CN-122024814-A - Respiratory tract virus inactivation data driven modeling prediction control method

CN122024814ACN 122024814 ACN122024814 ACN 122024814ACN-122024814-A

Abstract

The invention discloses a respiratory tract virus inactivation data driving modeling prediction control method, which relates to the technical field of biological medicine and comprises the following steps: obtaining initial virus characteristic data from a virus surface structure database through molecular dynamics simulation, and obtaining a structure change matrix of the virus under different inactivation conditions after treatment; extracting characteristics of virus characteristics and antibody binding sites by adopting a support vector machine classifier according to the structure change matrix, and determining a potential unexpected interaction site set; the respiratory tract virus inactivation data driving modeling prediction control method effectively improves the scientificity and stability of inactivation process parameter determination, reduces the safety risk caused by improper inactivation conditions, shortens the process optimization period, reduces the experiment cost, and is beneficial to maintaining good immune effectiveness while guaranteeing the virus inactivation safety, thereby better meeting the practical application requirements in vaccine research and development and virus prevention and control.

Inventors

XU ZIMU
ZHANG HUIMIN
QIAN ZHONG
SHEN HUILING
LI SHUYI
ZHANG MENGYU
HU SHUHENG
WU DANZHOU
LV BIN
Zhang Songkai
LI FENGWEN
WU QI
WANG YUANXIN

Assignees

合肥工业大学

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (10)

