Human body biological age detection method based on DNA methylation

Abstract

The invention provides a human body biological age detection method based on DNA methylation, which comprises the following steps: step S1, aiming at at least one subject, collecting a peripheral blood sample of the subject, and extracting DNA in blood from the sample; the method comprises the following steps: S1, extracting DNA, S2, measuring a methylation map of the extracted DNA to obtain detailed information of the methylation state of the DNA, and S3, calculating a corresponding methylation rate according to the strength of a methylation signal and a non-methylation signal in the obtained methylation map; and step S4, inputting the calculated DNA methylation rate into a biological age model obtained by pre-training to obtain the biological age corresponding to the subject. The method has the beneficial effects that the biological age can be accurately predicted by acquiring methylation information of DNA data, a novel biomarker-based assessment means is provided, and the method has important application value in health management, disease risk assessment and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of biological age detection, and in particular to a method for detecting human biological age based on DNA methylation. Background Art

[0002] Aging is a universal and inevitable biological process, accompanied by the gradual decline of physiological functions, increased risk of disease, and ultimately death. Although all organisms will age, there are significant differences in the aging rate and health status of different individuals within the same species. Calendar age is the main indicator for measuring aging, but it is difficult to capture the differences in individual aging rates. The health status, functional level and disease risk of individuals of the same age may be very different, which shows that relying solely on calendar age cannot effectively predict individual health risks. In contrast, biological age reflects the degree of individual functional decline and is closely related to age-related disease risk. It is a key indicator for quantifying the individual aging process. The difference in biological age can reveal the difference in individual aging rates, making it have great potential in capturing aging trajectories and predicting disease risks. However, how to accurately measure biological age remains a major challenge in current research.

[0003] At present, a variety of biomarkers and measurement strategies have been proposed, such as metabolic activity, inflammation level, body function and telomere length. These markers reveal the complexity of the aging process from different perspectives. However, these methods face challenges in measurement accuracy, repeatability and interpretation of their biological significance in application. Therefore, developing a tool that can accurately predict biological age and reflect the relationship between aging rate and health risk has become an urgent problem to be solved. Model organism studies have shown that the accumulation of cell damage is a core factor in aging, and epigenetic disorders play a key role in this process. DNA methylation mainly occurs at cytosine-guanine dinucleotide (CpG) sites, which is crucial in maintaining cell identity and gene expression stability. There are significant differences in methylation patterns between different tissues and their cell types, and methylation at CpG sites changes significantly with age. A large number of population studies have shown that DNA methylation is closely related to diseases such as Alzheimer's disease, cardiovascular disease and cancer. This dynamic regulation ability of DNA methylation makes it an important biomarker for studying aging and disease.

[0004] As one of the main ways to obtain DNA methylation information, blood biopsy has significant feasibility and importance. First of all, blood is one of the most widely contacted tissues in the body, which can reflect a variety of physiological and pathological states, especially changes related to the immune system. Through the detection of DNA methylation in the blood, individual information can be collected non-invasively, which is convenient and easy to repeat. This is particularly important in large-scale population studies and longitudinal studies. In addition, the cell types in the blood are diverse, especially the DNA methylation pattern of immune cells is closely related to the aging process, which makes blood an ideal material for studying aging and disease risks. Through these technologies, specific methylation sites related to the aging rate can be captured, and an "epigenetic clock" reflecting biological age can be constructed. These clock models can not only reflect the biological age of an individual, but also predict future health status and disease risks, especially in chronic diseases related to aging, such as atherosclerosis, cardiovascular disease and Alzheimer's disease, which have extremely high predictive value.

[0005] Therefore, using DNA methylation from blood biopsy to determine biological age is not only technically feasible, but also has significant application prospects in health management, disease prediction and aging research. This method provides a solid foundation for accurately measuring biological age, revealing differences in aging rates and predicting health risks. Summary of the invention

[0006] The technical problem to be solved by the present invention is to achieve accurate prediction of biological age by obtaining methylation information of DNA data. In order to overcome the defects of the above-mentioned prior art (or related technology), the present invention provides a method for detecting human biological age based on DNA methylation.

