A method for production quality control of medical mytilus mucus nasal spray

Abstract

This invention discloses a production quality control method for medical mussel adhesive protein nasal spray, belonging to the field of medical preparation production and quality management technology. The method acquires measured data on the physicochemical parameters of each process in historical batches, as well as the molecular aggregation state spectrum and oxidative crosslinking activity state of the finished product, and constructs a two-way control sample set corresponding to each batch. A multivariate gradient coupling fitting algorithm is used to construct a mapping model between physicochemical parameters and functional states. Real-time physicochemical parameters of the current production batch are input, and a hidden state time-series extrapolation algorithm is used to output dynamic evolution data and determine the quality level. Data from multiple batches are integrated to form a batch quality traceability chain. For batches with abnormal deviations, the production process is adjusted and the extrapolation is repeated until the functional state deviation converges. This invention achieves continuous monitoring and dynamic quality management of the production process of medical mussel adhesive protein nasal spray by mapping physicochemical parameters to functional states and combining time-series extrapolation and batch traceability.

Description

Technical Field

[0001] This invention relates to the field of medical preparation production and quality management technology, specifically to a production quality control method for a medical mussel adhesive protein nasal spray. Background Technology

[0002] Medical mussel adhesive protein nasal spray uses mussel adhesive protein as its core active ingredient. Its clinical efficacy depends on its ability to form a film on the nasal mucosa and promote mucosal repair. This ability is determined by the molecular aggregation spectrum and oxidative cross-linking activity of mussel adhesive protein, rather than simply by the physicochemical properties of the solution such as viscosity, protein concentration, or pH value. In industrial production, from raw material dissolution, solution preparation, pH adjustment, filtration to filling, mussel adhesive protein molecules are constantly undergoing a dynamic evolution process of spontaneous oxidative cross-linking. The migration of its molecular aggregation state over time directly affects the functional consistency of the product within and between batches.

[0003] For the quality control of the production of such bioactive preparations, existing technologies have developed various methods based on data acquisition and information management. These include establishing a correlation between process parameters and quality inspection results through electronic batch records to achieve full-process traceability; using statistical process control techniques to monitor trends in key quality attributes; and employing process analysis techniques such as near-infrared spectroscopy and Raman spectroscopy for online monitoring. However, the data acquisition and analysis frameworks of the above methods are mainly based on the internal space of physicochemical parameters, lacking cross-spatial correlations between physicochemical parameters and functional states, and failing to establish a mechanism for inferring unmeasurable functional states from physicochemical data.

[0004] The limitations of existing technologies include at least the following issues: The quality assessment of current medical mussel adhesive protein nasal sprays relies on threshold values ​​for physicochemical parameters such as viscosity, protein concentration, and pH. These parameters only characterize the basic physicochemical properties of the solution, while the actual film-forming retention and mucosal repair functions of the product in the nasal cavity depend on the molecular aggregation state spectrum and oxidative cross-linking activity of the mussel adhesive protein. The physicochemical parameter space and the functional state space belong to heterogeneous characterization systems lacking direct translation pathways, and existing technologies have not established a mapping model between the two. In actual production, physicochemical parameters being within acceptable ranges does not necessarily mean that the molecular aggregation state has not shifted at the functional level. Existing quality control systems lack an information correlation mechanism to infer the evolution trend of unmeasurable functional states from directly measurable physicochemical data. Quality evaluation remains at the level of physicochemical compliance, failing to extend to the data-driven characterization dimension of functional consistency. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a production quality control method for medical mussel adhesive protein nasal spray, which solves the problems in the existing production quality control of medical mussel adhesive protein nasal spray, namely, the lack of mapping relationship between physicochemical parameters and functional state, and the lack of an information association mechanism for inferring unmeasurable functional state from measurable physicochemical data.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a production quality control method for medical mussel adhesive protein nasal spray, comprising the following steps: obtaining physicochemical parameters of raw material dissolution, solution preparation, pH adjustment, filtration, and filling processes in multiple historical production batches, as well as measured data of molecular aggregation state spectrum and oxidative cross-linking activity state corresponding to each batch of finished product; constructing a two-way control sample set by batch-wise mapping of physicochemical parameters and measured data of molecular aggregation state spectrum and oxidative cross-linking activity state; using a multivariate gradient coupling fitting algorithm to perform correlation calculations on the two-way control sample set to construct a mapping relationship model between the physicochemical parameter space and the functional state space, wherein the functional state space is composed of the molecular aggregation state spectrum and oxidative cross-linking activity state; obtaining the current production... The real-time physicochemical parameters of each batch at each process node are input into the mapping relationship model. Using the hidden state time-series deduction algorithm, the dynamic evolution data of the molecular aggregation state spectrum and oxidative crosslinking activity state of the current production batch are output. The dynamic evolution data of the molecular aggregation state spectrum and oxidative crosslinking activity state are compared with the functional state offset judgment threshold to determine the quality level of the current production batch. The physicochemical parameters and dynamic evolution data of the molecular aggregation state spectrum and oxidative crosslinking activity state of multiple batches are integrated to form a batch quality traceability data chain. The state difference characteristics are extracted and the production process operating conditions are adjusted accordingly. The mapping relationship model construction, deduction and quality level judgment are repeated for the adjusted production process until the functional state offset of multiple batches converges within the functional state offset judgment threshold.

[0007] Furthermore, the specific steps for constructing a two-way control sample set are as follows: Physicochemical parameters of raw material dissolution, solution preparation, pH adjustment, filtration, and filling processes from multiple historical production batches are screened; measured data of molecular aggregation state profiles and oxidative crosslinking activity states corresponding to the finished products of each historical production batch are collected, and batch numbers are simultaneously retained; physicochemical parameters are matched with measured data of molecular aggregation state profiles and oxidative crosslinking activity states according to batch numbers, binding is completed, and secondary verification is performed to remove control content that deviates from the normal range; all screened and bound data are arranged in production order to form a two-way control sample set.

[0008] Furthermore, the specific steps for constructing the mapping relationship model using the multivariate gradient coupling fitting algorithm are as follows: The physicochemical parameters of each batch in the bidirectional control sample set are used as input vectors, and the measured data of the molecular aggregation state spectrum and oxidative crosslinking activity state corresponding to that batch are used as output vectors; a multivariate coupling objective function containing first-order terms, second-order terms, and interactive coupling terms of the physicochemical parameters is constructed; the weights of each physicochemical parameter term are iteratively updated along the gradient direction of the multivariate coupling objective function; the fitting residual of the multivariate coupling objective function after weight update is calculated in each iteration; the iteration stops when the fitting residual is lower than the preset convergence threshold, and the set of physicochemical parameter weights at this time is fixed as the model parameters of the mapping relationship model.

