Metabolic physiological age prediction and aging evaluation method
Through metabolomics detection and multivariate linear regression analysis, a metabolic physiological age prediction model was constructed, which solved the problem of inability to accurately predict physiological age and aging in the existing technology, and achieved accurate assessment and health management of individual aging status.
Patent Information
- Application Number
- CN202411156132.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-22
- Publication Date
- 2024-11-29
AI Technical Summary
Existing physiological age prediction methods such as telomeres length and DNA methylation markers cannot accurately predict individual aging degree, and there are significant differences in the number of immune cells, resulting in insufficient prediction.
Metabolomics is used to detect metabolites in plasma, and a metabolic physiological age prediction model is constructed by combining multiple linear regression analysis. By analyzing metabolites characteristic values, a metabolic physiological age prediction model is constructed, and personalized health management and intervention suggestions are provided.
Accurate prediction of physiological age and accurate assessment of aging degree are achieved, with high specificity, high sensitivity and high accuracy, which can improve health status and inhibit the aging process.
Smart Images

Figure CN119044349A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of life and health, and in particular to a method for predicting metabolic physiological age and assessing aging. Background Art
[0002] Aging is a natural phenomenon in which the functions of cells, tissues and organs of organisms gradually decline over time. It is caused by the combined effects of multiple biological mechanisms such as genetic factors, environmental factors and personal lifestyle. Against the backdrop of accelerating global population aging and continuous improvement in living standards, people are increasingly demanding healthy aging. However, aging-related diseases, including but not limited to cancer, diabetes, autoimmune diseases, cardiovascular diseases and neurodegenerative diseases, have become the main obstacles to the health of the elderly population. These diseases not only limit the quality of life of individuals, but also place a heavy burden on social medical resources.
[0003] At present, the methods for predicting physiological age mainly include telomere length, DNA methylation markers, and the number of immune cells. Telomeres are structures composed of repetitive sequences and related proteins at the ends of chromosomes. Telomeres shorten every time a cell divides. When they shorten to a certain extent, the cell stops dividing, leading to aging and death. However, due to the different division rates of different somatic cells, such as fast division of gastrointestinal mucosal cells and slow division of nerve cells, telomere length cannot accurately predict physiological age. With the deepening of the understanding of the role of epigenetic variation in the aging process, the DNA methylation model is called the epigenetic clock and is considered to be one of the most promising indicators for measuring physiological age. With age, the methylation status of different gene loci changes, and these changes can be used to determine physiological age. Metabolomics mainly studies small molecule metabolites in organisms (such as amino acids, fatty acids, sugars, etc.), which are the end products or intermediates of metabolic pathways. Metabolomics studies the changes in metabolite levels, provides instantaneous metabolic state information, and reflects the metabolic activity of cells or organisms at a specific time point. Epigenetic technology based on methylation studies changes in the gene regulation level, that is, affecting gene expression by adding methyl groups (CH3) on DNA molecules. The DNA methylation model provides longer-term gene expression regulation information, reflecting the regulation pattern of genes over a longer period of time. With age, the number and activity of immune cells decrease, which can cause immune decline and inflammation. People with a healthy immune system will be able to eliminate this inflammation to a certain extent. However, genetic and environmental variations cause many immune characteristics to differ significantly between individuals, and this difference increases with age. Therefore, the number of immune cells is not accurate enough to measure the degree of aging. Therefore, it is urgent to establish a more accurate physiological age prediction method to assess the degree of aging.
[0004] Metabolomics provides an instant snapshot of the metabolic activity of cells, tissues or organisms in a specific state, reflecting the instantaneous biochemical activity and metabolic status. By analyzing the changes in the metabolome spectrum, we can understand the metabolic basis of the disease, changes in physiological state (such as nutritional status, environmental stress, etc.), and the effects of drugs and toxins on the organism. Therefore, metabolomics helps to develop personalized treatment plans and nutritional recommendations by analyzing the changes in metabolites. Studies have shown that plasma metabolites such as tryptophan 2-C-mannoside, phenylalanyl aspartate and adenosine succinate are positively correlated with age, while steroid sulfates such as dehydroepiandrosterone sulfate, androsterone sulfate, testosterone sulfate and epiandrosterone sulfate are negatively correlated with aging. These findings indicate that metabolites have the potential to predict physiological age. Using metabolomics technology to detect metabolic small molecules in plasma can reveal more metabolic markers that are highly correlated with physiological age, build a metabolic physiological age prediction model, and thus accurately assess the degree of aging, with the advantages of high specificity, high sensitivity and high accuracy.
