Method for male biological age assessment and aging prediction based on DNA methylation data of peripheral blood

Abstract

The invention discloses a method for male biological age assessment and aging prediction based on DNA methylation data of peripheral blood. The method comprises the following steps: screening out DNA methylation sites related to male biological age and senescence by retrieving related literatures and research results; downloading DNA methylation chip sequencing data, and screening out male data; screening the modeling database and the DNA methylation sites to obtain a modeling sample and modeling sites; establishing an epigenetic clock model; comparing the epigenetic clock model with a traditional model, and performing model evaluation; carrying out biological age evaluation and aging prediction; according to the method, the regression models are established according to three methods based on screening of global male samples, few loci with top correlation ranking are selected from the regression models, few locus models are re-established to evaluate the age and the aging degree of people, and the method has the advantages of accurate male sample positioning, low prediction cost and convenience in later product popularization.

Description

Technical Field

[0001] The present invention relates to the field of life and health technology, and in particular to a method for assessing male biological age and predicting aging based on DNA methylation data of peripheral blood. Background Art

[0002] Population aging is a major challenge facing all of humanity today. According to statistics, by the end of 2023, China's population aged 60 and above will reach 297 million, accounting for 21.1% of the total population. The growth of population aging has attracted more attention to the aging process and the risk factors that mediate age-related health deterioration and promote age-related diseases. Age is the end of the human lifespan, which is characterized by functional decline and increased incidence of age-related degenerative diseases such as diabetes, cancer, cardiovascular disease and dementia. At the molecular level, aging is associated with a gradual loss of physiological, organismal and cellular integrity, leading to genomic instability in cells, organismal decompensation and ultimately death.

[0003] Among the various hallmarks of aging, epigenetics plays a crucial role in aging and age-related diseases. Studies have found that the functional aging of cells or tissues is associated with their epigenomic landscape. It is now generally accepted that impaired epigenetic marks contribute to the study of the relationship between biological aging, chronological age, and aging-related diseases. The epigenetic clock is a tool developed through machine learning that performs a multivariate weighted sum of cytosine methylation at cytosine-phosphate-guanine (CpG) sites throughout the genome to predict age or other phenotypic features of aging. Epigenetic clock measurements can accurately predict actual age, age-related diseases, and mortality. Therefore, the epigenetic clock can be used as a molecular biomarker of aging to assess the degree of aging in the individual being tested.

[0004] Studies have shown that there are significant pattern differences between men and women during the aging process. Specifically, the average life expectancy of men is about 5 years shorter than that of women. However, the so-called "male-female health-survival paradox" reveals a phenomenon: although women have a longer lifespan, the incidence of frailty and disability is relatively high in the elderly female group. This gender difference is partly due to different biological characteristics, such as differences in hormone levels and immune system function. Further studies have shown that during the aging process, the DNA methylation status of peripheral blood cells in men and women is also significantly different, reflecting gender differences in gene expression and epigenetic regulation. For example, the ELOVL2 gene plays an important role in the synthesis of long-chain polyunsaturated fatty acids, and its methylation level is highly correlated with age. It is considered to be one of the biomarkers of aging. Although changes in ELOVL2 methylation have been observed in both sexes, existing studies have shown that changes in the methylation level of this gene in men are more closely associated with age-related health problems such as cardiovascular disease. Another example is the KLF14 gene, which is involved in metabolic regulation and its methylation status is associated with metabolic syndrome. In men, changes in the methylation pattern of KLF14 are associated with fat distribution and insulin resistance, and these changes are particularly evident in the elderly male population. Animal experiments have further confirmed that gender not only affects the rate of aging, but also has a significant impact on a variety of anti-aging interventions, including dietary restrictions, genetic manipulation, and drug preparations. These research results emphasize the essential differences between men and women in the aging process and support the importance and rationality of constructing DNA methylation clocks for different sexes for accurately predicting individual biological age.