[0009] Furthermore, the specific steps for outputting dynamic evolution data using the hidden state time-series extrapolation algorithm are as follows: The real-time physicochemical parameters of the current production batch at each process node are arranged in the process sequence and used as the input mapping relationship model to obtain preliminary functional state extrapolation data for each process node; the preliminary functional state extrapolation data of each process node are used as the initial values ​​of the hidden state sequence, establishing the state transition probability matrix between adjacent nodes on the process time axis, and the observation probability matrix between each node in the hidden state sequence and the corresponding node's physicochemical parameters in the observation sequence; starting from the first process node, recursive estimation is performed node by node along the process time axis, using the hidden state estimate of the previous process node as the prior information of the current process node, and Bayesian correction of the hidden state estimate based on the measured physicochemical parameters of the current process node, outputting the corrected dynamic evolution data of the molecular aggregation state spectrum and oxidative crosslinking activity state of each process node.

[0010] Further, the steps for determining the functional state deviation threshold are as follows: Collect measured data of molecular aggregate state profiles and oxidative crosslinking activity states corresponding to historical batches whose functional states have been clinically verified as qualified; extract the distribution characteristic parameters of molecular aggregate state profiles and the activity characteristic parameters of oxidative crosslinking activity states to form a qualified sample set; use the mean of the distribution characteristic parameters of molecular aggregate state profiles in the qualified sample set as the center and the standard deviation of a preset multiple as the boundary to define the distribution threshold range of molecular aggregate state profiles; define the activity threshold range of oxidative crosslinking activity states in the same way; and use the above threshold range as the functional state deviation threshold.

[0011] Furthermore, the specific steps for determining the quality level of the current production batch are as follows: extract the dynamic evolution data of the molecular aggregate state spectrum and the oxidative crosslinking active state of each process node; calculate the degree of deviation of the distribution characteristic parameters of the molecular aggregate state spectrum of each process node from the distribution threshold range of the functional state offset determination threshold, and the degree of deviation of the active characteristic parameters of the oxidative crosslinking active state from the active threshold range; statistically analyze the proportion of process nodes with deviation in each level interval to determine the quality level of the current production batch; when the distribution characteristic parameters of the molecular aggregate state spectrum exceed the distribution threshold range or the active characteristic parameters of the oxidative crosslinking active state exceed the active threshold range, mark the offset direction.

[0012] Furthermore, the specific steps for forming a batch quality traceability data chain are as follows: integrate the real-time physicochemical parameters, molecular aggregation state spectrum, and dynamic evolution data of oxidative crosslinking activity state of the current batch, as well as the quality grade; associate the batch number with the basic production information, generate an independent identification code, and store it in a fixed manner; connect the independent identification codes of each batch in sequence according to the production order, build a batch quality traceability data chain, and establish a query index.

[0013] Further, the specific steps for extracting state difference characteristics are as follows: Arrange the dynamic evolution data of the molecular aggregate state spectrum and oxidative crosslinking active state of multiple batches in the batch quality traceability data chain according to the process sequence; split the dynamic evolution data layer by layer, compare the distribution characteristic parameters of the molecular aggregate state spectrum with the overall average distribution characteristic parameters of multiple batches, and the activity characteristic parameters of the oxidative crosslinking active state with the overall average activity characteristic parameters of multiple batches, and screen out abnormal batches that deviate from the preset difference threshold; classify the deviation data of abnormal batches, and mark the process node to which the deviation belongs and the corresponding physicochemical parameter category.

[0014] Furthermore, the specific steps for adjusting the production process operating conditions are as follows: The dynamic evolution data of the molecular aggregate state spectrum and oxidative crosslinking active state in the multi-batch quality traceability data chain are compared with the functional state deviation judgment threshold to identify key process nodes where the dynamic evolution data continuously deviates from the functional state deviation judgment threshold; the physicochemical parameter items corresponding to the key process nodes in the reverse mapping relationship model and their weight coefficients in the multivariate coupling objective function are retrieved to determine the target physicochemical parameter that contributes the most to the deviation of the molecular aggregate state spectrum and oxidative crosslinking active state; based on the gradient direction of the target physicochemical parameter in the mapping relationship model, the adjustment direction and adjustment amount of the target physicochemical parameter are determined, and the production process operating conditions of the key process nodes are adjusted accordingly.

[0015] Furthermore, the specific steps for determining the convergence of functional state offsets across multiple batches are as follows: Collect the molecular aggregated state distribution characteristic parameters and the oxidative crosslinking active state activity characteristic parameters of consecutive production batches; calculate the distance between the molecular aggregated state distribution characteristic parameters and the center of the distribution threshold in the functional state offset determination threshold, and the distance between the oxidative crosslinking active state activity characteristic parameters and the center of the activity threshold for each batch; when the above distances for multiple consecutive batches are all within the functional state offset determination threshold range, and the fluctuation range of the distance between adjacent batches is lower than the preset convergence determination threshold, the functional state offsets of multiple batches are determined to have converged; if the convergence condition is not met, repeat the mapping relationship model construction, dynamic evolution data deduction of molecular aggregated state and oxidative crosslinking active state, and quality level determination, and continue to adjust the production process operating conditions.

[0016] The present invention has the following beneficial effects:

[0017] (1) The production quality control method of this medical mussel adhesive protein nasal spray is to integrate the physicochemical parameters of the raw material dissolution, liquid preparation, pH adjustment, filtration and filling processes in the historical production batches with the measured data of the molecular aggregation state spectrum and oxidative cross-linking activity state of each batch of finished products to construct a two-way control sample set, and to establish a mapping model between the physicochemical parameter space and the functional state space by using a multivariate gradient coupling fitting algorithm. In the process of model construction, the physicochemical parameters are used as input vectors and the molecular aggregation state and oxidative cross-linking state are used as output vectors. The objective function introduces a first-order term, a second-order term and an interactive coupling term, and iteratively updates the weights of each parameter along the gradient direction until the fitting residual is lower than the convergence threshold. Through this model, only the physicochemical parameters need to be collected during the production process to deduce the molecular aggregation state spectrum and oxidative cross-linking state, realizing the data management from physicochemical detection to functional state.