[0005] Therefore, in-depth research on the aging mechanism and the development of methods that can accurately predict physiological age and objectively evaluate the degree of aging are of great significance for delaying the aging process and preventing and reducing the occurrence of aging-related diseases. The present invention aims to provide a comprehensive method based on metabolomics detection and multivariate linear regression analysis to predict metabolic physiological age and objectively evaluate the degree of aging, providing a scientific basis for individualized health management and intervention measures. Summary of the invention
[0006] The technical problem to be solved by the patent of this invention is to overcome the defects and shortcomings of the above-mentioned prior art, and provide a method for detecting metabolites in biological samples based on metabolomics, combining multiple linear regression analysis to construct a metabolic physiological age prediction model, which is used to accurately predict the physiological age of an individual, accurately assess the degree of aging, and make suggestions for improving diet or lifestyle, thereby improving health and inhibiting the aging process.
[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions: The present invention discloses a metabolic physiological age prediction method, comprising the following steps: 1) Prepare sample extracts to obtain organic phase test solutions and aqueous phase test solutions of the samples respectively; 2) Performing liquid chromatography and mass spectrometry on the organic phase test liquid and the aqueous phase test liquid respectively to obtain sample metabolomics data; 3) Metabolomics data processing and metabolite identification; 4) Construct a multivariate linear regression model between metabolite characteristics and actual age: Use the R language's linear model function (lm) to construct a regression model. The mathematical formula of the linear model function is as follows: ;
[0008] in: y is the dependent variable, x1, x2,…, xk are k independent variables, β0 is the intercept term, which is the expected value of the dependent variable when all independent variables are zero. β1, β2, …, βk are regression coefficients, which represent the influence of each independent variable on the dependent variable. ϵ is the error term, which represents the unexplained part between the dependent variable and the independent variable, and is usually assumed to follow a standard normal distribution with a mean of zero; where the independent variable is the metabolite feature of the metabolome spectrum and the dependent variable is the actual age; 5) Optimize the regression coefficients and independent variable combinations of the constructed multiple linear regression model to obtain an optimized multiple linear regression model; 6) Based on the optimized multivariate linear regression model, the predicted value of metabolic physiological age was obtained.
[0009] Preferably, in step 1), the sample extract is obtained by mixing the sample with a pre-cooled solution, and the pre-cooled solution is a mixture of methyl tert-butyl ether:methanol in a volume ratio of 3:1.
[0010] Preferably, in step 1), the preparation method of the organic phase test liquid is as follows: adding a mixture of methanol and water to the sample extract, ultrasonicating, standing, vortexing, centrifuging and stratifying, taking the upper organic phase and drying it, adding a mixture of acetonitrile and isopropanol, ultrasonicating, and centrifuging to obtain a supernatant as the organic phase test liquid.
[0011] Preferably, in step 1), the preparation method of the aqueous phase test liquid is as follows: adding a mixture of methanol and water to the sample extract, ultrasonicating, standing, vortexing, and centrifuging to separate layers, taking the lower aqueous phase, adding methanol to precipitate, centrifuging to obtain the supernatant, drying, adding water, oscillating and mixing, ultrasonicating, and centrifuging to obtain the supernatant as the aqueous phase test liquid.
[0012] Preferably, in step 2), the liquid chromatography conditions of the organic phase are: using a Waters ACQUTTY UPLC ® BEHC8 1.7µm 2.1*100mm column; mobile phase A is an aqueous solution containing 0.1% acetic acid and 1% ammonium acetate; mobile phase B is an acetonitrile-isopropanol (7:3 v / v) solution containing 0.1% acetic acid and 1% ammonium acetate; the separation elution gradient is as follows: 0-12 minutes is 55%-89% mobile phase B, 12-19.5 minutes is 100% mobile phase B.
[0013] Preferably, in step 2), the liquid chromatography conditions of the aqueous phase are: using a Waters ACQUTTY UPLC ® HSS T3 1.8µm 2.1*100mm column; mobile phase A is an aqueous solution containing 0.1% formic acid; mobile phase B is an acetonitrile solution containing 0.1% formic acid. The separation elution gradient is as follows: 0-13 minutes is 1%-70% mobile phase B, 13-18 minutes is 99% mobile phase B.
[0014] Preferably, in step 2), the mass spectrometry conditions are: acquisition in the form of Full MS and Full MS / dd-MS2, and the parameters used by QExactive are as follows: the resolution of the Full MS mode is 70,000, the scanning range is 100-1500m / z, the AGC is 3E+6, and the Maximum IT is 200 milliseconds; in the Full MS / dd-MS2 mode, the resolution of the secondary mass spectrometry is 17,500, the quadrupole window is 1.5m / z, the AGC is 1E+5, the maximum ion injection time is 50 ms, and the HCD relative collision energy is 30eV.