[0005] However, the existing biological age assessment and aging prediction methods are based on both male and female samples, which will inevitably reduce the accuracy of age assessment and aging prediction for male samples. In addition, in the existing prediction methods, more sites are used, resulting in higher prediction costs, which is not conducive to the promotion of later products. Summary of the invention

[0006] The technical task of the present invention is to address the above shortcomings and provide a method for male biological age assessment and aging prediction based on peripheral blood DNA methylation data. Based on the screening of global male samples, a penalized multivariate regression model, a stepwise linear regression model and a gradient boosting regression model of the elastic network are established respectively, and a small number of sites with high correlation rankings are selected from them, and a small number of site models are re-established to assess the age and aging degree of the population. It has the advantages of accurately positioning male samples, low prediction cost, and convenient for later product promotion.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A method for assessing male biological age and predicting aging based on DNA methylation data in peripheral blood comprises the following steps:

[0009] S1. Screening of DNA methylation sites: DNA methylation sites related to male biological age and aging were screened by searching relevant literature and research results on PubMed and Illumina methylation array platforms;

[0010] S2. Acquisition of modeling database: Download DNA methylation chip sequencing data from the GEO website, filter out male data, and obtain the modeling database;

[0011] S3. Determination of modeling samples and modeling sites: by calculating the sample detection rate and site detection rate, the modeling database and DNA methylation sites are screened to obtain modeling samples and modeling sites;

[0012] S4. Construction of epigenetic clock model: Use modeling samples and modeling sites to build an epigenetic clock model;

[0013] S5. Model comparison and evaluation: Compare the epigenetic clock model with the traditional model and perform model evaluation;

[0014] S6. Age assessment and aging prediction: Select external samples to conduct biological age assessment and aging prediction.

[0015] Furthermore, the DNA methylation sites associated with male biological age and aging described in S1 are cg01511567, cg01820374, cg02154074, cg04084157, cg04528819, cg04836038, cg05294243, cg05442902, cg06493994, cg07211259, cg09809672, cg13460409, cg13854874, cg13975369, cg14175438, cg16386080, cg17324128, cg19008809, cg19283806, cg19722847, cg19761273, cg22736354, cg22947000, cg25809905, cg26394940, cg26614073, cg26824091, cg26842024, cg27015931.

[0016] Furthermore, the DNA methylation chip sequencing data described in S2 includes country, biological age, gender, and peripheral blood DNA methylation chip sequencing data, wherein the peripheral blood DNA methylation chip sequencing data includes Illumina HumanMethylation 450K chip and Illumina Human Methylation 850K chip.

[0017] Furthermore, the sample detection rate and site detection rate described in S3 are calculated by a python script.

[0018] Furthermore, S4 also includes filling in the undetected data for the modeling samples using the makeX function included in the R package glmnet.

[0019] Furthermore, the epigenetic clock models described in S4 are a penalized multivariate regression model of an elastic network, a stepwise linear regression model, and a gradient boosting regression model.

[0020] Furthermore, S4 also includes randomly selecting a smaller number of sites from the modeling sites for re-modeling, screening the sites that are most significantly correlated with age based on the determination coefficient and the mean absolute error of the regression model evaluation indicators, and re-establishing the epigenetic clock model to obtain an epigenetic clock model with fewer sites, which is used to reduce the prediction cost while ensuring the prediction accuracy.

[0021] Furthermore, the model evaluation method described in S5 includes but is not limited to the mean absolute error between the predicted age and the biological age, the fit metric of the linear regression model, and the number of model sites.

[0022] Furthermore, the traditional models described in S5 are the Horvath S2018 model and the Levine M2018 model.

[0023] Furthermore, S6 also includes defining the aging rate of an individual as the difference between the predicted age and the biological age.

