A methylation age assessment method based on DNA methylation level data

Abstract

The present invention discloses a methylation age assessment method based on DNA methylation level data. The present invention constructs a methylation age prediction model requiring that the input file contains important methylation site information, and the important methylation site is a methylation site in a gene region reported in the literature and / or related to age changes. In the embodiment of the present invention, the R-square, MAE, MSE, and RMSE are calculated between the obtained validation set predicted methylation age and the known time age of the validation set sample. The result shows that the R-square is 0.855, indicating that the methylation age of the sample to be tested is predicted by using the prediction model constructed by using important sites, and the information is retained more completely, and the obtained methylation age prediction value is more accurate, and the degree of conformity with the real time age is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of bioinformatics analysis, and in particular to a methylation age assessment method based on DNA methylation level data. Background Art

[0002] DNA methylation refers to the chemical modification of a specific base on a DNA sequence with a methyl group under the catalysis of a methyltransferase. DNA methylation can have various effects on the growth and development of organisms, such as regulating gene expression and inhibiting transposable elements. Taking mammalian cells as an example, about 3 / 4 of the CpG (cytosine-phosphate-guanine) sites in cells are methylated. However, as individuals age, the methylation level in individuals is also in a dynamic process.

[0003] In the limited life process of human beings, chronological age is often used to describe the growth and aging status of individuals. This description through time is simple and applicable to all individuals. However, due to different individual genetic information, differences in external environmental factors in the environment of growth and development, and other reasons, there are great differences in the internal physiological conditions of individuals at the same chronological age, and this difference becomes more obvious as the chronological age increases. Therefore, chronological age can only describe the growth and development experience of an individual in terms of time, and the actual physiological condition of the individual still needs a more objective and accurate standard to measure.

[0004] In 2013, Horvath et al. established an epigenetic clock based on the DNA methylation levels in different individual tissues to assess individual age. However, due to genetic differences between individuals and differences in different tissues (such as skin, blood, etc.), the assessment results may be affected. Ribosomal DNA (rDNA) is a highly evolutionarily conserved genomic nucleic acid fragment. In 2019, Wang et al. used the ribosomal clock to assess the methylation age of humans, mice, dogs, etc. in different species, and believed that the ribosomal clock is a universal cross-species methylation age assessment tool. The DNA methylation levels in different tissues will affect the prediction results due to large differences between individuals or large differences between tissues and parts. At the same time, the physiological level of an individual is affected by the interaction of all genes in the body. If the methylation age is only predicted by the sites in the region where the conservative sequence is located, there is a possibility of missing important gene site information.

[0005] Establishing a methylation clock based on important sites and predicting the methylation age within an individual has become an important issue that needs to be solved urgently. Summary of the invention

[0006] The technical problem to be solved by the present invention is how to predict or evaluate the methylation age of an individual through important sites.

[0007] In order to solve the above technical problems, the present invention first provides a method for predicting methylation age. The method comprises the following steps: extracting methylation site information from a sample methylation level file to obtain a methylation site information file; splitting the methylation site information file into a known age data set and an unknown age data set according to age information; constructing a model for predicting methylation age using an elastic network function for the known age data set; and using the model for predicting methylation age to predict the methylation age of samples in the unknown age data set.

[0008] In the above method, the samples can be divided into known age dataset samples and unknown age dataset samples. The known age dataset samples mentioned above can be modeling samples. The sample unknown age dataset samples can be samples to be tested. The samples can be biological tissue and / or cell samples.

[0009] In the above method, the methylation site information may be important methylation site information. The important methylation site may be a methylation site in a region of a gene reported in the literature and / or associated with age changes.

[0010] The gene may be ELOVL2, FHL2, SMARCB1, AEBP2, ZYG11A, OTUD7A, TRIM59, TTC7A, EVL, WBP1L, KLF14, or GRIN1 gene on the human genome.

[0011] The important methylation sites can specifically be the methylation sites targeted by the following probes on the Illumina methylation chip EPIC methylation array (annotation file path: https: / / support.illumina.com.cn / array / array_kits / infinium-methylationepic-beadchip-kit / downloads.html, annotation file update time Mar 13, 2020): cg01912455, cg20793144, cg13868026, cg18071806, cg09038267, cg06784991, cg07955995, cg22454769, cg11667847, cg04875128, cg07553761, and cg16867657.

[0012] Among the genes mentioned above, the nucleotide sequence of the ELOVL2 gene may correspond to positions 10980992-11044538 of chromosome 6 of the human genome GRCh37 version (Gene ID: 54898, update date 17-Jan-2022).