[0015] Preferably, in step 3), the metabolite identification includes: matching the obtained spectral information of the primary parent ion (MS1) and secondary fragment ion (MS2) of the compound with the spectral information of the primary and secondary metabolites in the public database, and detecting and verifying the standard under the same chromatographic and mass spectrometric conditions to qualitatively identify the metabolites.
[0016] Preferably, the public database is selected from a human metabolite database, a metabolomics database, a mass spectrometry database, and a lipid metabolite database.
[0017] Preferably, step 3) further includes metabolite feature correlation analysis and screening. Preferably, the analysis and screening include eliminating metabolite features with correlation higher than a preset threshold value of 0.9.
[0018] Preferably, step 5) includes optimizing by minimizing mean absolute error (MAE) and stepwise regression analysis.
[0019] Preferably, step 5) includes using leave-one-out cross-validation to validate the multivariate linear regression prediction model.
[0020] Preferably, it further includes an individual aging degree assessment operation.
[0021] Preferably, the individual aging degree is assessed as follows: the aging rate is defined as the ratio of metabolic physiological predicted age to actual age. When the predicted age is higher than the actual age, the ratio is >1, indicating a fast aging rate; when the predicted age is lower than the actual age, the ratio is <1, indicating a slow aging rate.
[0022] The invention discloses a method for determining aging population.
[0023] The present invention discloses a method for determining an aging population, comprising the following steps: 1) Conducting individual aging assessment on a population according to the method described above to obtain the aging rate values of individuals in the population; 2) Arrange the individual aging rate values from small to large to obtain the first quartile and the third quartile; 3) Determination of aging: In the population, individuals whose aging rate is greater than the third quartile of the group are classified as fast aging; individuals whose aging rate is less than the first quartile of the group are classified as slow aging; individuals whose aging rate is between the first quartile and the third quartile are classified as normal aging.
[0024] The invention discloses a method for determining individual aging in a crowd.
[0025] Preferably, based on the above-mentioned calculation formula for the individual aging rate, the aging rate of the population can be obtained, and after sorting from small to large, the first quartile and the third quartile are obtained. In this population, an individual whose aging rate is greater than the third quartile of the group is considered to have an aging rate higher than the general range in this group and belongs to the fast-aging population; an individual whose aging rate is less than the first quartile of the group is considered to have an aging rate lower than the general range in this group and belongs to the slow-aging population; an individual whose aging rate is between the first quartile and the third quartile is considered to have an aging rate between the general range in this group and belongs to the general aging population.
[0026] Preferably, the sample is selected from blood, serum, body fluid or urine.
[0027] Compared with the prior art, the present invention uses characteristic values of plasma metabolites, which are more comprehensive and closer to the phenotype than other methods, and can therefore more comprehensively reflect the aging status of individuals and explain the degree of aging; the multiple linear regression analysis method has a fast modeling speed when facing large data sets, the calculation method is not cumbersome, and it is good at obtaining linear relationships in the data set, so the method combined with multiple linear regression analysis has a higher accuracy in predicting metabolic physiological age. BRIEF DESCRIPTION OF THE DRAWINGS
[0028] Figure 1 . Flowchart for metabolic physiological age prediction.
[0029] Figure 2 .Residual data analysis results graph.
[0030] Figure 3 .Diagram of the Leave-one-out Cross-Validation method.
[0031] Figure 4 .Linear correlation analysis between metabolic physiological predicted age (vertical axis) and actual age (horizontal axis) of 236 modeling group subjects. DETAILED DESCRIPTION
[0032] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.
[0033] Example 1 Subject Condition and Sample Collection 1) Inclusion criteria: (1) Male or female aged ≥18 years; (2) read and fully understood the patient instructions and signed the informed consent form; (3) according to the questionnaire, subjects with common chronic diseases (hypertension, diabetes, coronary heart disease), history of tumor treatment and major surgery were excluded; (4) self-reported no obvious clinical symptoms; (5) physical examination results were within the normal range and no obvious abnormalities were found; (6) examination results were beyond the normal range or abnormal, but the doctor judged that there was no clinical significance (physical examination items included: low-dose spiral CT (abbreviated as LDCT), abdominal color Doppler ultrasound, 4 tumor markers, blood pressure, etc.); subjects who met the above 6 conditions at the same time were considered healthy subjects; 2) Sample information: A total of 286 plasma samples were collected from subjects, including 236 in the modeling group and 50 in the validation group (Table 1). Plasma samples were collected in the early morning on an empty stomach and all collected plasma samples were stored in a -80 degree refrigerator.
