A method for constructing a model using stress response biomarkers as early warning indicators

Abstract

The application relates to the technical field of biological information analysis, and discloses a method for constructing a model by using stress response biomarkers as early warning indexes, which comprises the following method steps: collecting samples, defining research objects and sample amounts, selecting suitable biomarkers, standardizing a test environment, collecting biomarker data under a normal state as a benchmark, continuously measuring the biomarkers, and recording time-varying curves of the biomarkers; extracting and analyzing stress response biomarkers; constructing an early warning model based on the biomarkers; model training needs to divide a data set into a training set and a test set, uses the training set to train the model, and uses the test set to evaluate the performance of the model; and the early warning model is used for predicting new samples, including matching biomarker levels of the new samples with the early warning model by calculation, so that real-time monitoring and early warning of stress states of organisms can be realized, necessary measures can be taken in advance, and damage of stress to organisms is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biological information analysis, in particular to a method for constructing a model by using stress response biomarkers as early warning indicators. BACKGROUND

[0002] With the deepening of biological research, it is found that a series of special biomarkers are produced when organisms are under stress. These biomarkers can effectively reflect the physiological state of organisms and have very high research value. However, how to apply these biomarkers to the early warning system to realize real-time monitoring of the stress state of organisms is still a problem to be solved, and therefore a corresponding technical solution needs to be designed to solve it. SUMMARY

[0003] (I) Technical problems solved

[0004] In view of the deficiencies in the prior art, the present application provides a method for constructing a model by using stress response biomarkers as early warning indicators, which solves the technical problem of how to apply these biomarkers to the early warning system.

[0005] (II) Technical solutions

[0006] In order to achieve the above purpose, the present application is implemented by the following technical solutions: a method for constructing a model by using stress response biomarkers as early warning indicators, comprising the following method steps:

[0007] S1, collecting samples, defining research objects and sample size, selecting appropriate biomarkers, standardizing test environment, collecting biomarker data under normal state as a benchmark, introducing stress stimulation, measuring the change of biomarkers, continuously measuring biomarkers, and recording the time change curve thereof;

[0008] S2, extracting and analyzing stress response biomarkers, preprocessing the collected biomarker data, including removing invalid, missing values or standardizing units, analyzing the biomarker data using statistical methods, calculating the average value and standard deviation of each marker under normal state to reflect the baseline level of the marker, calculating the average value and standard deviation of each marker under stress state, comparing with the normal state, screening out the significantly changed markers as candidate stress response markers, performing correlation analysis on the candidate markers, deleting redundant variables with high correlation, obtaining a less redundant marker set, applying principal component analysis PCA multivariate statistical method to extract the main information in multiple markers, dimension reduction representation, obtaining several comprehensive features sensitive to stress response, establishing a discriminant model based on Logistic regression using these sensitive features to judge whether a sample is in stress state, and evaluating the performance of the discriminant model by cross-validation method to return the best feature set and model parameters;

[0009] S3, constructing an early warning model based on the biomarkers, model training needs to divide the data set into training set and test set, using the training set to train the model, using the test set to evaluate the performance of the model;

[0010] S4, using the early warning model to predict new samples, including matching the biomarker level of the new sample with the early warning model by calculation, to achieve stress early warning.

[0011] Preferably, in step S2, the stress response biomarkers include but are not limited to biomarkers in blood, urine and saliva.

[0012] Preferably, in step S2, the extraction of biomarkers includes using biochemical analysis methods or genomics methods, including ELISA, Western blot or PCR.

[0013] The analysis of biomarkers includes statistical and bioinformatics methods, including t-test, Mann-Whitney U, PCA or cluster analysis.

[0014] Preferably, in step S3, the construction of the early warning model includes using statistical methods, machine learning methods or deep learning methods, and using accuracy, sensitivity, specificity and AUC-ROC index to evaluate the model.

[0015] Preferably, in step S4, the early warning model is a logistic regression model, a decision tree model, a support vector machine model, an artificial neural network model or a deep neural network model.

[0016] Preferably, in step S4, the stress early warning includes early warning level and early warning time.

[0017] Preferably, the stress includes physiological stress, chemical stress and biological stress of organisms.

[0018] Preferably, in step S1, the test environment includes light, temperature and noise, and the sources of mental stress include time pressure, language stimulation and image stimulation.

