pore size of 100 nanometers was used for filtration, and the filtration pressure was 0.3 megapascals. The main parameters of the purified samples are shown in Table 1:

[0118] Table 1 Purification effect parameter table of exosomes

[0119] Indicators Patient group mean Control group mean P value Purity index 1.92±0.15 1.89±0.13 0.276 Particle size distribution 98 ± 12 nm 95 ± 10 nm 0.158 Concentration 2.8 x 10 10 mL 1.5 x 10 10 mL 0.003 Protein content 856 ± 95 μg / mL 425 ± 68 μg / mL 0.001

[0120] For example, Figure 2As shown, the distribution of vesicle size in the patient group and the control group is illustrated, reflecting the statistical difference in particle size between the two groups. Protein was extracted using lysis buffer, and the protein concentration was measured to be 1.8 mg / mL. Protein separation was performed using two-dimensional gel electrophoresis, with an isoelectric focusing voltage of 1500 V and an electrophoresis time of 18 hours. The main parameters of the electrophoretic separation results are shown in Table 2.

[0121] Table 2. Parameters for Protein Electrophoretic Separation Effect

[0122] Parameter Value range Mean Coefficient of variation Number of protein spots 850-1200 986 12.5％ Resolution 0.15-0.25 mm 0.18 mm 8.6％ Repeatability 0.85-0.92 0.89 5.8％ Dynamic range 10 3 -10 5 ]] 10 4 ]] 13.2％

[0123] Mass spectrometry analysis was performed using electrospray ionization. The raw data were processed and differentially expressed proteins were identified. The main analytical results are shown in Table 3.

[0124] Table 3. Results of Differentially Expressed Protein Proteomic Analysis

[0125] Category Up-regulated proteins Down-regulated proteins Total number Significance level P < 0.01 128 95 223 Significance level P < 0.05 256 187 443 Fold change > 2 86 64 150 Fold change > 1.5 195 146 341

[0126] A set of mathematical model equations was constructed for data analysis. The characteristics of the protein-protein interaction network obtained by game optimization equation calculation are shown in Table 4.

[0127] Table 4. Characteristics of Protein-Protein Interaction Networks

[0128] Network features Value Number of network nodes 443 Number of network edges 2856 Average degree 12.9 Clustering coefficient 0.386 Network density 0.029 Characteristic path length 3.25

[0129] Based on the analysis results of the embedded game optimization index model, the core functional proteins and their characteristic parameters were selected as shown in Table 5:

[0130] Table 5. Characteristic parameters of core functional proteins

[0131] Protein ID Interaction complexity index Fold change in expression Functional importance score P001 0.856 3.25 0.912 P002 0.823 2.86 0.875 P003 0.798 2.65 0.856 P004 0.765 2.42 0.823 P005 0.734 2.18 0.798

[0132] like Figure 3 As shown, the fold change in expression of core functional proteins is visually displayed, highlighting the expression characteristics of key proteins. A deep learning model was constructed for protein expression profiling analysis, and the model performance evaluation results are shown in Table 6.

[0133] Table 6 Performance Evaluation Table for Deep Learning Models

[0134]

[0135]

[0136] The quantitative analysis result of the final output is verified by a methodology, with a coefficient of variation of 12.8% and a correlation coefficient of 0.896. The method successfully identifies 443 differentially expressed proteins, of which 15 are identified as potential biomarkers. Traditional exovesicle proteome analysis methods mainly rely on conventional mass spectrometry technology and a single data analysis method, and have problems such as low analysis efficiency and low accuracy. The present application introduces an embedded game optimization index model and deep learning technology to establish a complete analysis process, significantly improving the analysis efficiency and accuracy.

[0137] It should be noted that the variables involved in the present application are explained in detail as shown in Table 7.