[0034] Table 1. Subjects in the modeling group and validation group Modeling Group Validation Group Number of people 236 50 Gender (male / female) 112 / 124 18 / 32 Age (range) 20-65 19-64 Example 2 Prediction of metabolic physiological age and aging assessment 2.1 Metabolite detection in plasma samples 1) Reagents: Methanol, acetonitrile, water, acetic acid, isopropanol of mass spectrometry grade purity, formic acid, ammonium acetate, and methyl tert-butyl ether of HPLC grade purity were purchased from Sigma-Aldrich, USA; 2) Sample preparation: Take 100 μL of plasma and place it in 1000 μL of pre-cooled solution (methyl tert-butyl ether: methanol, volume ratio 3:1), vortex and mix the extracted blood sample to obtain a sample extract; add 500 μL of solution (methanol: water, volume ratio 3:1) to the sample extract, sonicate, let stand, vortex and centrifuge to separate the layers, the upper layer is the organic phase, and the lower layer is the aqueous phase; - Organic phase: After the sample is layered, take 500μL of the upper organic phase into a centrifuge tube, dry it, add 200μL of solution (acetonitrile:isopropanol, volume ratio 3:1), and incubate at room temperature for 15 minutes; after incubation, vortex the centrifuge tube to mix, ultrasonically treat for 5 minutes, and then centrifuge the centrifuge tube for 5 minutes at room temperature (12000 rpm); take 180μl of the supernatant from the centrifuge tube into a 2 mL glass injection vial as the organic phase test solution, and detect it on the machine (LC-MS); - Aqueous phase: After the sample is layered, take 400 μL of the lower aqueous phase to a centrifuge tube, and add 1100 μL of ice methanol to precipitate the protein; after the protein in the centrifuge tube is precipitated, centrifuge the centrifuge tube, take 1000 μL of the supernatant and transfer it to a new centrifuge tube, and dry it overnight; add 200 μL of water to the dried centrifuge tube and incubate it at room temperature for 15 minutes; after incubation, vortex the centrifuge tube to mix, ultrasonically treat it for 5 minutes, and then centrifuge the centrifuge tube at room temperature for 5 minutes (12000rpm); take 180 μL of the supernatant from the centrifuge tube to a 2 mL glass injection vial as the aqueous phase test solution, and detect it on the machine (LC-MS).
[0035] 3) Detection of small molecule metabolites: The organic phase uses a Waters ACQUTTY UPLC ® BEH C8 1.7µm 2.1*100mm column, and the aqueous phase uses a Waters ACQUTTY UPLC ® HSS T3 1.8µm 2.1*100mmcolumn for small molecule separation; the liquid chromatography and mass spectrometer both use the ACQUITY UPLC I-Class liquid chromatography system (Waters) and the Q-Exactive mass spectrometry system (Thermo Fisher Scientific); - Mobile phase parameters for organic phase test liquid: Mobile phase A is an aqueous solution containing 0.1% acetic acid and 1% ammonium acetate; mobile phase B is an acetonitrile-isopropanol (7:3 v / v) solution containing 0.1% acetic acid and 1% ammonium acetate, and the separation elution gradient is as follows: 0-12 minutes is 55%-89% mobile phase B, 12-19.5 minutes is 100% mobile phase B; - Mobile phase parameters for aqueous test liquid: Mobile phase A is an aqueous solution containing 0.1% formic acid; Mobile phase B is an acetonitrile solution containing 0.1% formic acid. The separation elution gradient is as follows: 0-13 minutes is 1%-70% mobile phase B, 13-18 minutes is 99% mobile phase B; - Mass spectrometry parameters: Mass spectrometry data were collected in Full MS and Full MS / dd-MS2 modes (each including positive and negative modes). The parameters used by Q Exactive are as follows: Full MS mode resolution is 70,000, scan range is 100-1500 m / z, AGC is 3E+6, and Maximum IT is 200 milliseconds; in Full MS / dd-MS2 mode, the resolution of the secondary mass spectrometry is 17,500, the quadrupole window is 1.5 m / z, AGC is 1E+5, the maximum ion injection time is 50 ms, and the HCD relative collision energy is 30 eV; 2.2 Metabolomics data processing and metabolite identification 1) Metabolomics data processing: (1) Extract the peaks from the RAW format file of the mass spectrometer and convert it into a FeatureXML format file to reduce the dimension of the original mass spectrometry data and improve the signal-to-noise ratio; (2) Use the peak alignment algorithm of the OpenMS software to calibrate and align the retention time of the extracted peak format data between samples, thereby converting the mass spectrometry data into a data matrix; (3) Match and filter the isotope peaks in the data matrix obtained in step 2, and then replace the abnormal data (0, negative values, background noise, etc.) with vacant values; (4) Among all the characteristic peaks obtained in step 3, those with a detection rate of <80% are eliminated, and those with a detection rate of >80% are filled with the median value of the characteristic peak, and 5% random noise (obeying the standard normal distribution) is added; (5) In order to reduce the difference in metabolite concentrations between samples and make the data distribution more symmetrical, the Normalization Autoencoder (NormAE) is used for homogenization processing to remove systematic errors such as batch effects; 2) Identification of metabolites: The spectral information of the primary parent ion (MS1) and secondary fragment ion (MS2) of the compound is obtained after the software is parsed based on the original data. The metabolites are qualitatively identified by matching the spectral information of the primary and secondary metabolites in the public database, and the standard samples are tested and verified under the same chromatographic and mass spectrometric conditions. Commonly used metabolite databases include the Human Metabolite Database (HMDB, www.hmdb.ca), the Metabolomics Database (Metlin, metlin.scripps.edu), the Mass Spectrum Database (www.massbank.jp), and the Lipid Metabolite Database (Lipidmap, www.lipidmaps.org); Based on the metabolites identified in the relevant databases, the metabolites are finally verified according to the retention time, MS1, and MS2 mass spectrometric information when the standard samples are separated under the same chromatographic column and mass spectrometric conditions. The criteria for metabolite identification are that the retention time is within 0.1min difference, and the theoretical value and measured value of the metabolite molecular weight are less than 10ppm.