[0019] (Three) beneficial effects

[0020] Compared with the prior art, the present application has the beneficial effects that: the real-time monitoring and early warning of the stress state of organisms can be realized, so that necessary measures can be taken in advance to reduce the damage of stress to organisms, and the present application can be applied to various types of stress and has wide application prospects; the stress response biomarker is used as an early warning indicator, so that early warning can be performed at the early stage of stress, so that the timing of taking measures is earlier, the early warning effect is better, and the sensitivity is high; an advanced model construction method is adopted, so that the stress state can be accurately predicted according to the change of the biomarker, and the accuracy of early warning is greatly improved; the present application is suitable for various types of stress, including physiological stress, chemical stress and biological stress, has a wide application range and has wide application value; the method steps are simple and easy to implement, and the present application can be widely applied to the fields of biological research, medical diagnosis, environmental monitoring and the like; the early warning model of the present application can be updated and optimized according to new sample data, can adapt to complex and changing stress environments, and has strong adaptability. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0022] The embodiments of the present application provide a technical solution: a model construction method using a stress response biomarker as an early warning indicator, comprising the following method steps:

[0023] S1, collect samples, define the research object and sample size, select appropriate biomarkers, standardize the test environment, collect the biomarker data under normal state as the benchmark, introduce stress stimulation, measure the change of the biomarker, continuously measure the biomarker, and record the time change curve thereof;

[0024] S2, extract and analyze stress response biomarkers, preprocess the collected biomarker data, including removing invalid, missing values or standardizing units, analyze the biomarker data using statistical methods, calculate the average and standard deviation of each marker under normal conditions, reflect the baseline level of the marker, calculate the average and standard deviation of each marker under stress conditions, compare with the normal state, screen out the significantly changed markers as candidate stress response markers, perform correlation analysis on the candidate markers, delete redundant variables with high correlation, obtain a less redundant marker set, apply principal component analysis PCA multivariate statistical method to extract the main information in multiple markers, reduce dimension representation, obtain several comprehensive features sensitive to stress response, use these sensitive features to establish a discriminant model based on Logistic regression to judge whether a sample is in stress state, evaluate the performance of the discriminant model by cross-validation method, return the best feature set and model parameters;

[0025] S3, build an early warning model based on the biomarkers, model training needs to divide the data set into training set and test set, use the training set to train the model, use the test set to evaluate the performance of the model;

[0026] S4, use the early warning model to predict new samples, including matching the biomarker level of the new sample with the early warning model by calculating to realize stress early warning.

[0027] Further improvement, in step S2, the stress response biomarkers include but are not limited to biomarkers in blood, urine, saliva. Define the research object and sample size. Determine the target population of the study, and the sample size to be collected, the sample size should be large enough to ensure the statistical significance of the results.

[0028] Select appropriate biomarkers. Common stress response biomarkers include skin conductance, heart rate, blood pressure, skin temperature, respiratory rate, etc. The markers related to the research target should be selected.

[0029] Standardize the test environment. The light, temperature, noise, etc. of the test site should be controlled to make the subjects in a relatively constant state.

[0030] Collect biomarker data under normal conditions as a benchmark. Measure the biomarkers of the subjects in a relaxed state without external stimulation and record them as normal values.

[0031] Introduce stress stimulation and measure the changes in biomarkers. Different sources of mental stress can be used, such as time pressure, language stimulation, image stimulation, etc. to induce stress response in subjects.

[0032] The biomarkers are continuously measured and the changes over time are recorded. A sufficient period of time before, during and after stimulation is recorded to draw a complete response process.

[0033] Further improved, in step S2, the extracting biomarkers includes using biochemical analysis methods or genomics methods, including ELISA, Western Blot or PCR.

[0034] The analysis of biomarkers includes statistical and bioinformatics methods, including t-test, Mann-Whitney U, PCA or clustering analysis.

[0035] Biochemical analysis methods such as ELISA can specifically detect a certain protein in the sample, and Western Blot can detect multiple proteins. These methods can directly and quickly extract specific protein biomarkers from samples;

[0036] Genomics methods such as PCR can quickly amplify specific DNA sequences for detecting specific gene biomarkers;

[0037] Statistical methods such as t-test can test the difference in expression of two groups of markers, and Mann-Whitney U test can compare the difference distribution of two groups of markers, which can find the different markers;

[0038] Bioinformatics methods such as PCA can reduce the dimensionality of biomarkers, and clustering can detect the similarity between samples, which can effectively analyze and interpret complex high-dimensional biomarker data sets.