[0024] Compared with the prior art, the advantages and positive effects of the present invention are:

[0025] 1. The modeling data of the present invention is selected from male samples worldwide. The epigenetic clock model established based on male samples is more accurately targeted at the male population, and the biological age of men is more accurately assessed and aging is predicted;

[0026] 2. The present invention establishes three epigenetic clock models, namely, the penalized multivariate regression model of elastic network, the stepwise linear regression model and the gradient boosting regression model, and compares the three epigenetic clock models to select the optimal model for prediction;

[0027] 3. Based on the establishment of three epigenetic clock models, the present invention further simplifies the modeling sites according to the importance of the sites related to age, establishes an epigenetic clock model with fewer sites, reduces costs while ensuring prediction accuracy, and is more conducive to the promotion and application of later products;

[0028] 4. The present invention compares the established epigenetic clock model with the existing prediction model and verifies the epigenetic clock model using external samples, further illustrating the accuracy of age assessment and aging prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0030] Figure 1 is a flow chart of an embodiment of the present invention;

[0031] Figure 2 is a penalized multivariate regression model diagram of an elastic network in an embodiment of the present invention;

[0032] Figure 3 is a diagram of a penalized multivariate regression model of a less-site elastic network in an embodiment of the present invention;

[0033] Figure 4 is a stepwise linear regression model diagram of an embodiment of the present invention;

[0034] Figure 5 It is a stepwise linear regression model diagram of fewer sites in an embodiment of the present invention;

[0035] Figure 6 is a gradient boosting regression model diagram of an embodiment of the present invention;

[0036] Figure 7 This is a graph of a gradient boosting regression model with fewer sites in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0038] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0039] A method for assessing male biological age and predicting aging based on DNA methylation data in peripheral blood comprises the following steps:

[0040] S1. Screening of DNA methylation sites:

[0041] Through searching the Public Medicine (PubMed) database, a total of 9 articles related to this study were retrieved, and another one previous research result of this research team was added, a total of 10 articles were included in this technical solution research, among which the search keywords included DNA methylation, CpG, aging, and the search expression was (DNAmethylation[Title / Abstract])AND(CpG[Title / Abstract])AND(aging[Title / Abstract]), and the publication time of the literature was limited to 2013 to 2024, and the language was English. Through searching PubMed, literature related to this study and aging-related research results were selected to select DNA methylation sites, and the existing commonly used Illumina methylation chip platform (MSA 270k, 450k and 850k), and finally 29 DNA methylation sites were identified: cg01511567, cg01820374, cg02154074, cg04084157, cg04528819, cg04836038, cg05294243, cg05442902, cg06493994, cg07211259, cg09809672, cg13460409, cg1385487 4, cg13975369, cg14175438, cg16386080, cg17324128, cg19008809, cg19283806, cg19722847, cg19761273, cg22736354, cg22947000, cg25809905, cg26394940, cg26614073, cg26824091, cg26842024, cg27015931. Specific site annotation information is shown in Table 1.

[0042] Table 1: Annotation information of DNA methylation sites related to male biological age and aging

[0043]

[0044]

[0045] 1 Reference genome version: CRCh37 / h19

[0046] S2. Acquisition of modeling database:

[0047] The DNA methylation chip sequencing data including country, biological age, gender, and peripheral blood were downloaded from the Gene Expression Omnibus (GEO) website, where the DNA methylation chip sequencing data included the Illumina Human Methylation 450K chip and the Illumina Human Methylation 850K chip. The 450K chip can detect approximately 450,000 methylation sites in the human genome, fully covering 96% of CpG islands, and adding CpG sites outside CpG islands, non-CpG methylation sites in human stem cells, differentially methylated sites between normal tissues and tumor tissues, CpG islands outside coding regions, miRNA promoter regions and sites in disease-related regions identified by GWAS; the 850K chip can detect approximately 850,000 methylation sites in the human genome. It is a new generation of DNA methylation chip launched on the basis of the original 450K chip. In terms of data, it includes 91% of the sites of the original 450K chip and adds 413,745 sites, fully covering gene promoter regions, gene coding regions, CpG islands and enhancer regions.