[0036] 2.3 Construction of metabolic physiological age prediction model 1) Correlation analysis: In order to reduce the risk of overfitting and improve the stability of the prediction model, features with correlations higher than the preset threshold of 0.9 will be removed. Specifically: |R i ,j |≥α,for i,j=1,2,…,n; Among them, for n features X1, X2...X n , their correlation coefficient matrix is R, R i,j Represents feature X i and X j The correlation coefficient between them is set according to the correlation threshold α. If two features X i and X j The correlation coefficient R i,j If the absolute value of is greater than or equal to α, they are considered to be highly correlated feature pairs. Features that appear in multiple highly correlated feature pairs are preferentially eliminated; therefore, the correlation between each metabolite feature is first calculated, and all feature pairs with R values greater than 0.9 are extracted, starting with the metabolite feature with the most participation times, until the correlation of any pair of metabolite features is less than the threshold of 0.9; then, the screened metabolite features are used as independent variables, and the actual age is used as the dependent variable for multivariate linear regression analysis; 2) Construction of multiple linear regression model: Using the linear model fitting function (lm) of R language, a multiple linear regression model was constructed for the metabolome spectra of the subjects in the modeling group. The independent variable was the metabolite characteristics of the metabolome spectrum, and the dependent variable was the actual age. A multiple linear regression model between the metabolite characteristics and the actual age was constructed. Multiple linear regression is a commonly used data analysis method in statistics, which is usually used to study the relationship between two or more variables and to explain the relationship between a dependent variable and multiple independent variables. Its mathematical formula is as follows: ;
[0037] in: y is the dependent variable; x1,x2,…, x k are k independent variables; β0 is the intercept term, which is the expected value of the dependent variable when all independent variables are zero; β1,β2,…,β k is the regression coefficient, which indicates the effect of each independent variable on the dependent variable; ϵ is the error term, which represents the unexplained difference between the dependent variable and the independent variable, and is usually assumed to follow a standard normal distribution with mean zero.
[0038] The multiple linear regression model has been tested and evaluated through several parameters to determine whether the model can be used, including R square and mean absolute error (MAE), specifically: 2.1) R-squared (R 2 ) is used to measure the ability of the model to explain the variation of the dependent variable, and its value range is 0 to 1. Formula: ; · y i is the true value of the i-th sample; · y^ i is the predicted value of the i-th sample; · y - i is the mean of the sample set; Among them, when R 2 When it is close to 1, it indicates that the model can well explain the variation of the dependent variable and the fitting effect is good; when R 2 When it is close to 0, it means that the model has a weak ability to explain the variation of the dependent variable and the fitting effect is poor.
[0039] 2.2) Mean Absolute Error (MAE): Measures the average absolute error between the predicted value and the true value.
[0040] formula: ; · y i is the true value of the i-th sample; · y^ i is the predicted value of the i-th sample; n is the total number of samples.