[0013] The nucleotide sequence of the FHL2 gene may correspond to positions 105974169-106054986 of chromosome 2 of the human genome version GRCh37 (Gene ID: 2274, update date 5-Jan-2022).

[0014] The nucleotide sequence of the SMARCB1 gene may correspond to positions 24129153-24180196 of chromosome 22 of the human genome version GRCh37 (Gene ID: 6598, update date 17-Jan-2022).

[0015] The nucleotide sequence of the AEBP2 gene may correspond to positions 19592426-19675161 of chromosome 12 of the human genome GRCh37 version (Gene ID: 121536, update date 5-Jan-2022).

[0016] The nucleotide sequence of the ZYG11A gene may correspond to positions 53308432-53360670 of chromosome 1 of the human genome GRCh37 version (Gene ID: 440590, update date 5-Jan-2022).

[0017] The nucleotide sequence of the OTUD7A gene may correspond to positions 31767601-32162876 of chromosome 15 of the human genome GRCh37 version (Gene ID: 161725, update date 5-Jan-2022).

[0018] The nucleotide sequence of the TRIM59 gene may correspond to positions 160153291-160167574 of chromosome 3 of the human genome GRCh37 version (Gene ID: 286827, update date 5-Jan-2022).

[0019] The nucleotide sequence of the TTC7A gene may correspond to positions 47143005-47303262 of chromosome 2 of the human genome GRCh37 version (Gene ID: 57217, update date 5-Jan-2022).

[0020] The nucleotide sequence of the EVL gene may correspond to positions 100437759-100610573 of chromosome 14 of the human genome GRCh37 version (Gene ID: 51466, update date 5-Jan-2022).

[0021] The nucleotide sequence of the WBP1L gene may correspond to positions 104503705-104576019 of chromosome 10 of the human genome version GRCh37 (Gene ID: 54838, update date 5-Jan-2022).

[0022] The nucleotide sequence of the KLF14 gene may correspond to positions 130415525-130418967 of chromosome 7 of the human genome GRCh37 version (Gene ID: 136259, update date 5-Jan-2022).

[0023] The nucleotide sequence of the GRIN1 gene may correspond to positions 140033606-140063208 of chromosome 9 of the human genome GRCh37 version (Gene ID: 2902, update date 23-Jan-2022).

[0024] In the above method, the age information may be derived from the methylation level file. The age information of the sample to be tested may be unknown, and the age information of the modeling sample may be a known real time age.

[0025] In the above method, the methylation level file may be a file obtained by performing quality control on the raw methylation data.

[0026] The raw methylation data may be the raw data obtained by sequencing the sample through a methylation chip. The methylation chip may be an Illumina methylation chip.

[0027] The quality control may include the following steps: processing the raw methylation data according to the R package (ChAMP): using the parameters cors=5, method=BMIQ, arraytype=corresponding methylation chip type, standardizing the raw data to obtain standardized methylation level data. The obtained standardized methylation data is combined with the age data of the sample (obtained according to the raw data) and then merged into an integrated file. The integration process may be: if the age of the corresponding sample is known, then the corresponding age information is added to the age line of the merged file; if the age information of the corresponding sample is unknown, then the corresponding age line information of the merged file is supplemented with 'unknown', and the merged file is used for subsequent analysis.

[0028] In the method described above, the process of using the elastic net function to construct the model for predicting methylation age may include the steps of model construction and model evaluation.

[0029] The model building may include the following steps:

[0030] A1) Data format conversion: Use the pandas module to convert the known age dataset data into a known age dataset data frame;

[0031] A2) Data splitting: Split the known age dataset data frame into a training dataset and a validation dataset using the train_test_split function in the sklearn module; the validation dataset accounts for 20%;

[0032] A3) Optimal parameter selection: Use GridSearchCV in the model_selection submodule under the sklearn module to optimally select the parameters of the elastic network function, and the selected parameter pool is 'alpha':[0.0001,0.001,0.01,0.1,0,5,1], 'l1_ratio':[0.2,0.1,0.05,0.01];

[0033] A4) obtaining a model: using the elastic network function to train the training data set to obtain the model for predicting methylation age;

[0034] The model evaluation may include using the obtained model for predicting methylation age to predict the methylation age of the validation dataset samples, and calculating the correlation coefficient, mean absolute error, mean square error and / or root mean square error between the methylation age and the chronological age of the validation dataset samples to evaluate the prediction accuracy of the model.