[0018] (2) The production quality control method of this medical mussel adhesive protein nasal spray is to input the real-time physicochemical parameters of each process node of the current batch into the mapping relationship model in sequence to obtain preliminary functional state deduction data, and to use the hidden state time sequence deduction algorithm to use the preliminary data as the initial value of the hidden state sequence, establish the state transition probability matrix between nodes and the observation probability matrix between the hidden state and the observation sequence, and recursively estimate each node along the process time axis from the first process node. The hidden state estimation value of the previous node is used as the prior information for each node, and the hidden state is corrected with the physicochemical parameters of the current node as the condition. The corrected molecular aggregation state and oxidative crosslinking dynamic evolution data are output to realize the continuous deduction of functional state along the production process, timely identify the molecular aggregation state shift and mark the shift direction.

[0019] (3) The production quality control method of this medical mussel adhesive protein nasal spray integrates the real-time physicochemical parameters, molecular aggregation state and oxidative crosslinking dynamic data and quality grades of multiple batches, generates a batch quality traceability data chain according to the production sequence, and arranges and splits the dynamic evolution data according to the process sequence. The deviation of the functional state characteristics from the overall average value is compared batch by batch. Abnormal batches that exceed the threshold are screened and the process nodes and physicochemical parameter categories are marked. After locking the key nodes of continuous deviation, the corresponding physicochemical parameter weights in the mapping relationship model are retrieved in reverse. The target parameter with the greatest contribution is determined and the operating conditions are adjusted along the gradient direction. The model construction, deduction and judgment steps are repeated until the functional state offset of multiple batches converges, realizing the closed-loop management of quality control from deviation discovery to active correction.

[0020] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the production quality control method for a medical mussel adhesive protein nasal spray according to the present invention.

[0022] Figure 2 This is a flowchart illustrating the specific steps involved in constructing a two-way control sample set in the production quality control method for a medical mussel adhesive protein nasal spray according to the present invention.

[0023] Figure 3 This is a flowchart illustrating the specific steps involved in constructing a mapping relationship model using a multivariate gradient coupling fitting algorithm in the production quality control method for a medical mussel adhesive protein nasal spray according to the present invention. Detailed Implementation

[0024] Please see Figure 1This invention provides a technical solution: a production quality control method for a medical mussel adhesive protein nasal spray, comprising the following steps: obtaining physicochemical parameters of raw material dissolution, solution preparation, pH adjustment, filtration, and filling processes in multiple historical production batches, as well as measured data of molecular aggregation state spectrum and oxidative cross-linking activity state corresponding to each batch of finished product; constructing a two-way control sample set by batch-wise mapping of physicochemical parameters and measured data of molecular aggregation state spectrum and oxidative cross-linking activity state; using a multivariate gradient coupling fitting algorithm to perform correlation calculations on the two-way control sample set to construct a mapping relationship model between the physicochemical parameter space and the functional state space, wherein the functional state space is composed of molecular aggregation state spectrum and oxidative cross-linking activity state; obtaining the current production batch's... The real-time physicochemical parameters of the process nodes are input into the mapping relationship model. Using the hidden state time-series deduction algorithm, the dynamic evolution data of the molecular aggregation state spectrum and oxidative crosslinking activity state of the current production batch are output. The dynamic evolution data of the molecular aggregation state spectrum and oxidative crosslinking activity state are compared with the functional state offset judgment threshold to determine the quality level of the current production batch. The physicochemical parameters and dynamic evolution data of the molecular aggregation state spectrum and oxidative crosslinking activity state of multiple batches are integrated to form a batch quality traceability data chain. The state difference characteristics are extracted and the production process operating conditions are adjusted accordingly. The mapping relationship model construction, deduction and quality level judgment are repeated for the adjusted production process until the functional state offset of multiple batches converges within the functional state offset judgment threshold.

[0025] The number of historical production batches selected should be no less than 30, covering production data from different production periods and different raw material batches; Physicochemical parameters include the dissolution temperature and viscosity data corresponding to the stirring speed in the raw material dissolution process, the protein concentration data corresponding to the solution concentration in the solution preparation process, the pH value data corresponding to the amount of buffer added in the pH adjustment process, the turbidity data corresponding to the filtration pressure in the filtration process, and the liquid level stability data corresponding to the filling speed in the filling process. Basic production information includes production operators, production equipment numbers, raw material batch numbers, and start and stop times for each process.

[0026] Specifically, such as Figure 2 As shown, the specific steps for constructing a two-way control sample set are as follows: The physicochemical parameters of the raw material dissolution, solution preparation, pH adjustment, filtration, and filling processes in multiple historical production batches were screened, specifically as follows: The historical production batches to be screened must be qualified production batches with no abnormal shutdowns, no raw material changes, and no process adjustments during the production process. The physicochemical parameters must include the complete data of each process from start to finish, and the time interval between recording the physicochemical parameters of each process shall not exceed 10 minutes. In one implementation, 40 historical qualified production batches are selected, and the physicochemical parameters of each batch are retained for the raw material dissolution, solution preparation, pH adjustment, filtration and filling processes. Batches with missing data at any process node are eliminated.

[0027] Collect measured data on the molecular aggregation state spectrum and oxidative crosslinking activity state of finished products from each historical production batch, and simultaneously retain the batch number. Specifically: After each batch of production is completed, no less than 3 samples shall be randomly selected from the finished products of that batch. Each sample shall be tested for molecular aggregation state spectrum and oxidative crosslinking activity state using the same detection method. The average value of the sample shall be taken as the measured data of the finished products of that batch and recorded in correspondence with the batch number. In one embodiment, three samples are taken from each batch of finished products. The molecular aggregation state spectrum is detected by gel filtration chromatography, and the oxidative crosslinking activity state is detected by colorimetry. The arithmetic mean of the detection results of the three samples is calculated as the measured data of the batch, and the batch number is recorded at the same time.

[0028] Match the physicochemical parameters with the measured data of molecular aggregation state spectrum and oxidative crosslinking activity state according to the batch number, complete the binding and perform secondary verification, and remove the control content that deviates from the normal range. Specifically: Based on the batch number, the physicochemical parameters of each process corresponding to the same number are associated and bound with the measured data of the finished product of that batch, forming a single batch control data group; The secondary verification adopts the data consistency check method, comparing the change trend of physicochemical parameters of different processes within the same batch with the matching of the measured data of molecular aggregation state spectrum and oxidative cross-linking activity state. If the physicochemical parameters show abnormal fluctuations but the measured data are within the normal range, or if the physicochemical parameters do not show obvious fluctuations but the measured data are abnormal, they are judged to deviate from the normal range and are rejected. In one implementation, if the physicochemical parameters are abnormally inconsistent with the measured data, the batch of control data group is removed.