[0039] Various markers in biological samples can be obtained with high throughput and high specificity; various algorithms can be used to effectively analyze the expression patterns of biomarkers; and key markers with expression differences between samples can be found for disease diagnosis, etc.

[0040] Further improved, in step S3, the constructing early warning model includes using statistical methods, machine learning methods or deep learning methods, and using accuracy, sensitivity, specificity and AUC-ROC indicators to evaluate the model.

[0041] Model construction method: statistical method, construct logistic regression and other statistical models, calculate probability and coefficient based on sample data;

[0042] Machine learning methods such as SVM and random forest can handle non-linear and high-dimensional relationships;

[0043] Deep learning methods such as CNN can automatically learn features and complex rules;

[0044] Model evaluation indicators: accuracy, reflecting the overall correct degree of judgment;

[0045] Sensitivity, reflecting the proportion of positive classes correctly judged;

[0046] Specificity, reflecting the proportion of negative classes correctly judged;

[0047] AUC-ROC, comprehensive measure of false positive rate and false negative rate, the larger the value, the better the classification performance;

[0048] A high-precision early warning model can be constructed to improve the prediction accuracy;

[0049] Different methods can be combined to take advantage of each other and improve model robustness;

[0050] Comprehensive evaluation indicators can evaluate the model from different angles and select the best model;

[0051] Effective early warning of events can be achieved to shorten the response time.

[0052] Further improvement, in step S4, the early warning model is a logistic regression model, a decision tree model, a support vector machine model, an artificial neural network model or a deep neural network model.

[0053] Logistic regression model, based on statistical probability modeling, can give the probability of event occurrence and realize classification; calculation is simple and easy to understand, but the expression ability of non-linear relationship is weak; decision tree model, through recursive tree building, can represent the non-linear relationship between data; the model has strong interpretability, but may overfit; support vector machine model, based on interval maximization, realizes classification, and has good effect on small sample; can solve non-linear, high-dimensional mode, but needs to select kernel function; artificial neural network model, through network structure modeling, can learn complex non-linear relationship in data; has good fitting effect on high-dimensional data, but has poor interpretability; deep neural network model, deepening network layers, can learn deep features representation in data; has strong fitting effect on high-dimensional complex problems, but needs a large amount of data and parameter tuning.

[0054] Further improvement, in step S4, the stress early warning includes early warning level and early warning time.

[0055] Early warning level, multiple early warning levels can be set, such as general, important, serious, etc.;

[0056] Different levels can trigger different early warning response mechanisms;

[0057] The level can be divided according to the severity of the event to realize graded early warning;

[0058] The warning time can be set in advance, such as 1 hour, 2 hours, etc.

[0059] The warning time should be determined according to the urgency of the event;

[0060] The longer the warning time, the more adequate the response time;

[0061] Implement hierarchical warning, take different strategies for different levels of events;

[0062] Reserve enough response time to enhance the initiative of event handling;

[0063] Improve the pertinence and effectiveness of the warning, and reduce the false positive rate;

[0064] Provide more guidance for decision-making, guide event response.

[0065] Further improvement, the stress includes physiological stress, chemical stress and biological stress of the organism.

[0066] Physiological stress, such as physiological stress caused by environmental conditions such as temperature, radiation, etc.; Stress warning can be achieved through physiological parameter monitoring; Chemical stress, such as poisoning stress caused by chemical pollution such as heavy metals, pesticides, etc.; Stress warning can be achieved through environmental monitoring and biological detection; Biological stress, such as harmful stress caused by biological factors such as pathogens and pests; Stress warning can be achieved through epidemiological monitoring and pathogen detection; Directional monitoring technology can be implemented for different types of stress; The advantages of different technologies can be brought into play to improve the sensitivity and specificity of monitoring and warning; The type and level of stress can be distinguished for hierarchical warning; It is conducive to stress source tracing and risk control; It can improve the warning and response ability of various types of stress and protect the health and safety of organisms.

[0067] Specifically, in step S1, the test environment includes light, temperature and noise, and the stressors include time pressure, language stimulation and image stimulation.