[0048] The GEO database was searched for the raw data of DNA methylation chips from peripheral blood samples of 3530 healthy people, including 2200 males and 1330 females. The detected data were screened for gender and only male samples were retained, resulting in 10 modeling databases. The annotation information is detailed in Table 2.

[0049] Table 2: Annotation information of 10 modeling databases obtained from the Gene Expression Omnibus (GEO) database

[0050]

[0051]

[0052] S3. Determination of modeling samples and modeling sites:

[0053] (1) Sample filtering: The DNA methylation chip sequencing data in the modeling database were merged to obtain the sum of sites common to all samples, a total of 312,979 sites. The detection rate of each sample in the sum of sites was calculated using a python script, and samples with a detection rate greater than 96% were screened. The R package FactoMineR was used to perform principal component analysis (PCA), and samples with an FDR lower than 0.2 were designated as outliers and deleted. Finally, a total of 2,200 samples were obtained. The sample information is shown in Table 3.

[0054] Table 3: Description of modeling samples by age group

[0055] Age group male <10 169 >＝10,<20 23 >＝20,<30 23 >＝30,<40 329 >＝40,<50 653 >＝50,<60 542 >＝60,<70 330 >＝70,<80 93 >＝80,<90 19 >＝90 19 total 2200

[0056] (2) Site filtering: The detection rate of each site was calculated using a python script, and sites with a detection rate greater than 95% were screened, totaling 312,840 sites. The detection rates of the 29 sites used in the modeling were all qualified.

[0057] S4. Construction of epigenetic clock model:

[0058] A variety of methods were used to establish the epigenetic clock model, including the penalized multivariate regression method of the elastic network, the stepwise linear regression method, and the gradient boosting regression method. The 29 modeling sites were set as the dependent variable and the biological age was set as the independent variable. An epigenetic clock model containing 29 sites was constructed. Furthermore, a small number of sites among the 29 sites were randomly selected for reconstruction modeling. The coefficient of determination (R 2 By comparing the mean absolute error (R-Square) and the mean absolute error (MAE), a smaller number of sites with the most significant correlation with age were screened out, and an epigenetic clock model with a smaller number of sites was re-established based on the smaller number of sites. By reducing the number of sites, it aims to ensure the accuracy of age assessment and aging prediction, and to improve the operability and economic benefits of the prediction model in practical applications.

[0059] (1) Data imputation: For the modeling samples, the makeX function included in the R package glmnet was used to impute the undetected data.

[0060] (2) Calculate the Pearson correlation coefficient between the modeling site and age, see Table 4 for details.

[0061] Table 4: Pearson correlation coefficients between modeling sites and age

[0062]

[0063]

[0064] Note: The larger the absolute value of the Pearson coefficient, the stronger the correlation with age; if the Pearson coefficient is negative, it is negatively correlated with age, and vice versa.

[0065] (3) Use the R package glmnet to build a model of the elastic net penalized multivariate regression method. The elastic net is a linear regression model that combines the penalty terms of ridge regression and lasso regression to solve multicollinearity and variable selection problems. The penalty term of the elastic net can reduce the absolute value of the regression coefficient or reduce some coefficients to zero, thereby achieving variable selection and model stability.

[0066] In this embodiment, the α (selection parameter for ridge regression and lasso regression) value is set to 1, and the number of candidate λ (regularization parameter) values ​​is 28. The best λ value obtained by cross-validation is 27. The penalty multivariate regression model of the optimal elastic network obtained by fitting is shown in Table 5, and the model result diagram is shown in Figure 2 .

[0067] Table 5: Penalized multivariate regression model with elastic network based on male sample

[0068]

[0069]

[0070] Note: Among them, n is the total number of methylation sites included in the regression model, i is the number of corresponding methylation sites, corr is the correlation coefficient of each methylation site, and beta is the methylation level of each methylation site.