[0007] The present invention provides a method for detecting human biological age based on DNA methylation, comprising the following steps: Step S1, for at least one subject, obtaining a peripheral blood sample of the subject and extracting DNA in the blood from the peripheral blood sample; Step S2, measuring the DNA methylation profile of the DNA extracted from the blood; Step S3, calculating the methylation rate according to the intensity of the methylation signal and the intensity of the non-methylation signal obtained by the measurement; Step S4, inputting the methylation rate into a pre-trained biological age prediction model to obtain the biological age corresponding to the subject.

[0008] Compared with the prior art, the human biological age detection method based on DNA methylation in this application has the following advantages: In the present application, DNA data is obtained through step S1, methylation profile is measured through step S2, methylation rate is calculated through step S3, and biological age is predicted through step S4. The biological age prediction model is applied in the whole process, and the methylation rate is analyzed through the biological age prediction model to obtain the biological age, so as to achieve accurate prediction of biological age by obtaining methylation information of DNA data.

[0009] In a possible implementation, in step S2, the DNA methylation profile is determined using an Illumina methylation 450K chip or an EPIC chip.

[0010] In a possible implementation, in step S3, the methylation rate is obtained by the following calculation formula: Methylation rate = (signal intensity of methylated signal) / (signal intensity of unmethylated signal + signal intensity of methylated signal + 100) In a possible implementation manner, before executing step S1, a biological age prediction model training process is also included, and the biological age prediction model process includes: Step A1, obtaining DNA methylation map data corresponding to multiple historical peripheral blood samples, and calculating the methylation rate; Step A2, for each of the DNA methylation map data, subtract the mean of the methylation rates at each of the methylation sites from the methylation rate of each of the DNA methylation map data, and then divide the result by the standard deviation of the methylation rate of each of the methylation sites to obtain standardized methylation data; Step A3, based on the Lasso regression algorithm, an L1 regularization term is introduced to screen the methylation sites with non-zero coefficients in the methylation data as characteristic methylation sites related to age, and a methylation rate matrix corresponding to the characteristic methylation sites is extracted; Step A4, using the methylation rate of the characteristic methylation site as a characteristic variable, and the corresponding actual age of the historical peripheral blood sample as a target variable to obtain a model training set; Step A5, constructing the biological age prediction model from the extracted characteristic methylation sites related to age using an elastic network regression model, introducing L1 regularization and L2 regularization, and configuring two regularization parameters alpha and lambda to determine the mixing ratio of L1 regularization and L2 regularization; Step A6, determining the optimal alpha and lambda by cross-validation method; Step A7, dividing the model training set into multiple model training subsets, selecting one of the model training subsets as a validation set each time, and the remaining model training subsets as training sets to train the biological age prediction model.

[0011] In a possible implementation, in step A6, a 10-fold cross validation method is used to determine the optimal regularization parameter, including: Step B1, integrating the signal intensities of the characteristic methylation sites corresponding to the DNA methylation map data and the DNA methylation map data into a data set, dividing the data set into 10 equal data subsets, using one of the data subsets as a validation set each time, and using the remaining nine data subsets as training sets, and performing model training and validation on the biological age prediction model respectively; Step B2, obtaining the mean square error of the model after each verification, selecting the regularization parameter with the smallest mean square error of the model as the optimal value, and retraining the biological age prediction model based on the training set based on the optimal value.

[0012] In a possible implementation, in step A5, the value of alpha varies between 0 and 1. When alpha=1, the biological age prediction model is equivalent to Lasso regression; when alpha=0, the biological age prediction model is equivalent to Ridge regression; a value between 0 and 1 represents a mixture of Lasso regression and Ridge regression; the value of lambda is used to control the strength of regularization.