[0068] Air quality: Control dust, pollen, pollutants, etc. in indoor air to ensure air quality; electromagnetic environment: Shield electromagnetic wave interference to avoid the influence of electromagnetic radiation on the subjects; space layout: The test site space should be open and clean to avoid making the subjects feel cramped and nervous; physical contact: Avoid physical contact with the subjects during the measurement to reduce possible interference; mental state: Investigate the mental state of the subjects through questionnaires to exclude subjects with extremely unstable emotions; drug use: Understand whether the subjects are taking drugs that may affect biomarkers; diet control: Control the subjects' diet before the test to avoid affecting the results due to fasting, satiety, etc.; physical activity: Rest the subjects before the test to avoid the influence of large physical activity on physiological indicators; same-day plan: Try to arrange the subjects to take the test in the same time period to reduce the influence of intraday changes.

[0069] Arithmetic calculation stress: Let the subjects complete complex arithmetic operations within a limited time to increase cognitive load stress; cold water stimulation: Let the subjects immerse their hands in cold water to produce a stress response through temperature stimulation; body posture stimulation: Use a method of maintaining an uncomfortable body posture for a certain period of time to stimulate the subjects; social interaction stress: Set up a situation where the subjects need to speak or interact in front of strangers; fear stimulation: Use images, videos, or scenes related to the subjects' fear sources to stimulate the subjects; audio stimulation: Use noisy and harsh audio as a stimulus source; sleep deprivation: Control the subjects' sleep time to cause mild sleep deprivation; social exclusion: Set up a simulation scenario where the subjects are excluded or isolated by their peers; emotional images: Use images with specific emotional colors to stimulate the subjects; the comprehensive use of different types and intensities of mental stressors can trigger stable stress responses in the subjects.

[0070] Early warning source: Clearly indicate which biomarker parameters triggered the early warning, such as heart rate, blood pressure, skin conductance, etc.; early warning accuracy: Give the credibility of the system's judgment of the early warning, such as 90% certainty; early warning explanation: Analyze the early warning and explain what factors caused the stress response; early warning duration: Monitor how long the early warning state lasts to determine the degree of warning; early warning trend: Determine whether the early warning signal is escalating to predict the possibility of increasing stress; self-defined early warning threshold: Allow setting the early warning threshold according to individual differences; early warning repetition rate: Statistically analyze the frequency of repeated occurrence of a certain early warning signal; early warning management suggestion: Provide disposal suggestions for relieving stress corresponding to the early warning level; early warning history record: Store early warning records for signal analysis and model optimization; early warning feedback channel: The output method of the early warning signal, such as audio, video, and tactile feedback; by considering these factors, a more targeted and practical stress early warning system can be designed.

[0071] Real-time monitoring and early warning of the stress state of organisms can be realized, so that necessary measures can be taken in advance to reduce the damage of stress to organisms, and meanwhile, the present application can be applied to various types of stress, and has wide application prospects.

[0072] The present application uses stress response biomarkers as early warning indicators, can early warn at the early stage of stress, so that the timing of taking measures is earlier, the early warning effect is better, and the sensitivity is high, advanced model construction methods are adopted, the stress state can be accurately predicted according to the change of biomarkers, the accuracy of early warning is greatly improved, the present application is suitable for various types of stress, including physiological stress, chemical stress and biological stress, has wide application range, and has wide application value, the method steps are simple and easy to implement, can be widely applied to biological research, medical diagnosis, environmental monitoring and other fields, the early warning model of the present application can be updated and optimized according to new sample data, can adapt to complex and changing stress environment, and has strong adaptability.

[0073] The above shows and describes the basic principles and main features of the present application and the advantages of the present application, for those skilled in the art, it is obvious that the present application is not limited to the details of the above exemplary embodiments, and the present application can be realized in other specific forms without departing from the spirit or basic characteristics of the present application. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-limiting, the scope of the present application is defined by the appended claims rather than the above description, and therefore all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application.

[0074] In addition, it should be understood that, although the present application is described in the specification, each embodiment only contains one independent technical solution, and the description manner of the specification is only for the sake of clarity, those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be properly combined to form other embodiments that those skilled in the art can understand.