[0029] All filtered and bound data are arranged in production order to form a two-way control sample set, specifically as follows: The single-batch control data sets that passed the second verification were arranged in the order of production batches. Each batch of control data sets included the physicochemical parameters of each process, the actual measured data of the finished product, and the batch number. After the arrangement was completed, a two-way control sample set was formed.

[0030] Specifically, such as Figure 3 As shown, the specific steps for constructing a mapping relationship model using the multivariate gradient coupling fitting algorithm are as follows: The physicochemical parameters of each batch in the two-way control sample set are used as the input vector, and the measured data of the molecular aggregation state spectrum and oxidative crosslinking activity state corresponding to that batch are used as the output vector. Specifically: The input vector for each batch consists of the physicochemical parameters of each process in that batch, including the final viscosity value of the raw material dissolution process, the final protein concentration value of the solution preparation process, the final pH value of the pH adjustment process, the average turbidity value of the filtration process, and the average liquid level stability value of the filling process. Each parameter is a component of the input vector. The output vector consists of the measured data of the molecular aggregation state spectrum and the measured data of the oxidative crosslinking activity state of the finished product in this batch, with the two data serving as the two components of the output vector respectively. In one implementation, the input vector for a certain batch includes the raw material solubility viscosity of 15 mPa·s, the protein concentration of the solution of 2.5 mg / mL, the pH value of 7.2, the average turbidity of filtration of 0.02 NTU, and the average liquid level stability of filling of 0.05 mm. The output vector includes the molecular aggregation state spectrum data and the oxidative crosslinking activity state data of the batch. The two are bound together to form a set of input and output data pairs.

[0031] A multivariate coupled objective function is constructed, comprising linear, quadratic, and interactive coupling terms of physicochemical parameters, as follows: The multivariate coupling objective function takes each physicochemical parameter in the input vector as a variable and includes a first-order term for each physicochemical parameter, a second-order term for each physicochemical parameter, and an interactive coupling term between any two physicochemical parameters. In one embodiment, the multivariate coupling objective function includes a first-order term of the raw material solubility viscosity, a second-order term of the raw material solubility viscosity, a first-order term of the prepared protein concentration, a second-order term of the prepared protein concentration, a first-order term of the pH value, and a second-order term of the pH value. It also includes an interactive coupling term between the raw material solubility viscosity and the prepared protein concentration, and an interactive coupling term between the pH value and the filtration turbidity.

[0032] The weights of each physicochemical parameter are iteratively updated along the gradient direction of the multivariate coupled objective function. In each iteration, the fitting residual of the multivariate coupled objective function after the weight update is calculated. The iteration stops when the fitting residual is lower than a preset convergence threshold, and the set of physicochemical parameter weights at this time is fixed as the model parameters of the mapping relationship model. Specifically: During iterative updates, with the goal of minimizing the fitting residual, the weights of the first-order, second-order, and interactive coupling terms of each physicochemical parameter are gradually adjusted along the gradient direction of the multivariate coupled objective function. After each weight update, the fitting residual of the objective function under the current weight is calculated, which is the deviation between the output value of the objective function and the measured value of the output vector in the two-way control sample set. When the fitting residual is lower than the preset convergence threshold, the iteration stops, and the weights of all current physicochemical parameters are organized into a weight set, which serves as the fixed parameters of the mapping relationship model. In one implementation, the preset convergence threshold is set to 0.01, the initial weights are all set to 0.2, the step size of the weight adjustment in each iteration is 0.001, and after 1000 iterations, the fitting residual drops to 0.008, which is lower than the preset convergence threshold, the iteration stops, and the weight set of each physicochemical parameter item at this time is solidified as the model parameters.

[0033] Specifically, the steps for outputting dynamic evolution data using the hidden state time series deduction algorithm are as follows: The real-time physicochemical parameters of the current production batch at each process node are arranged in process sequence and used as the observation sequence input into the mapping relationship model to obtain preliminary functional state projection data for each process node, specifically: The current production batch's process nodes are arranged in the order of raw material dissolution, solution preparation, pH adjustment, filtration, and filling. Real-time physicochemical parameters are collected for each process node, with the collection time interval consistent with historical batches. The real-time physicochemical parameters of each process node are organized into an observation sequence according to the process sequence, and the length of the observation sequence is consistent with the number of process nodes. The observed sequence is input node by node into the mapping relationship model of the solidified parameters. The model calculates and outputs the preliminary inference data of the molecular aggregation state spectrum and oxidative crosslinking activity state corresponding to each process node based on the physicochemical parameters and fixed weights of each node. In one implementation, the current production batch is arranged in the order of raw material dissolution, solution preparation, pH adjustment, filtration, and filling to form an observation sequence containing 5 nodes. This sequence is input into a mapping model, which outputs preliminary inference data on the molecular aggregation state spectrum and oxidative crosslinking activity state corresponding to the 5 process nodes.

[0034] Using the preliminary functional state projection data of each process node as the initial value of the hidden state sequence, a state transition probability matrix is ​​established between adjacent nodes of the hidden state sequence on the process time axis, as well as an observation probability matrix between the physical and chemical parameters of each node in the hidden state sequence and the corresponding node in the observation sequence. Specifically: The hidden state sequence is arranged in the order of the process time axis, with the preliminary functional state projection data of each process node as the initial value. The state transition probability matrix is ​​used to describe the probability of the hidden state of the previous process node transitioning to the hidden state of the current process node. This probability is determined according to the change pattern of the functional state of adjacent process nodes in the historical batch. The observation probability matrix is ​​used to describe the probability of the physicochemical parameters of a certain process node in the observation sequence corresponding to the hidden state of that node. This probability is determined based on the correspondence between physicochemical parameters and functional states in the two-way comparison sample set. In one implementation, based on historical batch data, the probability of the hidden state of the raw material dissolving process transitioning to the hidden state of the solution preparation process is set to 0.92, and the probability of the hidden state of the solution preparation process transitioning to the hidden state of the pH adjustment process is set to 0.95. At the same time, the observation probability of the hidden state of each process and the corresponding physicochemical parameters is determined based on historical data, and a complete state transition probability matrix and observation probability matrix are constructed.