[0041] When fitting the regression model on the modeling group data, the model parameters were trained by minimizing the mean absolute error (MAE), i.e., the loss function, to obtain the optimal regression coefficient and its prediction model, determine the linear correlation between the metabolite characteristics and the actual age, and calculate the R 2 To evaluate the goodness of fit of the regression model. 2 The closer to 1 means that the linear regression model explains the variability of its actual age based on the given metabolite characteristics better, that is, the stronger the degree of data point fitting is, and the closer the MAE value is to 0, the smaller the difference between the model's predicted value and the actual value is, indicating that the prediction performance of the linear regression model is better. Specifically, first randomly assign regression coefficients β0, β1, …, β k The initial value of is used, and then the stochastic gradient descent method is used to iteratively optimize the above regression coefficients. This step is repeated 30,000 times or until the difference in the loss function value decreases within 5 consecutive iterations is less than 0.01. At this time, the repetition is terminated to obtain the trained linear regression model parameters and prediction model; 3) Construction of metabolic physiological age prediction model and reliability assessment: Based on the multivariate linear regression model fitted in step 2), the combination of independent variables is further optimized through stepwise regression analysis. The specific operations are as follows: 3.1) Perform an F test on the independent variables with non-zero coefficients in the multivariate linear regression model established in step 2), and record the corresponding F1,…Fk values. Since there is a Fi lower than the F(1,df) threshold, the feature combination can be further optimized and the independent variable with the lowest F value can be eliminated. Specific method: The degrees of freedom (df) of the model is the number of data that can change freely during the model parameter estimation process.
[0042] formula: df = nk-1 n is the sample size; k is the number of independent variables with non-zero coefficients in the model; In the established multiple linear regression model, df is automatically generated after each modeling. According to the F-test critical value table with a significance level of 0.05, look up the threshold corresponding to F(1, df) (the F-test critical value table is a well-known table, http: / / socr.ucla.edu / Applets.dir / F_Table.html). The method for looking up the threshold corresponding to F(1, df) is as follows: the 1 in front refers to the first column, so from the numbers in the first column, select the first column value on the far left corresponding to df. Compare all the F-test F1, … F k values with the threshold corresponding to F(1, df). If there is an F-test value lower than the threshold corresponding to F(1, df),剔除 the independent variable with the lowest F value.
[0043] 3.2) Re-establish the multiple linear regression model with the updated independent variables based on the method in step 2), and repeat step 3.1) until the F-test values F corresponding to all independent variables with non-zero coefficients i are all greater than F(1, df). This indicates that the model has obtained the optimal combination of independent variables, and a multiple linear regression prediction model is obtained. The comparison results of some parameters between the original regression model and the optimized model are shown in Table 2.
[0044] Table 2. Comparison of Some Parameters between the Original Model and the Optimized Model parameter Original Model Optimizing the model Residual standard error 9.218 6.304 p-value 6.375e-05 2.2e-16 <![CDATA[R 2 ]]> 0.8963 0.9399 F-test value p<0.05 p<0.05 Furthermore, conduct reliability assessment on the final multiple linear regression prediction model through residual analysis, normality analysis, and collinearity analysis. The residual standard error is a commonly used statistic for evaluating the fitting degree of a regression model. The closer the value is to 0, the closer the prediction result of the model is to the true value (the results are shown in Figure 2 ). The Shapiro-Wilk normality test result shows that p > 0.05 is not sufficient to reject the null hypothesis, that is, the independent variables all follow a normal distribution, meeting the prerequisite assumptions of linear regression analysis. Collinearity will interfere with the true contribution factors of the independent variables in the linear regression model. Therefore, evaluate the correlation between each independent variable by calculating the variance inflation factors (VIFs) of the model to reduce the impact of collinearity. The test results show that the maximum VIF value of the model is 4.60 < 5, and p < 0.05, indicating that collinearity will not damage the correct estimation of the model coefficients (explanation of the statistic VIF: VIF_i = 1: there is no collinearity between the i-th independent variable and other independent variables; 1 < VIF_i ≤ 5 indicates that there is moderate collinearity between the i-th independent variable and other independent variables, but it does not damage the model evaluation; VIF_i > 5 indicates that there is severe collinearity between the i-th independent variable and other independent variables, resulting in inaccurate estimation of the model coefficients).
[0045] 2.4 Evaluation of the Prediction Results of Metabolic Physiological Age The flow chart for predicting metabolic physiological age is shown in Figure 1 .
[0046] In order to verify the discriminative ability of the multivariate linear regression prediction model between metabolite characteristics and actual age during the construction and optimization process, the Leave-one-out Cross-Validation method was used. Figure 3 As shown, cross-validation means randomly selecting one piece of data from the 236 sample data of the modeling group as the test set, and the remaining 235 pieces as the training set. The 235 samples of the training set are used to build a multivariate linear regression prediction model, and then the samples of the test set are tested to obtain the metabolic physiological age values of the test set samples. Similarly, this process is repeated 236 times, so that each sample in the data set is modeled 235 times as a training set and verified once as a test set. The advantage of this method is that it uses as much data as possible as a training set, and each training set and test set are independent of each other and completely cover the entire data set. Based on the established multivariate linear regression prediction model between metabolite characteristics and actual age, the predicted value of metabolic physiological age is predicted; where Figure 4 The linear correlation analysis between the metabolic physiological predicted age (vertical axis) and the actual age (horizontal axis) of 236 subjects was performed. The results showed that the R² in the linear regression was 0.940, which was close to 1, indicating that the model had a high degree of fit to the data, the linear correlation between the two variables was strong, and the predicted results were more accurate.