1. The respiratory tract virus inactivation data driven modeling prediction control method is characterized by comprising the following steps of: S1, acquiring initial virus characteristic data from a virus surface structure database through molecular dynamics simulation, and obtaining a structure change matrix of the virus under different inactivation conditions after treatment; S2, extracting characteristics of virus characteristics and antibody binding sites by adopting a support vector machine classifier according to a structure change matrix, and determining a potential unexpected interaction site set; S3, if the potential unexpected interaction site set exceeds a preset threshold, predicting influence distribution of the inactivation process variable on the interaction strength through a random forest regression model, and obtaining an optimized inactivation condition range; S4, obtaining simulated immune response data from the optimized inactivation condition range, judging the immune chaotic response probability, and obtaining a regulation and control accurate adjustment vector; S5, iterating a virus characteristic model by adopting a gradient descent optimization algorithm aiming at the regulation and control accurate adjustment vector, and determining a final inactivation process parameter set; S6, simulating a vaccine development process through a final inactivation process parameter set to obtain a safe and effective evaluation index set; and S7, outputting a complete regulation and control scheme description if the safety and effectiveness evaluation index set meets a preset standard.
2. The method for modeling and predicting respiratory tract virus inactivation data driving according to claim 1, wherein S1 comprises: obtaining initial virus characteristic data from a virus surface structure database through molecular dynamics simulation, and obtaining a preliminary structure model after processing; For the primary structure model, introducing inactivation condition variables including temperature and pH value, and simulating protein folding changes under different inactivating agent concentrations to obtain a condition influence data set; According to the condition influence data set, constructing a row vector of a structural change matrix to represent an inactivation condition variable, and a column vector to represent a virus surface structure locus, determining a matrix element value as a locus displacement difference value, wherein the locus displacement difference value is obtained by comparing Euclidean distances of the virus surface structure locus coordinates before and after inactivation; if the matrix element value exceeds a preset threshold value, adjusting simulation parameters to optimize protein folding change simulation, and obtaining a refining structure change matrix; And extracting change trend tracking features from the refined structure change matrix, and combining control data acquired from the condition influence data set to confirm the accuracy of the matrix to obtain the structure change matrix of the virus under different inactivation conditions.
3. The method for modeling and predicting respiratory tract virus inactivation data driving according to claim 1, wherein S2 comprises: Protein folding change data corresponding to the inactivation condition variable are obtained from the structure change matrix, and a support vector machine classifier is adopted to carry out preliminary division on virus characteristics, so that a feature vector set is obtained; Aiming at the feature vector set, introducing an inactivating agent type variable binding site displacement difference value, extracting relevant properties of the antibody binding site, and determining a preliminary interaction site list; Acquiring change trend tracking information from a condition influence data set according to the preliminary interaction site list, judging that potential unexpected sites are marked if the site displacement difference exceeds a preset threshold value, and obtaining a screening site group; integrating virus surface structure data to optimize a support vector machine classifier by screening a site group, wherein the support vector machine classifier takes a site displacement difference value as input, constructs a decision boundary by maximizing a classification interval, outputs a site classification result, constructs a site stability evaluation index based on the site classification result, and obtains a refined interaction site matrix; and (3) aggregating unexpected interaction attributes from the refined interaction site matrix, and determining a potential unexpected interaction site set by combining with change trend tracking verification.
4. The method for modeling and predicting respiratory tract virus inactivation data driving according to claim 1, wherein S3 comprises: Acquiring inactivation process variable data from a potential unexpected interaction site set, and if the number of the site sets exceeds a preset threshold, acquiring interaction strength influence distribution by taking the inactivation process variable data as input process variable influence through a random forest regression model; Aiming at interaction strength influence distribution, introducing a condition range optimization attribute as a distribution boundary adjustment basis, adopting a distribution data integration site set threshold value, and determining a variable influence prediction result; according to the variable influence prediction result, combining the regression model distribution attribute as an intensity deviation calculation basis, extracting intensity distribution to obtain information, judging that if the distribution deviation exceeds the range, adjusting the process variable distribution to obtain a preliminary optimization condition; And polymerizing unexpected interaction sites and preset threshold judgment attributes to serve as verification inputs through preliminary optimization conditions, and verifying the optimization condition range to obtain a final inactivation condition range.
5. The method for modeling and predicting respiratory tract virus inactivation data driving according to claim 1, wherein S4 comprises: obtaining simulated immune response data from the optimized inactivation condition range, and processing site interaction strength through response data aggregation, wherein the aggregation processing comprises the steps of summarizing the response data according to site classification, calculating an interaction strength average value, and judging immune confusion reaction probability; Aiming at the immune confusion reaction probability, introducing probability threshold calibration as an adjustment basis, wherein the threshold calibration integrates the process variable influence by comparing the probability with a preset standard value through adopting distribution boundary adjustment, wherein the distribution boundary adjustment comprises expanding a variable influence interval, and determining a preliminary form of a regulation and control accurate adjustment vector; according to the preliminary form of the regulation and control accurate adjustment vector, aggregating simulation response data and intensity deviation calculation, wherein the intensity deviation calculation is carried out by subtracting an interaction intensity value from an average value of the response data to obtain an immune response simulation verification attribute, and judging that if the verification attribute exceeds a preset threshold value, the vector distribution is adjusted; Extracting influence distribution prediction aiming at a final inactivation range by adjusting vector distribution, wherein the influence distribution prediction comprises extracting peak points from the distribution to obtain a regulation and control accurate adjustment vector optimization version; and acquiring response data aggregation from the regulation and control precise regulation vector optimized version, wherein the response data aggregation integrates the previous response data through the optimized version vector, and a final regulation and control precise regulation vector is determined.
6. The method for modeling and predicting respiratory tract virus inactivation data driving according to claim 1, wherein S5 comprises: obtaining immune response data from virus inactivation conditions, introducing a gradient descent optimization algorithm aiming at a regulation and control precise adjustment vector, wherein the gradient descent optimization algorithm takes the vector as input, iteratively updating model parameters by minimizing a loss function, and iterating a virus characteristic model, wherein the virus characteristic model takes the immune response data as input, and outputting the optimized characteristic parameters to obtain a preliminary process parameter set; The response data are aggregated through the preliminary process parameter set, an interaction intensity average value is calculated, wherein the interaction intensity average value is obtained by summing the response data and dividing the response data by the number of data points, the probability threshold calibration condition is judged, the probability threshold calibration comparison average value is compared with a preset threshold value, and distribution boundary adjustment is determined; Performing intensity deviation calculation according to distribution boundary adjustment, wherein the intensity deviation calculation is obtained by subtracting an interaction intensity value from an average value, so as to obtain an influence distribution prediction result, and optimizing, regulating and controlling a precise adjustment vector form; Extracting an inactivation temperature interval aiming at an optimization regulation and control accurate regulation vector form, and introducing a support vector machine algorithm to process interval data, wherein the support vector machine algorithm takes the interval data as input, and outputs a classification result by maximizing a classification boundary to obtain temperature influence distribution; And extracting peak points from the temperature influence distribution, integrating the previous response data, and determining a final inactivation process parameter set.
7. The method for modeling and predicting respiratory tract virus inactivation data driving according to claim 1, wherein S6 comprises: Obtaining simulation input data from a final inactivation process parameter set, and introducing dose response simulation aiming at vaccine process modeling, wherein the dose response simulation maps a response curve through the input data to obtain a preliminary safety index set; And aggregating immune response predictions through the preliminary safety index set, and calculating a validity measurement average value, wherein the validity measurement average value is obtained by dividing response prediction summation by the number of response prediction sample points, and determining distribution boundary adjustment through judgment threshold calibration.
8. The method for modeling and predicting respiratory tract virus inactivation data drive according to claim 7, wherein S6 further comprises: Performing intensity deviation calculation according to distribution boundary adjustment, wherein the intensity deviation calculation is obtained by subtracting intensity values from the average value of the effectiveness index on all dose response nodes, extracting temperature interval data, and obtaining peak point extraction results, wherein the peak point extraction results select maximum points from the interval data; Optimizing a process iteration form aiming at a peak point extraction result, wherein the process iteration form processes by updating parameter vectors, and introduces a support vector machine algorithm to process response data, wherein the support vector machine algorithm takes the response data as input, and the response data comprises safety index output and validity index output obtained under different parameter vectors; outputting a classification boundary through a maximized interval to obtain an optimized parameter set; and extracting an evaluation index from the optimized parameter set, and integrating the previous simulation data to obtain a safe and effective evaluation index set.
9. The method for modeling and predicting respiratory tract virus inactivation data driving according to claim 1, wherein S7 comprises: Acquiring data meeting preset standards from a safe and effective evaluation index set, extracting key control point data in a prediction result aiming at the condition that the indexes meet the standards, and matching the key control point data with the virus inactivation control conditions in a data mapping relation mode to obtain a preliminary process control boundary dividing result; According to the preliminary process regulation boundary dividing result, environmental constraint data in inactivation condition setting is acquired according to the requirement of process parameter adjustment, and the parameter range interval in a process regulation scheme is determined by integrating the content of safety index verification.
10. The method for modeling and predicting respiratory tract virus inactivation data drive according to claim 9, wherein S7 further comprises: For a parameter range interval, obtaining boundary conditions required by control flow optimization, adopting a pre-established mapping table, combining the parameter range interval with inactivation environment constraint condition setting, judging the feasibility of a regulation scheme, and obtaining an optimized process regulation scheme frame; and integrating the result of the validity data acquired from the safety and effectiveness evaluation index set according to the actual demand of virus inactivation control through an optimized process control scheme framework, outputting a complete process control scheme description, and determining a control flow finally used for the respiratory tract virus inactivation process.