[0035] Starting from the first process node, recursive estimation is performed node by node along the process time axis. The hidden state estimate of the previous process node is used as the prior information of the current process node. The measured physicochemical parameters of the current process node are used as conditions to perform Bayesian correction on the hidden state estimate. The dynamic evolution data of the corrected molecular aggregation state spectrum and oxidative crosslinking activity state of each process node are output. Specifically: The recursive estimation starts from the raw material dissolution process. The preliminary functional state data of this process is used as the initial hidden state estimate. Then, the process moves to the solution preparation process. The hidden state estimate of the raw material dissolution process is used as the prior information of the solution preparation process. Combined with the real-time measured values ​​of the physicochemical parameters of the solution preparation process, the hidden state estimate of the solution preparation process is adjusted by the Bayesian correction method to obtain the corrected functional state data of the solution preparation process. By analogy, recursive estimation and Bayesian correction are completed node by node, and finally the molecular aggregation state spectrum and oxidative crosslinking activity state data after correction of each process node are output, forming complete dynamic evolution data; In one implementation, the molecular aggregation state spectrum data after the raw material dissolution process is used as prior information. Combined with the real-time measured protein concentration value of the solution preparation process, Bayesian correction is performed on the molecular aggregation state spectrum and oxidative crosslinking activity state of the solution preparation process. The correction of all process nodes is completed in sequence, and dynamic evolution data is output.

[0036] Specifically, the steps for determining the functional state offset threshold are as follows: The measured data of molecular aggregation state profiles and oxidative crosslinking activity states of historical batches that have been clinically validated as qualified were collected. Distribution characteristic parameters of the molecular aggregation state profiles and activity characteristic parameters of the oxidative crosslinking activity states were extracted to form a qualified sample set, specifically: Historical batches whose functional status has been clinically verified as qualified refer to production batches whose finished products can meet the requirements of nasal mucosal film formation and retention and mucosal repair in clinical applications, and have no adverse clinical reactions. Distribution characteristic parameters of molecular aggregated state spectrum were extracted from this type of historical batch, including the main distribution range and distribution peak of molecular aggregated state. Activity characteristic parameters of oxidative crosslinking active state were extracted, including activity peak and activity stability range. All extracted characteristic parameters were sorted by batch to form a qualified sample set. In one implementation, 20 batches of historically qualified clinically validated samples are collected, and the main distribution range of the molecular aggregate state spectrum of each batch is extracted as 100-500kDa, the distribution peak is 300kDa, the activity peak of the oxidative crosslinking active state is 85%, and the activity stability range is 75%-95%, forming a qualified sample set.

[0037] Centered on the mean of the distribution characteristic parameters of the molecular aggregate state spectrum in the qualified sample set, and with the standard deviation of a preset multiple as the boundary, the distribution threshold range of the molecular aggregate state spectrum is defined. Similarly, the activity threshold range of the oxidative crosslinking active state is defined. The above threshold range is used as the functional state shift determination threshold, specifically as follows: Calculate the mean and standard deviation of the distribution characteristic parameters of the molecular aggregation state spectrum in the qualified sample set. Use the mean minus a preset multiple of the standard deviation as the lower limit of the distribution threshold range, and use the mean plus a preset multiple of the standard deviation as the upper limit of the distribution threshold range. Using the same method, the mean and standard deviation of the activity characteristic parameters of the oxidative crosslinking active state were calculated, and the activity threshold range was defined. In one embodiment, the mean of the molecular aggregated state spectrum distribution characteristic parameters is 300 kDa, the standard deviation is 50 kDa, the preset multiple is 1.2, and the defined distribution threshold range is 240-360 kDa. The mean value of the activity characteristic parameters of the oxidative crosslinking active state is 85%, the standard deviation is 5%, the preset multiple is 1.2, and the defined activity threshold range is 79%-91%. The molecular aggregation spectrum distribution threshold and the oxidative crosslinking active state activity threshold together constitute the functional state shift judgment threshold.

[0038] Specifically, the steps for determining the quality level of the current production batch are as follows: The dynamic evolution data of the molecular aggregation state spectrum and oxidative crosslinking activity state of each process node were extracted, specifically as follows: From the dynamic evolution data output by the hidden state time-series deduction algorithm, the molecular aggregation state spectrum distribution characteristic parameters and oxidative cross-linking activity characteristic parameters corresponding to each process of raw material dissolution, solution preparation, pH adjustment, filtration and filling of the current production batch are extracted node by node; In one implementation, the distribution range and peak value of the molecular aggregation state spectrum of the five process nodes of the current production batch, as well as the activity peak value and activity stability range of the oxidative crosslinking active state, are extracted to form complete dynamic evolution data of each node.

[0039] The deviation of the molecular aggregation state spectrum distribution characteristic parameters of each process node from the distribution threshold range of the functional state shift judgment threshold, and the deviation of the activity characteristic parameters of the oxidative crosslinking active state from the activity threshold range are calculated. Specifically: The degree of deviation is calculated using the relative deviation calculation method. For the distribution characteristic parameter of the molecular aggregate system, the center value of the distribution threshold range is used as the benchmark. The difference between the actual value and the center value of the parameter is calculated and then divided by the center value to obtain the relative deviation. For the activity characteristic parameters of the oxidative crosslinking active state, the relative deviation is calculated using the same method with the center value of the activity threshold range as the benchmark. A positive relative deviation indicates that the parameter is higher than the center value, while a negative relative deviation indicates that the parameter is lower than the center value. In one implementation, the actual value of the molecular aggregation state spectrum distribution characteristic parameter of a certain process node is 330kDa, the center value of the distribution threshold range is 300kDa, and the calculated relative deviation is 10%. The actual value of the activity characteristic parameter of the oxidative crosslinking active state is 82%, the center value of the activity threshold range is 85%, and the calculated relative deviation is -3.5%.

[0040] The percentage of process nodes with varying degrees of deviation within each level is used to determine the quality grade of the current production batch. When the distribution characteristic parameters of the molecular aggregate state spectrum exceed the distribution threshold range or the activity characteristic parameters of the oxidative crosslinking active state exceed the activity threshold range, the offset direction is marked, specifically as follows: Three quality level ranges are preset: qualified range, warning range, and unqualified range. The qualified range is when the absolute value of the relative deviation does not exceed 5%, the warning range is when the absolute value of the relative deviation is greater than 5% but not more than 10%, and the unqualified range is when the absolute value of the relative deviation is greater than 10%. Count the number of nodes in each process node of the current production batch that are in each level interval, and calculate the node proportion of each level interval; The quality level is determined based on the proportion. If the proportion of qualified nodes in the qualified range is not less than 90%, it is judged as qualified. If the proportion of nodes in the warning interval is not less than 50%, it is determined to be a warning level; If the proportion of non-compliant nodes is not less than 10%, it is judged as non-compliant. If the molecular aggregation spectrum distribution characteristic parameter of a node exceeds the distribution threshold range, or the activity characteristic parameter of the oxidative crosslinking active state exceeds the activity threshold range, mark the offset direction of the parameter, i.e., above or below the threshold range.