[0047] 2.5 Methods for assessing aging Based on the predicted metabolic physiological age, in order to further evaluate the degree of individual aging, we define the aging rate as the ratio of metabolic physiological predicted age to actual age.
[0048] Aging rate = predicted age / actual age When the predicted age is higher than the actual age, the ratio is >1, the aging speed is fast, indicating that the current physical health status is poor; when the predicted age is lower than the actual age, the ratio is <1, the aging speed is slow, indicating that the current physical health status is good. Based on the above individual aging speed calculation formula, the aging speed of 236 subjects in the modeling group can be obtained. After sorting from small to large, the third quartile (75%) and the first quartile (25%) of the aging speed of the modeling group population can be obtained. Among the 236 subjects, the third quartile is 1.059, and the individual whose aging speed is greater than the third quartile of the group is considered to have a higher aging speed than the general range in this group and belongs to the fast aging population; the first quartile is 0.942, and the individual whose aging speed is less than the first quartile of the group is considered to have a lower aging speed than the general range in this group and belongs to the slow aging population; the population whose aging speed is between the first quartile and the third quartile belongs to the general aging population.
[0049] 2.6 Prediction results of validation group sample data Metabolite detection and data processing of the validation group plasma samples were the same as steps 2.1-2.5. After the validation group samples were placed as unknown samples in the metabolic physiological age prediction model constructed above, specific prediction results were obtained (Table 3). Linear correlation analysis was also performed between the metabolic physiological predicted age and the actual age of the 50 validation group subjects. The results showed that the R² in the linear regression was 0.998, which is close to 1, indicating that the linear correlation between the two variables is strong and the prediction results are more accurate.
[0050] Table 3. Metabolic physiological age prediction results Subject number Actual age Predicted age Aging speed Degree of aging error Absolute value of error 1 36 38.7 1.075 Fast aging 2.7 2.7 2 31 30.2 0.974 General aging -0.8 0.8 3 19 18.3 0.963 General aging -0.7 0.7 4 25 24.8 0.992 General aging -0.2 0.2 5 33 35.6 1.079 Fast aging 2.6 2.6 6 32 33.8 1.056 General aging 1.8 1.8 7 26 24.3 0.935 Slow aging -1.7 1.7 8 25 24.3 0.972 General aging -0.7 0.7 9 25 23.3 0.932 Slow aging -1.7 1.7 10 27 28.6 1.059 Fast aging 1.6 1.6 11 26 28.4 1.092 Fast aging 2.4 2.4 12 33 30.8 0.933 Slow aging -2.2 2.2 13 32 32.3 1.009 General aging 0.3 0.3 14 30 29.4 0.980 General aging -0.6 0.6 15 41 40.2 0.980 General aging -0.8 0.8 16 40 39.5 0.988 General aging -0.5 0.5 17 39 40.4 1.036 General aging 1.4 1.4 18 30 30.2 1.007 General aging 0.2 0.2 19 29 31.2 1.076 Fast aging 2.2 2.2 20 38 36.9 0.971 General aging -1.1 1.1 21 37 39.2 1.059 Fast aging 2.1 2.1 22 36 35.4 0.983 General aging -0.6 0.6 23 33 30.5 0.924 Slow aging -2.5 2.5 24 32 29.4 0.919 Slow aging -2.6 2.6 25 28 27.7 0.989 General aging -0.3 0.3 26 27 26.2 0.970 General aging -0.8 0.8 27 26 26.9 1.035 General aging 0.9 0.9 28 35 35.7 1.020 General aging 0.7 0.7 29 34 36.3 1.068 Fast aging 2.3 2.3 30 33 31.6 0.958 General aging -1.4 1.4 31 50 49.4 0.988 General aging -0.6 0.6 32 49 50.4 1.029 General aging 1.4 1.4 33 48 45.2 0.942 Slow aging -2.6 2.6 34 29 31.0 1.069 Fast aging 2.0 2.0 35 28 25.7 0.918 Slow aging -2.3 2.3 36 67 69.4 1.036 General aging 2.4 2.4 37 53 52.4 0.989 General aging -0.6 0.6 38 57 59.3 1.040 General aging 2.3 2.3 39 64 62.5 0.977 General aging -1.5 1.5 40 66 67.8 1.027 General aging 1.8 1.8 41 56 55.9 0.998 General aging -0.1 0.1 42 55 56.4 1.025 General aging 1.4 1.4 43 53 53.5 1.009 General aging 0.5 0.5 44 60 60.8 1.013 General aging 0.8 0.8 45 56 58.0 1.036 General aging 2.0 2.0 46 51 52.9 1.037 General aging 1.9 1.9 47 63 63.3 1.005 General aging 0.3 0.3 48 61 59.2 0.970 General aging -1.8 1.8 49 57 59.0 1.035 General aging 2.0 2.0 50 52 54.6 1.050 General aging 2.6 2.6 The present invention illustrates the process method of the present invention through the above-mentioned embodiments, but the present invention is not limited to the above-mentioned process steps, that is, it does not mean that the present invention must rely on the above-mentioned process steps to be implemented. Those skilled in the art should understand that any improvement of the present invention, equivalent replacement of the raw materials selected by the present invention, addition of auxiliary components, selection of specific methods, etc., all fall within the protection scope and disclosure scope of the present invention.