[0013] In a possible implementation, in step A6, the mean square error of the biological age prediction model is calculated for each pair of alpha and lambda, and the alpha and lambda with the smallest mean square error are selected as the optimal values. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a flow chart of the steps of the present invention; Figure 2 is a flow chart of the steps of the biological age prediction model process of the present invention; Figure 3 It is a flow chart of the steps of the Lasso regularization parameter determination process of the present invention; Figure 4 Schematic diagram of model performance in an embodiment of the present invention. DETAILED DESCRIPTION

[0015] First, those skilled in the art should understand that these implementations are only used to explain the technical principles of the embodiments of the present application, and are not intended to limit the protection scope of the embodiments of the present application. Those skilled in the art can make adjustments to them as needed to adapt to specific application scenarios.

[0016] The present application is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] See also Figure 1 The present application embodiment discloses a method for detecting human biological age based on DNA methylation, comprising: Step S1, for at least one subject, obtaining a peripheral blood sample of the subject and extracting DNA in the blood from the peripheral blood sample; Step S2, measuring the DNA methylation profile of the DNA extracted from the blood; Step S3, calculating the methylation rate according to the intensity of the methylation signal and the intensity of the non-methylation signal obtained by the measurement; Step S4, inputting the methylation rate into a pre-trained biological age prediction model to obtain the corresponding biological age of the subject.

[0018] Continue to see Figure 1 In step S1, DNA extraction is performed. First, DNA is extracted from the subject's peripheral blood sample. Peripheral blood specifically refers to a blood sample collected through a vein. After the blood sample is collected, it should be stored under appropriate conditions and DNA extraction should be performed as soon as possible to ensure the quality of the DNA.

[0019] Continue to see Figure 1 The extracted DNA data was then used to measure the DNA methylation map using the Illumina methylation 450K model chip. The chip can efficiently analyze the DNA methylation status within the genome and generate a detailed methylation map. The measurement results include the signal intensity of the methylation signal and the signal intensity of the non-methylation signal.

[0020] Continue to see Figure 1 In step S3, the methylation rate is obtained by the following calculation formula: Methylation rate = (signal intensity of methylated signal) / (signal intensity of unmethylated signal + signal intensity of methylated signal + 100) See also Figure 2 Before executing step S1, a biological age prediction model training process is also included, and the biological age prediction model process includes: Step A1, obtaining DNA methylation map data corresponding to multiple historical peripheral blood samples, and calculating the methylation rate; Step A2, for each DNA methylation map data, subtract the mean of the methylation rates at each methylation site from the methylation rate of each DNA methylation map data, and then divide it by the standard deviation of the methylation rate of each methylation site to obtain standardized methylation data; Step A3, based on the Lasso regression algorithm, an L1 regularization term is introduced to screen the methylation sites with non-zero coefficients in the methylation data as the characteristic methylation sites related to age, and the methylation rate matrix corresponding to the characteristic methylation sites is extracted; Step A4, using the methylation rate of the characteristic methylation site as a characteristic variable, and the actual age of the corresponding historical peripheral blood sample as a target variable to obtain a model training set; Step A5, using the elastic network regression model to construct a biological age prediction model from the extracted age-related characteristic methylation sites, and introducing L1 regularization and L2 regularization, configuring two regularization parameters alpha and lambda to determine the mixing ratio of L1 regularization and L2 regularization; Step A6, determining the optimal alpha and lambda by cross-validation method; Step A7, divide the model training set into multiple model training subsets, select one of the model training subsets as the validation set each time, and use the remaining model training subsets as the training sets to train the biological age prediction model.