[0041] Specifically, the steps to form a batch quality traceability data chain are as follows: The data integrates the dynamic evolution of real-time physicochemical parameters, molecular aggregation state spectrum, and oxidative crosslinking activity state of the current batch, as well as its quality grade, specifically: The real-time physicochemical parameters of each process node of the current production batch, the corrected molecular aggregation state spectrum and dynamic evolution data of the oxidative crosslinking activity state of each process node, and the overall quality level of the current batch are collected into the same data document, with each data point labeled with the corresponding process node and collection time. In one implementation, the real-time physicochemical parameters of the five process nodes of the current batch, the dynamic evolution data of each node, and the determined pass level are collected into a data document named with the batch number. Each data is labeled with the corresponding process and collection time, forming a complete single batch data set. Associate the batch number with basic production information to generate an independent identification code and store it permanently. Specifically: The batch number of the current batch is linked and bound to the basic production information, which includes the name of the production operator, the position of the operator, the production equipment number, the raw material batch number, the start and stop time of each process, and the temperature and humidity of the production environment. Based on the associated information, a unique identification code is generated. The code format includes the batch number, equipment number, and production date. The unique identification code and all corresponding data are stored in a fixed location using an encrypted storage method.

[0042] By linking the independent identification codes of each batch in the production sequence, a batch quality traceability data chain is built, and a query index is established. Specifically: All the unique identification codes of the production batches are concatenated in chronological order, and each unique identification code is associated with the complete data of the corresponding batch, forming a coherent batch quality traceability data chain. Create a query index with keywords including batch number, unique identification code, production date, and quality grade.

[0043] Specifically, the steps for extracting state difference features are as follows: The dynamic evolution data of the molecular aggregation state spectrum and oxidative crosslinking activity state of multiple batches in the batch quality traceability data chain are arranged in the order of the production process, specifically as follows: Select dynamic evolution data of no less than 10 consecutive production batches from the batch quality traceability data chain, organize the dynamic evolution data of each batch according to the process sequence of raw material dissolution, liquid preparation, pH adjustment, filtration and filling, and arrange the dynamic evolution data of the same process node according to the batch sequence to form a dynamic evolution data matrix classified by process and sorted by batch. In one implementation, dynamic evolution data of 15 consecutive production batches are selected, and the dynamic evolution data of molecular aggregation state spectrum and oxidative crosslinking activity state of 5 process nodes in each batch are classified by process and sorted by batch to form dynamic evolution data sequences corresponding to 5 processes, with each sequence containing 15 batches of data.

[0044] The dynamic evolution data is split into layers, and the distribution characteristics of molecular aggregation state are compared with the overall average distribution characteristics of multiple batches, as well as the activity characteristics of oxidative crosslinking active state with the overall average activity characteristics of multiple batches, to screen out abnormal batches that deviate from the preset difference threshold. Specifically: The dynamic evolution data is split into layers, that is, layered by process node, with each process node as an independent analysis layer; Calculate the average value of the molecular aggregation state spectrum distribution characteristic parameters and the average value of the oxidative crosslinking active state activity characteristic parameters of multiple batches as a whole, compare the difference between the parameters of each process node and the corresponding average value for each batch, and calculate the difference value. The preset difference threshold is determined based on the allowable range of production quality fluctuations. When the difference value of a certain batch at a certain process node exceeds the preset difference threshold, the batch is determined to be an abnormal batch. In one implementation, a preset difference threshold of 8% is set, the overall average value of 15 batches of data is calculated, and each batch is compared. If the difference between the molecular aggregation state spectrum distribution characteristic parameter of a certain batch of liquid preparation process and the average value is 10%, which exceeds the preset difference threshold, the batch is determined to be an abnormal batch.

[0045] The deviation data of abnormal batches are categorized, and the process node to which the deviation belongs and the corresponding physicochemical parameter category are marked. Specifically: All deviation data exceeding the preset difference threshold in the abnormal batch are classified according to the type of deviation parameter, into two categories: deviation of molecular aggregated state spectrum and deviation of oxidative cross-linking active state. At the same time, the process node to which each deviation data belongs, as well as the physicochemical parameter category corresponding to that process node, are labeled; In one implementation, the deviation data of an abnormal batch is a deviation of the molecular aggregate state spectrum, the process node to which it belongs is the solution preparation process, and the corresponding physicochemical parameter category is the solution protein concentration. The deviation data is classified as a deviation of the molecular aggregate state spectrum, and the solution preparation process and solution protein concentration parameter are labeled.

[0046] Specifically, the steps for adjusting the operating conditions of the production process are as follows: By comparing the dynamic evolution data of molecular aggregation state spectrum and oxidative crosslinking activity state in multiple batches of quality traceability data chain with the functional state deviation judgment threshold, the key process nodes where the dynamic evolution data continuously deviates from the functional state deviation judgment threshold are identified. Specifically: Select dynamic evolution data from multiple batches of continuous production, compare the dynamic evolution data with the functional state deviation judgment threshold batch by batch and node by node, and count the number of times the dynamic evolution data deviates from the threshold for each process node. If the dynamic evolution data of a certain process node deviates from the functional state deviation judgment threshold in three or more consecutive production batches, and the direction of deviation is consistent, then the process node is locked as a critical process node. In one implementation, 10 consecutive production batches are selected, and the dynamic evolution data of each batch and each node is compared with the functional state deviation judgment threshold. Statistical analysis shows that the dynamic evolution data of the liquid preparation process node is lower than the functional state deviation judgment threshold in 4 consecutive production batches, and the liquid preparation process is locked as a key process node.

[0047] By reversing the retrieval of the mapping relationship model, the physicochemical parameters corresponding to key process nodes and their weight coefficients in the multivariate coupling objective function are determined, identifying the target physicochemical parameters that contribute most to the shift in molecular aggregation state spectrum and oxidative crosslinking active state. Specifically: From the mapping relationship model of the solidified parameters, all physicochemical parameters corresponding to the key process nodes are retrieved in reverse. The weight coefficients of these physicochemical parameters in the multivariate coupling objective function are extracted. The larger the weight coefficient, the greater the influence of the physicochemical parameter on the molecular aggregation state spectrum and the oxidative crosslinking activity state. The physicochemical parameter with the largest weight coefficient is determined as the target physicochemical parameter. In one implementation, the physicochemical parameters corresponding to the reverse retrieval of the solution preparation process are the protein concentration and the stirring time of the solution preparation. Their weight coefficients in the multivariate coupling objective function are 0.35 and 0.15, respectively. The protein concentration has a larger weight coefficient, so it is determined as the target physicochemical parameter.