[0035] The closer the correlation coefficient is to 1, the higher the degree of consistency between the methylation age predicted by the model and the real time age. The closer the mean absolute error, mean square error and / or root mean square error is to 0, the higher the degree of consistency between the methylation age predicted by the model and the real time age.

[0036] In the above method, the correlation coefficient may be greater than 0.80.

[0037] In order to solve the above technical problems, the present invention also provides a device for predicting the methylation age of a sample to be tested. The device comprises:

[0038] B1) Data identification module: used to identify whether the sample data is a standard methylation level file; when the sample data to be tested is the original data of methylation sequencing, execute B2) the data quality control module; when the sample data to be tested is the standard methylation level file, execute B3) the methylation site information extraction module; the sample includes a sample to be tested and a modeling sample; the sample data contains age information; the age information of the sample to be tested is unknown, and the age information of the modeling sample is a known real time age.

[0039] B2) Data quality control module: Perform quality control on the data to obtain a standard methylation level file.

[0040] B3) Extracting methylation site information module: used to extract the methylation site information of the standard methylation level file. It can be established by the following method steps: splitting the methylation site information file into a known age data set and an unknown age data set according to age information.

[0041] B4) Constructing a model module for predicting methylation age: used to select the known age dataset described in module B3) and construct a model for predicting methylation age using an elastic network function.

[0042] B5) A module for predicting the methylation age of the sample to be tested: used to predict the methylation age of the sample to be tested using the unknown age data set in B3) and the model for predicting methylation age in B4).

[0043] In the above device, the methylation site information in B3) may be important methylation site information. The important methylation sites may be methylation sites reported in literature or in gene regions related to age changes.

[0044] The gene may be ELOVL2, FHL2, SMARCB1, AEBP2, ZYG11A, OTUD7A, TRIM59, TTC7A, EVL, WBP1L, KLF14, or GRIN1 gene on the human genome.

[0045] The important methylation sites can specifically be the methylation sites targeted by the following probes on the Illumina methylation chip EPIC methylation array (annotation file path: https: / / support.illumina.com.cn / array / array_kits / infinium-methylationepic-beadchip-kit / downloads.html, annotation file update time Mar 13, 2020): cg01912455, cg20793144, cg13868026, cg18071806, cg09038267, cg06784991, cg07955995, cg22454769, cg11667847, cg04875128, cg07553761, and cg16867657.

[0046] Among the genes mentioned above, the nucleotide sequence of the ELOVL2 gene may correspond to positions 10980992-11044538 of chromosome 6 of the human genome GRCh37 version (Gene ID: 54898, update date 17-Jan-2022).

[0047] The nucleotide sequence of the FHL2 gene may correspond to positions 105974169-106054986 of chromosome 2 of the human genome version GRCh37 (Gene ID: 2274, update date 5-Jan-2022).

[0048] The nucleotide sequence of the SMARCB1 gene may correspond to positions 24129153-24180196 of chromosome 22 of the human genome version GRCh37 (Gene ID: 6598, update date 17-Jan-2022).

[0049] The nucleotide sequence of the AEBP2 gene may correspond to positions 19592426-19675161 of chromosome 12 of the human genome GRCh37 version (Gene ID: 121536, update date 5-Jan-2022).

[0050] The nucleotide sequence of the ZYG11A gene may correspond to positions 53308432-53360670 of chromosome 1 of the human genome GRCh37 version (Gene ID: 440590, update date 5-Jan-2022).

[0051] The nucleotide sequence of the OTUD7A gene may correspond to positions 31767601-32162876 of chromosome 15 of the human genome GRCh37 version (Gene ID: 161725, update date 5-Jan-2022).

[0052] The nucleotide sequence of the TRIM59 gene may correspond to positions 160153291-160167574 of chromosome 3 of the human genome GRCh37 version (Gene ID: 286827, update date 5-Jan-2022).

[0053] The nucleotide sequence of the TTC7A gene may correspond to positions 47143005-47303262 of chromosome 2 of the human genome GRCh37 version (Gene ID: 57217, update date 5-Jan-2022).

[0054] The nucleotide sequence of the EVL gene may correspond to positions 100437759-100610573 of chromosome 14 of the human genome GRCh37 version (Gene ID: 51466, update date 5-Jan-2022).

[0055] The nucleotide sequence of the WBP1L gene may correspond to positions 104503705-104576019 of chromosome 10 of the human genome version GRCh37 (Gene ID: 54838, update date 5-Jan-2022).

[0056] The nucleotide sequence of the KLF14 gene may correspond to positions 130415525-130418967 of chromosome 7 of the human genome GRCh37 version (Gene ID: 136259, update date 5-Jan-2022).