[0021] Continue to see Figure 2 The present invention uses the lasso regression algorithm to extract characteristic methylation sites related to age. The lasso regression algorithm can effectively screen out characteristic methylation sites that have a significant impact on predicting biological age by introducing L1 regularization. First, the data is standardized. Specifically, the signal intensity of each methylation site is subtracted from its mean and divided by its standard deviation to obtain standardized methylation data. Subsequently, the lasso regression algorithm is used for feature selection, using the actual age of the subject as the target variable and all standardized DNA methylation sites as feature variables. The lasso regression algorithm can automatically screen out characteristic methylation sites that are significantly related to age by introducing L1 regularization.

[0022] See also Figure 3 To determine the optimal value of the regularization parameter, a 10-fold cross validation method is used. The specific steps include: a. Randomly divide the data set into 10 equal parts, use one of them as the validation set each time, and the remaining 9 as the training set, for model training and validation respectively; b. Through multiple verifications, calculate the mean square error (MSE) of the model under different regularization parameters, and select the regularization parameter that minimizes the MSE as the optimal regularization parameter. The calculation method of the mean square error is as follows: , in represents the true value, represents the estimated value, and n represents the sample size; c. After determining the optimal regularization parameter, the biological age prediction model was retrained using the entire data set. After the training was completed, methylation sites with non-zero coefficients were screened out based on the regression coefficient vector of the model. These sites were the characteristic methylation sites related to age, indicating that they made an important contribution to predicting biological age. d. In order to verify the effectiveness of the extracted characteristic methylation sites, an independent DNA methylation dataset was used to test the biological age prediction model. The root mean square error (RMSE) and correlation coefficient between the biological age predicted by the model and the actual age were calculated to evaluate the prediction performance of the model.

[0023] Continue to see Figure 1 Based on the extracted characteristic methylation sites, the present invention further uses machine learning and deep learning algorithms including the elastic network algorithm to construct a biological age prediction model. The elastic network algorithm combines the advantages of L1 and L2 regularization, can handle multicollinearity problems, and improve the robustness of the model. By combining other machine learning and deep learning algorithms, the biological age prediction model finally established has a higher prediction accuracy. The specific scheme is as follows: According to the age-related characteristic methylation sites extracted by the Lasso regression algorithm mentioned above, a feature matrix containing these characteristic methylation sites is constructed. The characteristic variable of each sample is the methylation rate corresponding to these characteristic methylation sites, and the target variable is the actual age of the subject; The elastic network regression model is used to construct a biological age prediction model from the extracted age-related characteristic methylation sites. The model can handle the multicollinearity problem in high-dimensional data and improve the robustness of the prediction model. First, the regularization parameter of the elastic network model is selected. The elastic network regression model combines the advantages of L1 (Lasso) and L2 (Ridge) regularization. In order to optimize the model performance, two key regularization parameters need to be selected: alpha: used to determine the mixing ratio of L1 and L2 regularization. The alpha value varies between 0 and 1. When alpha=1, the model is equivalent to Lasso regression; when alpha=0, the model is equivalent to Ridge regression; values ​​between 0 and 1 represent a mixture of the two. lambda: controls the strength of regularization. A larger lambda value will increase the regularization effect, thereby reducing the complexity of the model and preventing overfitting. The alpha value used in the present invention is 0.5, and the lambda value is 0.5; Then the optimal alpha and lambda values ​​are determined through cross-validation. In order to ensure the best performance of the model, the cross-validation method is used to select the optimal alpha and lambda values; First, divide the data set: randomly divide the model training set into multiple model training subsets (such as 10 subsets), select one of the model training subsets as the validation set each time, and the other model training subsets as the training set. Then, for each pair of alpha and lambda combinations, train the model on the training set and evaluate the model performance on the validation set. Finally, calculate the mean square error (MSE) under each combination and select the parameter combination with the smallest MSE as the optimal alpha and lambda. Subsequently, the elastic network regression model was trained using the optimal alpha and lambda. After determining the optimal alpha and lambda values, the entire training data set was used for model training. First, the feature matrix and target variable were input, and the methylation rate of the characteristic methylation sites extracted previously was used as the input feature matrix. The actual age of the subject was used as the target variable, and then the model was fitted. The elastic network algorithm was used to train the model on the entire model training set so that it could capture the relationship between the input feature matrix and the target variable. The final elastic network regression model, namely the biological age prediction model, was obtained through model fitting. This model can predict the biological age based on the input methylation rate.