[0048] Based on the gradient direction of the target physicochemical parameters in the mapping relationship model, the adjustment direction and amount of the target physicochemical parameters are determined, and the production process operating conditions of key process nodes are adjusted accordingly. Specifically: Analyze the gradient direction of the target physicochemical parameters in the multivariate coupled objective function. If the gradient direction is positive, adjust the direction to increase the target physicochemical parameters. If the gradient direction is negative, the adjustment direction is to decrease the target physicochemical parameters; The adjustment amount is determined based on the weighting coefficients and the degree of deviation of the target physicochemical parameters; Based on the direction and amount of adjustment, modify the production process operating conditions of key process nodes; In one embodiment, the target physicochemical parameter is the protein concentration of the solution, the gradient direction is positive, the current protein concentration of the solution is 2.0 mg / mL, the adjustment direction is to increase, the adjustment amount is 0.3 mg / mL, and the protein concentration of the solution in the solution preparation process is adjusted to 2.3 mg / mL.

[0049] Specifically, the steps for determining the convergence of functional state offsets across multiple batches are as follows: Collect the molecular aggregation state distribution characteristics and oxidative crosslinking active state activity characteristics of consecutive production batches. Calculate the distance between the molecular aggregation state distribution characteristics and the center of the functional state shift judgment threshold, and the distance between the oxidative crosslinking active state activity characteristics and the center of the activity threshold for each batch. Specifically: Collect molecular aggregation state spectrum distribution characteristics and oxidative crosslinking activity characteristics of no less than three batches of products produced continuously after adjusting the production process operating conditions. The distance between the two types of parameters and the corresponding threshold center is calculated in batches. The distance calculation adopts the absolute difference method, that is, the absolute difference between the actual value of the parameter and the threshold center value. In one implementation, data from three batches of adjusted products are collected, and the absolute difference between the molecular aggregation state spectrum distribution characteristic parameter and the distribution threshold center, as well as the absolute difference between the oxidative crosslinking active state activity characteristic parameter and the activity threshold center, are calculated for each batch to obtain two distance values ​​for each batch.

[0050] When the distances of multiple consecutive batches are all within the functional state offset determination threshold range, and the fluctuation range of the distance between adjacent batches is lower than the preset convergence determination threshold, it is determined that the functional state offsets of multiple batches have converged, specifically as follows: The preset number of consecutive batches is no less than 3 batches, and the preset convergence judgment threshold is determined according to the production quality stability requirements. First, check whether the distances of the two parameters of three or more consecutive batches of products are all within the corresponding threshold range. If they are all within the range, then calculate the fluctuation range of the distance of the same parameter between adjacent batches, that is, the absolute difference between the distance of the later batch and the distance of the earlier batch. If the fluctuation amplitude of all adjacent batches is lower than the preset convergence judgment threshold, it is determined that the functional state offset of multiple batches has converged. In one implementation, the preset number of consecutive batches is 3, the preset convergence judgment threshold is 0.5, the distance between the two types of parameters of the 3 consecutive batches of products is within the threshold range, and the fluctuation range of the distance between adjacent batches is 0.3 and 0.2 respectively, both lower than the preset threshold, and it is determined that the functional state offset of multiple batches has converged.

[0051] If the convergence condition is not met, the mapping relationship model construction, dynamic evolution data deduction of molecular aggregation state spectrum and oxidative crosslinking activity state, and quality level determination are repeated, and the production process operating conditions are adjusted continuously, specifically as follows: If the number of consecutive batches does not reach the preset number, or the parameter distance of some batches exceeds the threshold range, or the fluctuation range between adjacent batches is higher than the preset convergence judgment threshold, then the convergence condition is not met. At this point, multiple batches of historical data are re-collected to construct a two-way comparison sample set, and the mapping relationship model construction steps are re-executed. The dynamic evolution data of the new production batch is extrapolated and the quality level is determined. At the same time, based on the new deviation data, the production process operating conditions are adjusted until the convergence conditions are met.

[0052] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0053] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for quality control in the production of a medical mussel adhesive protein nasal spray, characterized in that, Includes the following steps: The physicochemical parameters of raw material dissolution, solution preparation, pH adjustment, filtration and filling processes in multiple historical production batches were obtained, as well as the measured data of molecular aggregation state spectrum and oxidative cross-linking activity state of the finished products of each batch. The physicochemical parameters and the measured data of molecular aggregation state spectrum and oxidative cross-linking activity state were matched by batch to construct a two-way control sample set. A multivariate gradient coupling fitting algorithm was used to perform correlation calculations on the two-way control sample set to construct a mapping relationship model between the physicochemical parameter space and the functional state space. The functional state space consists of the molecular aggregation state spectrum and the oxidative cross-linking active state. The system obtains real-time physicochemical parameters of the current production batch at each process node, inputs the mapping relationship model, and uses the hidden state time series deduction algorithm to output the dynamic evolution data of the molecular aggregation state spectrum and oxidative crosslinking activity state of the current production batch. The dynamic evolution data of molecular aggregation state spectrum and oxidative cross-linking active state are compared with the functional state shift judgment threshold to determine the quality level of the current production batch. By integrating dynamic evolution data of physicochemical parameters, molecular aggregation state spectrum, and oxidative crosslinking activity state from multiple batches, a batch quality traceability data chain is formed. State difference characteristics are extracted and production process operating conditions are adjusted accordingly. The adjusted production process is repeatedly mapped to construct, deduce, and determine the quality level until the functional state offset of multiple batches converges within the functional state offset determination threshold.

2. The production quality control method for medical mussel adhesive protein nasal spray according to claim 1, characterized in that, The specific steps for constructing a two-way control sample set are as follows: Screening the physicochemical parameters of raw material dissolution, solution preparation, pH adjustment, filtration and filling processes in multiple historical production batches; Collect measured data on the molecular aggregation state spectrum and oxidative cross-linking activity state of finished products from each historical production batch, and retain the batch number simultaneously. Match the physicochemical parameters with the measured data of molecular aggregation state spectrum and oxidative crosslinking activity state according to the batch number, complete the binding and perform secondary verification, and remove the control content that deviates from the normal range; All the filtered and bound data are arranged in production order to form a two-way control sample set.