Description

Respiratory tract virus inactivation data driven modeling prediction control method Technical Field The invention relates to the technical field of biological medicine, in particular to a respiratory tract virus inactivation data driving modeling prediction control method. Background In the biomedical field, respiratory viruses form a continuous threat to public health safety due to various transmission ways and easy initiation of group infection, and are always important research objects in virology research, vaccine research and biological safety control. Particularly, in the processes of vaccine research and development and virus prevention and control, the virus inactivation technology is used as a key link for guaranteeing the safety and the immune effectiveness, and the scientificity and the controllability of the virus inactivation technology directly influence the quality and the prevention and control effect of a final product. Therefore, the development of systematic studies around the respiratory virus inactivation process has become an important development direction in this field of technology. In the prior art, virus inactivation usually relies on physical or chemical means to disable the infectious ability of the virus by setting specific process conditions. However, due to the complex surface structure of viruses and the highly conformational dependence of the environmental conditions, different inactivation conditions often lead to significant changes in the viral structure and its immunological properties. In practical application, the conventional method is used for determining the inactivation conditions based on experience or single parameter adjustment, and is difficult to comprehensively describe the structural change rule of viruses under the multivariable inactivation process, so that uncertainty exists in the inactivation effect and the immune safety. Further, changes in the critical structure of the viral surface during inactivation may affect the interaction relationship between the virus and the antibody. The prior art lacks systematic analysis means for the correlation between virus structural changes and antibody binding behavior, and is difficult to identify potential unexpected interaction risks in time. In some cases, improper setting of inactivation conditions may induce non-target immune or immune disorder responses, thereby adversely affecting vaccine safety and effectiveness, but the associated risks are often difficult to accurately predict and evaluate at the stage of existing process design. Meanwhile, with the continuous development of virus data scale and calculation means, how to effectively utilize multi-source data to perform modeling analysis on a virus inactivation process is still a weak link in the prior art. The existing scheme generally lacks of system modeling and prediction capability based on data driving, and prospective evaluation and regulation decision of the inactivation process are difficult to realize under the condition of complex process variables, so that the inactivation process has long optimization process period and high trial and error cost. Disclosure of Invention The invention aims to provide a respiratory tract virus inactivation data driven modeling prediction control method, which solves the problems existing in the prior art. In order to achieve the purpose, the invention provides the following technical scheme that the respiratory tract virus inactivation data driving modeling prediction control method comprises the following steps: S1, acquiring initial virus characteristic data from a virus surface structure database through molecular dynamics simulation, and obtaining a structure change matrix of the virus under different inactivation conditions after treatment; S2, extracting characteristics of virus characteristics and antibody binding sites by adopting a support vector machine classifier according to a structure change matrix, and determining a potential unexpected interaction site set; S3, if the potential unexpected interaction site set exceeds a preset threshold, predicting influence distribution of the inactivation process variable on the interaction strength through a random forest regression model, and obtaining an optimized inactivation condition range; S4, obtaining simulated immune response data from the optimized inactivation condition range, judging the immune chaotic response probability, and obtaining a regulation and control accurate adjustment vector; S5, iterating a virus characteristic model by adopting a gradient descent optimization algorithm aiming at the regulation and control accurate adjustment vector, and determining a final inactivation process parameter set; S6, simulating a vaccine development process through a final inactivation process parameter set to obtain a safe and effective evaluation index set; and S7, outputting a complete regulation and control scheme description if the safety and effecti