Claims
1. A method for predicting metabolic physiological age, characterized in that: The steps include: 1) Prepare sample extracts to obtain organic phase test solutions and aqueous phase test solutions of the samples respectively; 2) Performing liquid chromatography and mass spectrometry on the organic phase test liquid and the aqueous phase test liquid respectively to obtain sample metabolomics data; 3) Metabolomics data processing and metabolite identification; 4) Construct a multivariate linear regression model between metabolite characteristics and actual age: Use the R language's linear model function (lm) to construct a regression model. The mathematical formula of the linear model function is as follows: ; in: y is the dependent variable, x1,x2,…, xk are k independent variables, β0 is the intercept term, which is the expected value of the dependent variable when all independent variables are zero. β1, β2, …, βk are regression coefficients, which represent the influence of each independent variable on the dependent variable. ϵ is the error term, which represents the unexplained part between the dependent variable and the independent variable, and is usually assumed to follow a standard normal distribution with a mean of zero; where the independent variable is the metabolite feature of the metabolome spectrum and the dependent variable is the actual age; 5) Optimize the regression coefficients and independent variable combinations of the constructed multiple linear regression model to obtain an optimized multiple linear regression model; 6) Based on the optimized multivariate linear regression model, the predicted value of metabolic physiological age was obtained.
2. The method according to claim 1, characterized in that In step 1), the sample extract is obtained by mixing the sample with a pre-cooled solution, and the pre-cooled solution is a mixture of methyl tert-butyl ether and methanol in a volume ratio of 3:
1.
3. The method according to claim 1, characterized in that In step 3), the metabolite identification includes: matching the obtained spectral information of the primary parent ion (MS1) and secondary fragment ion (MS2) of the compound with the spectral information of the primary and secondary metabolites in the public database, and detecting and verifying the standard under the same chromatographic and mass spectrometric conditions to qualitatively identify the metabolites.
4. The method according to claim 1, characterized in that: Step 3) further includes metabolite feature correlation analysis and screening.
5. The method according to claim 1, characterized in that Step 5) includes optimization using minimization of Mean Absolute Error (MAE) and stepwise regression analysis.
6. The method according to claim 1, characterized in that Step 5) includes using Leave-one-out Cross-Validation to validate the multivariate linear regression prediction model.
7. The method according to claim 1, characterized in that Further includes individual aging degree assessment operation.
8. The method according to claim 7, characterized in that The individual aging degree is assessed as follows: the aging rate is defined as the ratio of metabolic physiological predicted age to actual age. When the predicted age is higher than the actual age, the ratio is >1, indicating a fast aging rate; when the predicted age is lower than the actual age, the ratio is <1, indicating a slow aging rate.
9. The method according to claim 1, characterized in that: The sample is selected from blood, serum, body fluid or urine.
10. A method for determining aging population, characterized in that: The steps include: 1) Performing individual aging assessment on a population according to the method of claim 8 to obtain aging rate values of individuals in the population; 2) Arrange the individual aging rate values from small to large to obtain the first quartile and the third quartile; 3) Determination of aging: In the population, individuals whose aging rate is greater than the third quartile of the population are classified as fast aging; Individuals whose aging rate is less than the first quartile of the population are classified as slow aging; Individuals whose aging rate is between the first and third quartiles are classified as having normal aging.
Citation Information
Cited By
Method and device for measuring and calculating physiological age of human endocrine system, electronic equipment and storage medium
CN121601234A
Metabolin spectrum aging degree prediction method based on non-local variance enhancement
CN122291045A
A Metabolite Profile Aging Prediction Method Based on Nonlocal Variance Enhancement
CN122291045B