[0138] Table 7: Variable explanation table

[0139]

[0140]

[0141] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for rapid quantitative analysis of an exovesicle proteome, characterized by, The method comprises the following steps: Preprocessing the sample to be tested by removing cell debris through low-speed centrifugation to obtain a pretreated sample; Purifying exovesicles in the pretreated sample by ultracentrifugation and membrane filtration to obtain purified exovesicles; Extracting proteins in the purified exovesicles to obtain an exovesicle protein sample; Obtaining electrophoretic mobility, electrophoretic spot intensity and mass spectrometry data by two-dimensional gel electrophoresis and mass spectrometry analysis; and constructing a mathematical model equation set for data analysis and quantification, wherein the mathematical model equation set comprises a game optimization equation, a feature extraction equation, a quantitative analysis equation and a contribution evaluation equation, and is used for analyzing protein interaction, feature extraction, quantitative analysis and contribution evaluation; Constructing an embedded game optimization index model, wherein the embedded game optimization index model comprises a parameter input layer, an interaction mapping layer, a complexity evaluation layer and an optimization iteration layer, and is used for optimizing the structure of a protein interaction network; Based on the data, a deep learning model is established to analyze the protein expression profile and generate quantitative analysis data.

2. The method of claim 1, wherein the method is a rapid quantitative analysis method of exovesicle proteins. In the preprocessing step, the low-speed centrifugation condition is 300 to 800 revolutions per minute, and the centrifugation time is 10 to 15 minutes; in the step of purifying exovesicles by ultracentrifugation, the centrifugation condition is 100,000 to 120,000 revolutions per minute, the centrifugation temperature is 4°C, and the centrifugation time is 70 to 90 minutes; in the membrane filtration method, a polyether sulfone membrane with a pore size of 50 to 200 nanometers is used, and the filtration pressure is controlled at 0.1 to 0.5 megapascals.

3. The method of claim 1, wherein the method is a rapid quantitative analysis method of exovesicle proteins. In the protein extraction step, the concentration of urea in the lysis buffer is 8 moles per liter, and the concentration of dithiothreitol is 10 millimoles per liter; the isoelectric focusing voltage of the two-dimensional gel electrophoresis is 500 to 2,000 volts, and the electrophoresis time is 16 to 20 hours.

4. The method of claim 1, wherein the exovesicle proteome is rapidly quantified. The game optimization equation inputs protein concentration, exovesicle extraction batch, total protein amount, extraction efficiency and sample purity, and outputs protein interaction complexity index; the feature extraction equation inputs peptide concentration, ionization intensity, mass spectrometry scan number, background noise level and sample purity, and outputs denoised standardized feature vector.

5. The method of claim 1, wherein the method is a rapid exosome proteome quantification method. The quantitative analysis equation inputs protein concentration, total protein amount, electrophoretic mobility, sample purity and exovesicle extraction batch, and outputs standardized protein expression level; the contribution evaluation equation inputs protein concentration, total protein amount, electrophoretic spot intensity, sample purity and exovesicle extraction batch, and outputs multidimensional protein expression contribution index.

6. The method of claim 1, wherein the exovesicle proteome is rapidly quantified. The embedded game optimization index model quantifies the interaction intensity and contribution degree between proteins by constructing a multidimensional parameter space, and dynamically adjusts the model parameters by a stochastic gradient descent algorithm, and outputs the optimized protein interaction network structure.

7. The method of claim 1, wherein the method is a rapid exosome proteome quantification method. The deep learning model is constructed based on a residual network, contains 4 to 6 residual modules, the module channel numbers are 64, 128, 256 and 512 in sequence, the convolution kernel size is 3 by 3, and a channel attention mechanism is introduced, wherein the channel attention mechanism calculates weights through a multilayer perceptron and global pooling and processes them through a sigmoid function.

8. The method of claim 1, wherein the exovesicle proteome is quantified rapidly. The normalized feature vector output by the feature extraction equation is subjected to deep feature extraction and noise reduction processing by constructing a convolution parameter adaptive function, to obtain feature data for training the deep learning model.

9. The method of claim 1, wherein the exovesicle proteome is quantified rapidly. The quantitative analysis data table contains the expression level, relative abundance, functional importance score and biological annotation information of each protein, and the method repeatability and reproducibility are verified by calculating the coefficient of variation, and the coefficient of variation is less than 15%.

10. The method of claim 1, wherein the method is a rapid quantitative analysis method of exovesicle proteins. The mass spectrometry analysis adopts an electrospray ionization mode to obtain peptide segment concentration, ionization intensity, mass spectrometry scanning times and background noise level data.

Images