[0024] Continue to see Figure 1The biological age prediction model constructed in the present invention needs to be verified to ensure its accuracy. The present invention obtains additional public database information and applies statistical analysis methods including root mean square error (RMSE) and correlation coefficient to compare the biological age predicted by the model with the actual age, evaluate the performance of the model, and optimize the parameter configuration of the model through multiple verifications to ensure the reliability and consistency of the prediction results.

[0025] The present invention will be further described below in conjunction with a specific embodiment: Obtaining DNA methylation microarray data from public databases Select the GEO database and download the human data set numbered GSE223748. This data set contains rich clinical information, such as gender, age, tissue source, and health status. In order to ensure the pertinence and effectiveness of the study, healthy human samples with peripheral blood as the tissue source were specially selected, and their corresponding DNA methylation maps were downloaded. After screening, 173 samples were finally collected, including 116 male samples and 57 female samples. The age range of the participants was from the youngest 26 to the oldest 92 years old, with an average age of 66 years old. The diversity of these samples not only covers different genders, but also provides a wide range of data sets across multiple age groups, which enables in-depth exploration of the relationship between DNA methylation and age, as well as its performance differences in different genders; Organizing DNA methylation chip map data The data downloaded from the public database is in .idat format. Although this format contains important methylation information, it cannot be directly used for subsequent analysis. Therefore, in this embodiment, the 'ChAMP' package in R is used to organize and calculate these data to obtain a usable methylation rate matrix. Specifically, the 'ChAMP' package provides a series of powerful functions that can efficiently process and parse .idat format data and ultimately generate the required methylation rate matrix. In this process, the methylation rate is calculated as follows: Methylation rate = (signal intensity of methylated signal) / (signal intensity of unmethylated signal + signal intensity of methylated signal + 100); The design of this calculation method takes into account the dynamic range of the signal. By adding a constant (such as 100) in the denominator, the influence of the minimum signal intensity on the methylation rate is effectively reduced, ensuring the stability and reliability of the calculation results. Ultimately, the methylation rate matrix obtained in this embodiment will provide a solid data foundation for subsequent analysis and help in-depth research on the relationship between DNA methylation and age and health status. Extract characteristic methylation sites First, the obtained methylation rate matrix was z-score normalized to obtain normalized data. This process ensures that all features have the same scale by subtracting the mean of the feature from each data point and dividing by the standard deviation, which is convenient for subsequent analysis. Subsequently, the Lasso algorithm was used for feature screening, which was implemented using the glmnet package in R. The main steps include: (1) Parameter settings: a. Set alpha=1 to specify Lasso regression; b. Use cross validation (via cv.glmnet function) to select an appropriate penalty parameter λ, and use 10-fold cross validation to improve the stability of the model; (2) Model fitting: Fit the Lasso model on the training set and calculate the coefficient of each feature. This process will generate a series of models corresponding to λ values. The best model is found by selecting the λ value that minimizes the mean square error (MSE). The λ finally selected in this embodiment is 0.439. (3) Feature selection: By observing the coefficient of each feature, the features with non-zero coefficients are selected as the key features to be finally screened out. The Lasso algorithm helps identify the features that are most closely related to the target variable by reducing the coefficients of unimportant features to zero. Ultimately, the features screened by the Lasso algorithm will provide key input for subsequent model construction and biological analysis, ensuring the accuracy and effectiveness of the research; Construction of biological age prediction model Next, the 173 samples were divided into a training set and a prediction set in a ratio of 4:1. The training set contained 141 samples and the prediction set contained 32 samples. This division method ensures that the model has enough data during training to improve its generalization ability and prediction accuracy. Specifically, the number of samples in the training set is sufficient to capture the variability between samples, so that the model can learn the relationship between effective features and age.