[0071] (4) Establish a penalized multivariate regression model of the elastic network with fewer sites. Randomly select 5-8 sites from the 29 sites for reconstruction modeling, and evaluate the coefficient of determination (R 2 , R-Square) and mean absolute error (MAE), from which the most significant correlation with age was screened out, and the penalized multivariate regression model of the elastic network of the few sites was re-established based on the few sites. Since the model could not be established at 5-8 sites in this model, 17 sites were selected as the minimum site. The few-site model is shown in Table 6, and the model prediction results are shown in Figure 6. Figure 3 .

[0072] Table 6: Penalized multivariate regression model based on the elastic network with fewer sites for male samples

[0073]

[0074]

[0075] Note: Among them, n is the total number of methylation sites included in the regression model, i is the number of corresponding methylation sites, corr is the correlation coefficient of each methylation site, and beta is the methylation level of each methylation site.

[0076] (5) The R package MASS was used to establish a stepwise linear regression model. Stepwise linear regression is a technique used for model selection. When establishing a linear regression model, it selects the optimal feature subset through automated steps and uses the Akaike information criterion (AIC), Bayesian information criterion (BIC) and determination coefficient R 2 Evaluate model performance. The main goal of stepwise regression is to find a model that can explain the data while avoiding overfitting. Stepwise regression includes three main methods: forward selection, backward elimination, and bidirectional stepwise regression.

[0077] In this embodiment, a stepwise linear regression model is established using a bidirectional stepwise regression method. The optimal stepwise linear regression model obtained by fitting is shown in Table 7. The model prediction result diagram is shown in Figure 4 .

[0078] Table 7: Stepwise linear regression model based on male samples

[0079]

[0080]

[0081] Note: Among them, n is the total number of methylation sites included in the regression model, i is the number of corresponding methylation sites, corr is the correlation coefficient of each methylation site, and beta is the methylation level of each methylation site.

[0082] (6) Establish a stepwise linear regression model with fewer sites; randomly select 5-8 sites from the 29 sites for remodeling, and evaluate the index determination coefficient (R 2 , R-Square) and mean absolute error (MAE), from which a few sites with the most significant correlation with age were screened out, and a stepwise linear regression model of a few sites was re-established based on the few sites, as shown in Table 8 for details, and the model prediction results are shown in Figure 5 .

[0083] Table 8: Stepwise linear regression model based on fewer loci for male samples

[0084]

[0085]

[0086] Note: Among them, n is the total number of methylation sites included in the regression model, i is the number of corresponding methylation sites, corr is the correlation coefficient of each methylation site, and beta is the methylation level of each methylation site.

[0087] (7) Use the R package gbm to build a gradient boosting regression model. Gradient Boosting Regression (GBR) is an ensemble learning method that gradually builds multiple weak learners and weights the prediction results of these weak learners to improve the prediction performance of the overall model. GBR combines the ideas of gradient descent and boosting, using the gradient descent method to gradually reduce the loss function, thereby improving the prediction accuracy of the model.

[0088] In this embodiment, the number of weak learners set in the model is 5000, the learning rate is 0.01, the maximum depth of each weak learner (decision tree) is 1, and the sample ratio used in each training of the weak learner is 1. The optimal gradient boosting regression model obtained by fitting is shown in Table 9, and the model prediction results are shown in Figure 6 .

[0089] Table 9: Gradient boosting regression model based on male samples

[0090]

[0091]

[0092] Note: The second column in the table is the relative influence value of each site in the model, normalized to a total of 100.

[0093] (8) Establishing a gradient boosting regression model with fewer sites: According to the relative influence ranking of the sites, the first 8 sites were used to re-model the model, and the gradient boosting regression model with fewer sites was re-established based on the fewer sites. See Table 10 for details, and the model prediction results are shown in Figure 7 .