3. The production quality control method for medical mussel adhesive protein nasal spray according to claim 1, characterized in that, The specific steps for constructing a mapping relationship model using the multivariate gradient coupling fitting algorithm are as follows: The physicochemical parameters of each batch in the two-way control sample set are used as input vectors, and the measured data of the molecular aggregation state spectrum and oxidative crosslinking activity state corresponding to that batch are used as output vectors. Construct a multivariate coupled objective function that includes linear, quadratic, and interactive coupling terms of physicochemical parameters; The weights of each physicochemical parameter are iteratively updated along the gradient direction of the multivariate coupled objective function. In each iteration, the fitting residual of the multivariate coupled objective function after the weight update is calculated. When the fitting residual is lower than the preset convergence threshold, the iteration stops and the set of weights of the physicochemical parameter at this time is fixed as the model parameters of the mapping relationship model.

4. The production quality control method for medical mussel adhesive protein nasal spray according to claim 1, characterized in that, The specific steps for outputting dynamic evolution data using the hidden state time series extrapolation algorithm are as follows: The real-time physicochemical parameters of the current production batch at each process node are arranged in the process sequence and used as the observation sequence input into the mapping relationship model to obtain preliminary functional state projection data for each process node. The preliminary functional state projection data of each process node is used as the initial value of the hidden state sequence. The state transition probability matrix between each adjacent node of the hidden state sequence on the process time axis is established, as well as the observation probability matrix between each node in the hidden state sequence and the corresponding node in the observation sequence. Starting from the first process node, recursive estimation is performed node by node along the process time axis. The hidden state estimate of the previous process node is used as the prior information of the current process node. The hidden state estimate is Bayesian corrected based on the measured physicochemical parameters of the current process node. The dynamic evolution data of the corrected molecular aggregation state spectrum and oxidative crosslinking active state of each process node are output.

5. The production quality control method for medical mussel adhesive protein nasal spray according to claim 1, characterized in that, The steps for determining the functional state offset threshold are as follows: Collect measured data of molecular aggregate state spectrum and oxidative cross-linking activity state corresponding to historical batches whose functional status has been clinically verified as qualified, extract the distribution characteristic parameters of molecular aggregate state spectrum and the activity characteristic parameters of oxidative cross-linking activity state, and form a qualified sample set. Centered on the mean value of the distribution characteristic parameters of the molecular aggregate spectrum in the qualified sample set, and with the standard deviation of a preset multiple as the boundary, the distribution threshold range of the molecular aggregate spectrum is defined. Similarly, the activity threshold range of the oxidative crosslinking active state is defined, and the above threshold range is used as the functional state deviation judgment threshold.

6. The production quality control method for medical mussel adhesive protein nasal spray according to claim 1, characterized in that, The specific steps for determining the quality level of the current production batch are as follows: Extract dynamic evolution data of molecular aggregation state spectrum and oxidative crosslinking activity state at each process node; The deviation of the molecular aggregation state spectrum distribution characteristic parameters of each process node from the distribution threshold range in the functional state offset judgment threshold, and the deviation of the oxidative crosslinking active state activity characteristic parameters from the activity threshold range; The percentage of process nodes with varying degrees of deviation within each level is used to determine the quality grade of the current production batch. When the distribution characteristic parameters of the molecular aggregate spectrum exceed the distribution threshold range or the activity characteristic parameters of the oxidative crosslinking active state exceed the activity threshold range, the offset direction is marked.

7. The production quality control method for medical mussel adhesive protein nasal spray according to claim 1, characterized in that, The specific steps for forming a batch quality traceability data chain are as follows: Integrate the dynamic evolution data and quality grade of the current batch's real-time physicochemical parameters, molecular aggregation state spectrum, and oxidative crosslinking activity state; Associate batch numbers with basic production information to generate unique identification codes and store them permanently. By linking the independent identification codes of each batch in the production sequence, a batch quality traceability data chain is built and a query index is established.

8. The production quality control method for medical mussel adhesive protein nasal spray according to claim 1, characterized in that, The specific steps for extracting state difference features are as follows: The dynamic evolution data of the molecular aggregation state spectrum and oxidative cross-linking activity state of multiple batches in the batch quality traceability data chain are arranged in the order of the process. The dynamic evolution data is split into layers, and the distribution characteristics of molecular aggregation state spectrum are compared with the overall average distribution characteristics of multiple batches, as well as the activity characteristics of oxidative crosslinking active state and the overall average activity characteristics of multiple batches, to screen out abnormal batches that deviate from the preset difference threshold. The deviation data of abnormal batches are categorized, and the process node to which the deviation belongs and the corresponding physicochemical parameter category are marked.

9. The production quality control method for the medical mussel adhesive protein nasal spray according to claim 1, characterized in that, The specific steps for adjusting the operating conditions of the production process are as follows: By comparing the dynamic evolution data of molecular aggregation state spectrum and oxidative crosslinking activity state in multiple batches of quality traceability data chain with the functional state deviation judgment threshold, the key process nodes where the dynamic evolution data continuously deviates from the functional state deviation judgment threshold are identified. By reversing the retrieval of the mapping relationship model, the physicochemical parameters corresponding to the key process nodes and their weight coefficients in the multivariate coupling objective function are determined to identify the target physicochemical parameters that contribute the most to the shift of the molecular aggregation state spectrum and the oxidative crosslinking active state. Based on the gradient direction of the target physicochemical parameters in the mapping relationship model, the adjustment direction and amount of the target physicochemical parameters are determined, and the production process operating conditions of key process nodes are adjusted.

10. The production quality control method for medical mussel adhesive protein nasal spray according to claim 1, characterized in that, The specific steps for determining the convergence of functional status offsets across multiple batches are as follows: Collect molecular aggregated state spectrum distribution characteristic parameters and oxidative crosslinking active state activity characteristic parameters of continuous production batches, and calculate the distance between the molecular aggregated state spectrum distribution characteristic parameters and the distribution threshold center in the functional state offset judgment threshold, as well as the distance between the oxidative crosslinking active state activity characteristic parameters and the activity threshold center for each batch. When the distances of multiple consecutive batches are all within the functional state offset judgment threshold range, and the fluctuation range of the distance between adjacent batches is lower than the preset convergence judgment threshold, it is determined that the functional state offsets of multiple batches have converged. If the convergence condition is not met, the mapping relationship model construction, dynamic evolution data deduction of molecular aggregation state spectrum and oxidative crosslinking activity state, and quality level determination are repeated, and the production process operation conditions are adjusted again.

Images