[0026] After completing the feature screening, the selected feature sites are constructed into a methylation matrix, and these features are combined with the age labels corresponding to the samples for subsequent model construction. This embodiment adopts the elastic network model (ElasticNet), which can take into account the advantages of both Lasso and ridge regression, effectively deal with multicollinearity problems, and improve the model's explanatory power and prediction accuracy.

[0027] When constructing the elastic network model, the parameter α is first set to 0.5 to balance the effects of Lasso and ridge regression. The use of the elastic network helps to retain features with strong correlation during feature selection and avoid information loss caused by over-penalizing certain features. The model is trained by inputting the methylation feature matrix of the training set and the corresponding age labels into the model to achieve fitting of the target variable.

[0028] To ensure the stability and reliability of the model, this embodiment adopts a cross-validation method. Specifically, the training set is further divided into 10 folds, and model training and validation are performed on each fold to obtain the model performance index for each λ value. By comparing the mean square error (MSE) under different λ values, the λ value that minimizes the MSE is selected as the parameter setting of the final model. At this time, the λ value is 1.028. This process ensures the consistency of the model in different data partitions and reduces the model's dependence on a specific training set.

[0029] Finally, the constructed elastic network model obtained a total of 62 loci related to age weight, with an intercept of 67.33. The coefficients of each locus are as follows: Location Correlation coefficient cg00463693 -0.02735148 cg00729699 -1.733843076 cg00925802 -15.40753415 cg01343642 -0.275705674 cg01447660 -15.36541358 cg01710150 -0.945663356 cg02228185 -4.122315254 cg02981187 -7.764872692 cg03607117 3.725213887 cg04084157 21.90814829 cg04836038 8.571952563 cg05369916 -5.588514419 cg05442902 -24.12674612 cg05542681 2.802320506 cg05706534 1.138232672 cg06639320 11.30066843 cg06646690 -2.334177542 cg06683527 -0.060747715 cg07553761 1.061882165 cg08097417 19.65707445 cg08743477 4.309780009 cg09175318 -0.334125123 cg09809672 -0.825867434 cg10211416 -1.751681854 cg10501210 -0.153678804 cg10820947 -1.778754008 cg10939684 -8.244916188 cg11071401 16.53335408 cg11793825 10.80588759 cg11992861 -0.17433611 cg12941369 -0.438367279 cg13741927 -6.167385726 cg14361627 3.098059237 cg14568069 -0.262048183 cg15013019 -4.750856894 cg16756969 -1.001419777 cg16867657 30.00089843 cg16871561 -1.059493026 cg17148169 -0.267130927 cg17243289 16.91769897 cg18468088 9.529819526 cg18470619 -3.738770677 cg18473521.1 5.687992919 cg18582260 -5.616210737 cg18992104 -0.3468091 cg19283806 -19.81399412 cg19663246 -2.325928065 cg19796371 -0.935073021 cg19917435 -2.893896005 cg22454769 5.260551801 cg22468573 -1.639865905 cg22512531 1.471264068 cg22943590 -3.718850151 cg23124451 -3.355111235 cg23500537 3.523726603 cg23798387 -3.711716182 cg24183320 0.508105947 cg24639882 -0.762436864 cg25427880 1.859044482 cg25478614 10.82009728 cg26842024 18.26666486 cg27537626 -0.303093174 It will be used to evaluate the prediction set. By comparing the predicted values ​​with the actual age labels, the prediction accuracy and robustness of the model are calculated. The specific model performance is as follows Figure 4 shown.

[0030] This model not only reveals the potential relationship between methylation status and age, but also provides important data support and theoretical basis for subsequent biological research and clinical applications. Through this method, it can provide more accurate tools for individual biological age assessment and promote research on health monitoring and disease prevention.