[0057] The nucleotide sequence of the GRIN1 gene may correspond to positions 140033606-140063208 of chromosome 9 of the human genome GRCh37 version (Gene ID: 2902, update date 23-Jan-2022).

[0058] The raw methylation data may be the raw data obtained by sequencing the sample through a methylation chip. The methylation chip may be an Illumina methylation chip. The sample may be a biological tissue and / or cell sample.

[0059] The quality control may include the following steps: processing the raw methylation data according to the R package (ChAMP): using the parameters cors=5, method=BMIQ, arraytype=corresponding methylation chip type, standardizing the raw data to obtain standardized methylation level data. The obtained standardized methylation data is combined with the age data of the sample (obtained according to the raw data) and then merged into an integrated file. The integration process may be: if the age of the corresponding sample is known, then the corresponding age information is added to the age line of the merged file; if the age information of the corresponding sample is unknown, then the corresponding age line information of the merged file is supplemented with 'unknown', and the merged file is used for subsequent analysis.

[0060] In the above-mentioned device, the module for constructing a model for predicting methylation age described in B4) may specifically include a model construction module and a model evaluation module.

[0061] The model building module may include the following modules:

[0062] A1) Data format conversion module: used to convert the known age dataset data into a known age dataset data frame using the pandas module.

[0063] A2) Data splitting module: used to split the known age data set data frame into a training data set and a validation data set using the train_test_split function in the sklearn module; the validation data set accounts for 20%.

[0064] A3) Optimal parameter selection module: used to use GridSearchCV in the model_selection submodule under the sklearn module to optimally select the parameters of the elastic network function. The selected parameter pool may be 'alpha': [0.0001, 0.001, 0.01, 0.1, 0, 5, 1], 'l1_ratio': [0.2, 0.1, 0.05, 0.01].

[0065] A4) Obtaining a model module: Specifically, ElasticNet may be used to train the training data set to obtain the model for predicting methylation age.

[0066] The model evaluation includes using the obtained model for predicting methylation age to predict the methylation age of the validation data set samples, and calculating the correlation coefficient, mean absolute error, mean square error, and root mean square error between the methylation age and the chronological age of the validation data set samples to evaluate the prediction accuracy of the model.

[0067] The closer the correlation coefficient is to 1, the higher the degree of consistency between the methylation age predicted by the model and the real time age. The closer the mean absolute error, mean square error and / or root mean square error is to 0, the higher the degree of consistency between the methylation age predicted by the model and the real time age.

[0068] The sample mentioned above may be a human, an animal, a plant or a microorganism.

[0069] In order to solve the above technical problem, the present invention further provides a computer-readable storage medium storing a computer program, wherein the computer program can enable a computer to execute the steps of the method described above.

[0070] The present invention provides a methylation clock age prediction method based on methylation sites corresponding to methylation levels. The present invention constructs a methylation age prediction model requiring the input file to contain important methylation site information at the same time. In the embodiment of the present invention, the R-square, MAE, MSE, and RMSE are calculated between the obtained validation set predicted methylation age and the known time age of the validation set sample to evaluate the prediction accuracy of the obtained model. The results show that the R-square is 0.965, close to 1, indicating that the methylation age of the validation set sample predicted by the model is highly consistent with the true time age of the sample; the MAE value is 3.575, indicating that the age result predicted by using the constructed methylation age prediction model is 3.57 years with the true value.

[0071] The beneficial effects of the present invention are mainly manifested in: using important sites to predict the methylation age of the sample to be tested, retaining more complete information, and obtaining an accurate methylation age prediction value, which is more consistent with the real time age. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 Analysis process of the methylation age assessment method based on DNA methylation levels. DETAILED DESCRIPTION

[0073] The present invention is further described in detail below in conjunction with specific embodiments, and the examples provided are only for illustrating the present invention, rather than for limiting the scope of the present invention. The examples provided below can be used as a guide for further improvements by those of ordinary skill in the art, and do not constitute a limitation of the present invention in any way.

[0074] The experimental methods in the following examples, unless otherwise specified, are all conventional methods, and are performed according to the techniques or conditions described in the literature in the field or according to the product instructions. The materials, reagents, etc. used in the following examples, unless otherwise specified, can all be obtained from commercial channels.

[0075] Example 1: Establishment of a methylation age assessment method

[0076] 1.1 Input data quality control

[0077] When the input sample data used for methylation age assessment is the raw data of methylation chip sequencing, quality control of the raw data is required.