[0094] Table 10: Gradient boosting regression model with fewer sites based on male samples

[0095]

[0096]

[0097] Note: The second column in the table is the relative influence value of each site in the model, normalized to a total of 100.

[0098] S5. Model comparison and evaluation:

[0099] (1) Comparison of the epigenetic clock models established above is shown in Table 11. Among them, the models established by the gradient boosting regression method are all the optimal prediction models.

[0100] Table 11: Comparison of epigenetic clock models

[0101]

[0102] Note: MAE is the mean absolute error between predicted age and chronological age, R 2 is the fit measure of the linear regression model, where R 2 The larger it is, the better the model fit is.

[0103] (2) Randomly extract some sample data from the DNA methylation chip sequencing data downloaded from the GEO database, and compare the optimal prediction model constructed in this embodiment with other classic models. The model comparison results are shown in Table 12.

[0104] Table 12: Comparison of various prediction models based on GEO male sample data

[0105]

[0106]

[0107] The results show that each model has a good R 2 value (>0.85), although the prediction model involved in the CN115240761A patent only requires 18 sites, its MAE is 9.6786. Among the models tested, the difference between the average predicted age and the actual age calculated by this model is the largest; the traditional Horvath S2018 model and Levine M2018 model require 391 sites and 513 sites, respectively, to achieve a good fitting effect. The gradient boosting regression model and the gradient boosting regression model with fewer sites constructed in this embodiment can achieve a better fitting effect based on a smaller number of sites, and have a relatively ideal MAE, which is suitable for relatively low-cost promotion and application in the population.

[0108] (3) The original data of peripheral blood DNA methylation (850k chip) from 100 male subjects who had undergone physical examinations in the hospital health management center were selected, and the gradient boosting regression model constructed in this embodiment was compared with other classic models. The model results are shown in Table 13.

[0109] Table 13: Comparison results of various methylation prediction models from external male sample data

[0110]

[0111] The results show that since only 15 of the 18 sites required by the model in the CN115240761A patent exist in the data, the effect of the model cannot be evaluated; compared with the traditional Horvath S2018 model and Levine M2018 model, the gradient boosting regression model constructed in this embodiment and the gradient boosting regression model with fewer sites can achieve a smaller MAE based on a smaller number of sites.

[0112] S6. Age assessment and aging prediction:

[0113] (1) Based on the predicted age value, the aging speed of an individual is defined as the difference between the predicted age and the actual age. If the difference is positive, the aging speed is fast, and the larger the value, the worse the physical health; if the difference is negative, the aging speed is slow, and the smaller the value, the better the physical health.

[0114] (2) DNA methylation data of pooled samples of Chinese longevity towns were selected, which were equal amounts of DNA from 66 middle-aged men, with an average chronological age of 43.64 years. The average biological age of the samples was predicted by the average effect value of the methylation sites in the samples, and the prediction effects of the optimal model of the present invention and other classical models were compared. The model comparison results are shown in Table 14. Compared with the Horvath S2018 model and the Levine M2018 model, this embodiment has an advantage in the number of sites and is more feasible for promotion.

[0115] Table 14 Comparison results of various prediction models based on male sample data from China's longevity villages

[0116] Model Number of sites Predicted age Prediction Difference Gradient Boosting Regression Model 29 49.20 5.56 Gradient boosting regression model with fewer sites 8 47.73 4.09 Prediction model in CN115240761A patent 18 37.05 -6.59 HorvathS2018 model 391 45.00 1.36 LevineM2018 model 513 40.78 -2.86

[0117] The results show that compared with the CN115240761A patent model, the gradient boosting regression model constructed in this embodiment and the gradient boosting regression model with fewer sites have smaller prediction differences and better model fitting effects. Compared with the Horvath S2018 model and the Levine M2018 model, this embodiment has an advantage in the number of sites and is more feasible for promotion.