[0031] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "in the present embodiment", "specific example", or "some examples" etc. means that the specific features, mechanisms, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, mechanisms, materials or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0032] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for detecting human biological age based on DNA methylation, characterized in that: The following steps are involved: Step S1, for at least one subject, obtaining a peripheral blood sample of the subject and extracting DNA in the blood from the peripheral blood sample; Step S2, measuring the DNA methylation profile of the DNA extracted from the blood; Step S3, calculating the methylation rate according to the intensity of the methylation signal and the intensity of the non-methylation signal obtained by the measurement; Step S4, inputting the methylation rate into a pre-trained biological age prediction model to obtain the biological age corresponding to the subject.

2. The method for detecting human biological age based on DNA methylation according to claim 1, characterized in that: In the step S2, the DNA methylation profile is measured using an Illumina methylation 450K chip or an EPIC chip.

3. The method for detecting human biological age based on DNA methylation according to claim 1, characterized in that: In step S3, the methylation rate is obtained by the following calculation formula: Methylation rate = (signal intensity of methylated signal) / (signal intensity of unmethylated signal + signal intensity of methylated signal + 100).

4. The method for detecting human biological age based on DNA methylation according to claim 1, characterized in that: Before executing step S1, a biological age prediction model training process is also included, and the biological age prediction model process includes: Step A1, obtaining DNA methylation map data corresponding to multiple historical peripheral blood samples, and calculating the methylation rate; Step A2, for each of the DNA methylation map data, subtract the mean of the methylation rates at each of the methylation sites from the methylation rate of each of the DNA methylation map data, and then divide the result by the standard deviation of the methylation rate of each of the methylation sites to obtain standardized methylation data; Step A3, based on the Lasso regression algorithm, an L1 regularization term is introduced to screen the methylation sites with non-zero coefficients in the methylation data as characteristic methylation sites related to age, and a methylation rate matrix corresponding to the characteristic methylation sites is extracted; Step A4, using the methylation rate of the characteristic methylation site as a characteristic variable, and the corresponding actual age of the historical peripheral blood sample as a target variable to obtain a model training set; Step A5, constructing the biological age prediction model from the extracted characteristic methylation sites related to age using an elastic network regression model, introducing L1 regularization and L2 regularization, and configuring two regularization parameters alpha and lambda to determine the mixing ratio of L1 regularization and L2 regularization; Step A6, determining the optimal alpha and lambda by cross-validation method; Step A7, dividing the model training set into multiple model training subsets, selecting one of the model training subsets as a validation set each time, and the remaining model training subsets as training sets to train the biological age prediction model.

5. The method for detecting human biological age based on DNA methylation according to claim 4, characterized in that: In step A6, a 10-fold cross validation method is used to determine the optimal regularization parameter, including: Step B1, integrating the signal intensities of the characteristic methylation sites corresponding to the DNA methylation map data and the DNA methylation map data into a data set, dividing the data set into 10 equal data subsets, using one of the data subsets as a validation set each time, and using the remaining nine data subsets as training sets, and performing model training and validation on the biological age prediction model respectively; Step B2, obtaining the mean square error of the model after each verification, selecting the regularization parameter with the smallest mean square error of the model as the optimal value, and retraining the biological age prediction model based on the training set based on the optimal value.

6. The method for detecting human biological age based on DNA methylation according to claim 4, characterized in that: In step A5, the value of alpha varies between 0 and 1. When alpha=1, the biological age prediction model is equivalent to Lasso regression; when alpha=0, the biological age prediction model is equivalent to Ridge regression; values ​​between 0 and 1 represent a mixture of Lasso regression and Ridge regression; the value of lambda is used to control the strength of regularization.

7. The method for detecting human biological age based on DNA methylation according to claim 4, characterized in that: In step A6, the mean square error of the biological age prediction model is calculated for each pair of alpha and lambda, and the alpha and lambda with the smallest mean square error are selected as the optimal values.

Images