Claims

1. A method of constructing a model using stress response biomarkers as early warning indicators, characterized by, The method comprises the following steps: S1, collecting samples, defining subjects and sample size, selecting appropriate biomarkers, standardizing test environment, collecting biomarker data in normal state as a benchmark, introducing stress stimulation, measuring changes in biomarkers, continuously measuring biomarkers, and recording their time-varying curves; S2, extracting and analyzing stress response biomarkers, preprocessing the collected biomarker data, including removing invalid, missing values or standardizing units, analyzing biomarker data using statistical methods, calculating the average and standard deviation of each marker in the normal state to reflect the baseline level of the marker, calculating the average and standard deviation of each marker in the stress state, comparing with the normal state, screening out significantly changed markers as candidate stress response markers, performing correlation analysis on candidate markers, deleting highly correlated redundant variables, obtaining a less redundant marker set, applying principal component analysis PCA multivariate statistical method to extract the main information in multiple markers, dimension reduction representation, obtaining several comprehensive features sensitive to stress response, using these sensitive features to establish a discriminant model based on Logistic regression to judge whether a sample is in a stress state, and evaluating the performance of the discriminant model using cross-validation method to return the best feature set and model parameters; S3, constructing a warning model based on the biomarkers, model training needs to divide the data set into training set and test set, using the training set to train the model, and using the test set to evaluate the performance of the model; S4, using the warning model to predict new samples, including matching the biomarker level of the new sample with the warning model by calculation to realize stress warning, The form of the Logistic regression model is P(Y=1|X) = 1 / (1+e^-(b0+b1x1+...+bpxp)) Where: Y: the state of the sample, 1 indicates stress, and 0 indicates normal, P(Y=1|X): the probability of the sample being in a stress state given the feature X, b0, b1...bp: parameters of the Logistic regression model, x1, x2...xp: p feature variables of the sample, For a new sample, we first extract its biomarker data, preprocess it in accordance with the model to obtain the feature vector X = [x1, x2,... xp], Then bring X into the Logistic model to calculate P(Y=1|X), the probability of the sample being in a stress state, Compare with the preset threshold τ, when P(Y=1|X)>τ, the model judges that the sample is in a stress state, triggering the warning, τ is taken near 0.5, or adjusted through cross-validation results to balance the false positive rate and the false negative rate, The calculation formula is summarized as: If P(Y=1|X) > τ, Where P(Y=1|X) is calculated by the Logistic regression model.

2. The method according to claim 1, wherein the model is constructed by using the stress response biomarker as an early warning indicator. In step S2, the stress response biomarkers include but are not limited to biomarkers in blood, urine, and saliva.

3. The method according to claim 2, wherein the model is constructed by using the stress response biomarker as an early warning indicator. In step S2, the biomarker extraction includes biochemical analysis or genomics methods, including ELISA, Western blot or PCR. The biomarker analysis includes statistical and bioinformatics methods, including t-test, Mann-Whitney U, PCA or cluster analysis.

4. The method according to claim 1, wherein the model is constructed by using stress response biomarkers as early warning indicators. In step S3, the early warning model construction includes statistical methods, machine learning methods or deep learning methods, and the model is evaluated by accuracy, sensitivity, specificity and AUC-ROC.

5. The method according to claim 1, wherein the model is constructed using the stress response biomarker as an early warning indicator. In step S4, the early warning model is a logistic regression model, a decision tree model, a support vector machine model, an artificial neural network model or a deep neural network model.

6. The method according to claim 1, wherein the model is constructed using the stress response biomarker as an early warning indicator. In step S4, the stress early warning includes early warning level and early warning time, and the stress early warning system further includes early warning source, early warning accuracy, early warning explanation, early warning duration, early warning trend, custom early warning threshold, early warning repetition rate, early warning management suggestion, early warning history record or early warning feedback channel.

7. The method according to claim 6, wherein the model is constructed by using the stress response biomarker as an early warning indicator. The stress includes physiological stress, chemical stress and biological stress of the organism.

8. The method according to claim 1, wherein the model is constructed using the stress response biomarker as an early warning indicator. In step S1, the test environment includes light, temperature and noise, and the mental stressors include time pressure, language stimuli and image stimuli.

9. The method according to claim 8, wherein the model is constructed by using the stress response biomarker as an early warning indicator. The test environment further includes air quality, electromagnetic environment, space layout, physical contact, mental state, medication, dietary control, physical activity or same-day plan.

10. The method according to claim 8, wherein the model is constructed by using stress response biomarkers as early warning indicators. The mental stressors as stress stimuli further include arithmetic calculation stress, cold water stimulation, body posture stimulation, social interaction stress, fear stimulation, audio stimulation, sleep deprivation, social exclusion or emotional image.