[0118] Through the above specific implementations, those skilled in the art can easily implement the present invention. However, it should be understood that the present invention is not limited to the above specific implementations. On the basis of the disclosed implementations, those skilled in the art can arbitrarily combine different technical features to achieve different technical solutions.

Claims

1. A method for assessing male biological age and predicting aging based on DNA methylation data of peripheral blood, characterized in that: The following steps are involved: S1. Screening of DNA methylation sites: DNA methylation sites related to male biological age and aging were screened by searching relevant literature and research results on PubMed and Illumina methylation array platforms; S2. Acquisition of modeling database: Download DNA methylation chip sequencing data from the GEO website, filter out male data, and obtain the modeling database; S3. Determination of modeling samples and modeling sites: by calculating the sample detection rate and site detection rate, the modeling database and DNA methylation sites are screened to obtain modeling samples and modeling sites; S4. Construction of epigenetic clock model: Use modeling samples and modeling sites to build an epigenetic clock model; S5. Model comparison and evaluation: Compare the epigenetic clock model with the traditional model and perform model evaluation. S6. Age assessment and aging prediction: Select external samples to conduct biological age assessment and aging prediction.

2. The method for assessing male biological age and predicting aging based on DNA methylation data of peripheral blood according to claim 1, characterized in that: The DNA methylation sites associated with male biological age and aging described in S1 are cg01511567, cg01820374, cg02154074, cg04084157, cg04528819, cg04836038, cg05294243, cg05442902, cg06493994, cg07211259, cg09809672, cg13460409, cg13854874, cg 13975369, cg14175438, cg16386080, cg17324128, cg19008809, cg19283806, cg19722847, cg19761273, c g22736354, cg22947000, cg25809905, cg26394940, cg26614073, cg26824091, cg26842024, cg27015931.

3. The method for assessing male biological age and predicting aging based on DNA methylation data of peripheral blood according to claim 1, characterized in that: The DNA methylation chip sequencing data described in S2 includes country, biological age, gender, and peripheral blood DNA methylation chip sequencing data, wherein the peripheral blood DNA methylation chip sequencing data includes Illumina Human Methylation450K chip and Illumina Human Methylation 850K chip.

4. The method for assessing male biological age and predicting aging based on DNA methylation data of peripheral blood according to claim 1, characterized in that: The sample detection rate and site detection rate described in S3 were calculated using a python script.

5. The method for assessing male biological age and predicting aging based on DNA methylation data of peripheral blood according to claim 1, characterized in that: S4 also includes imputation of undetected data for the modeling samples using the makeX function included in the R package glmnet.

6. The method for assessing male biological age and predicting aging based on DNA methylation data of peripheral blood according to claim 1, characterized in that: The epigenetic clock models described in S4 are the penalized multivariate regression model of elastic networks, the stepwise linear regression model, and the gradient boosting regression model.

7. The method for assessing male biological age and predicting aging based on DNA methylation data of peripheral blood according to claim 6, characterized in that: S4 also includes randomly selecting a smaller number of sites from the modeling sites for re-modeling, screening the sites that are most significantly correlated with age based on the determination coefficient and mean absolute error of the regression model evaluation indicators, and re-establishing the epigenetic clock model to obtain an epigenetic clock model with fewer sites, which is used to reduce the prediction cost while ensuring the prediction accuracy.

8. The method for assessing male biological age and predicting aging based on DNA methylation data of peripheral blood according to claim 1, characterized in that: The model evaluation methods described in S5 include, but are not limited to, the mean absolute error between the predicted age and the biological age, the fit metric of the linear regression model, and the number of model sites.

9. The method for assessing male biological age and predicting aging based on DNA methylation data of peripheral blood according to claim 1, characterized in that: The traditional models described in S5 are the Horvath S2018 model and the Levine M2018 model.

10. The method for assessing male biological age and predicting aging based on peripheral blood DNA methylation data according to claim 1, characterized in that: S6 also includes defining an individual's aging rate as the difference between predicted age and biological age.

Images