[0078] The specific steps of quality control are: use R package (ChAMP) to process the raw methylation data: use parameters cors = 5, method = BMIQ, arraytype = corresponding methylation chip type, standardize the raw data to obtain standardized methylation level data.

[0079] When the input sample data for methylation age assessment is a standardized methylation level data file, the quality control step can be omitted.

[0080] The above-mentioned samples include samples to be tested and samples to be modeled; the sample data includes age information; the age information of the samples to be tested is unknown, and the age information of the samples to be modeled is the known real time age.

[0081] The obtained standardized methylation data were combined with the age data of the sample (obtained from the original data) and then merged into an integrated file. The integration process was as follows: if the age of the corresponding sample was known (modeling sample), the corresponding age information was added to the age row of the merged file; if the age information of the corresponding sample was unknown, the corresponding age row information in the merged file was supplemented with 'unknown' (sample to be tested), and the merged file was used for subsequent analysis.

[0082] 1.2 Extracting methylation site information

[0083] In the integrated file obtained in step 1.1, the methylation site information of the sample is extracted to obtain the methylation site information file. The methylation site is an important methylation site, and the important methylation site is a methylation site in the region of 12 genes reported in the literature or related to age changes, and the screening correlation threshold is 0.6. The 12 genes include ELOVL2, FHL2, SMARCB1, AEBP2, ZYG11A, OTUD7A, TRIM59, TTC7A, EVL, WBP1L, KLF14 and GRIN1 of the human genome.

[0084] Among them, the nucleotide sequence of the ELOVL2 gene corresponds to positions 10980992-11044538 of chromosome 6 of the human genome GRCh37 version (Gene ID: 54898, update date 17-Jan-2022);

[0085] The nucleotide sequence of the FHL2 gene corresponds to positions 105974169-106054986 of chromosome 2 of the human genome version GRCh37 (Gene ID: 2274, update date 5-Jan-2022);

[0086] The nucleotide sequence of the SMARCB1 gene corresponds to positions 24129153-24180196 of chromosome 22 of the human genome version GRCh37 (Gene ID: 6598, update date 17-Jan-2022);

[0087] The nucleotide sequence of the AEBP2 gene corresponds to positions 19592426-19675161 of chromosome 12 of the human genome version GRCh37 (Gene ID: 121536, update date 5-Jan-2022);

[0088] The nucleotide sequence of the ZYG11A gene corresponds to positions 53308432-53360670 of chromosome 1 of the human genome version GRCh37 (Gene ID: 440590, update date 5-Jan-2022);

[0089] The nucleotide sequence of the OTUD7A gene corresponds to positions 31767601-32162876 of chromosome 15 of the human genome version GRCh37 (Gene ID: 161725, update date 5-Jan-2022);

[0090] The nucleotide sequence of the TRIM59 gene corresponds to positions 160153291-160167574 of chromosome 3 of the human genome version GRCh37 (Gene ID: 286827, update date 5-Jan-2022);

[0091] The nucleotide sequence of the TTC7A gene corresponds to positions 47143005-47303262 of chromosome 2 of the human genome version GRCh37 (Gene ID: 57217, update date 5-Jan-2022);

[0092] The nucleotide sequence of the EVL gene corresponds to positions 100437759-100610573 of chromosome 14 of the human genome version GRCh37 (Gene ID: 51466, update date 5-Jan-2022);

[0093] The nucleotide sequence of the WBP1L gene corresponds to positions 104503705-104576019 of chromosome 10 of the human genome version GRCh37 (Gene ID: 54838, update date 5-Jan-2022);

[0094] The nucleotide sequence of the KLF14 gene corresponds to positions 130415525-130418967 of chromosome 7 of the human genome version GRCh37 (Gene ID: 136259, update date 5-Jan-2022);

[0095] The nucleotide sequence of the GRIN1 gene corresponds to positions 140033606-140063208 of chromosome 9 of the human genome version GRCh37 (Gene ID: 2902, update date 23-Jan-2022).

[0096] The methylation site information file is split into samples according to the age data to obtain a known age data set (information of the modeling sample) and an unknown age data set (information of the sample to be tested). The known age data set is used to establish the methylation clock model; the unknown age sample is used as the sample to be tested, and the methylation age of the sample to be tested is obtained by substituting the unknown age data set into the methylation clock model.

[0097] 1.3 Establishing and validating the methylation clock model

[0098] 1.3.1 Establishing a methylation clock model

[0099] The known age data set (information of known age samples) obtained in step 1.2 was processed using the ElasticNet function to construct a methylation clock model. The processing and construction processes were both processed using Python 3.7.

[0100] 1) Data format conversion: Read the known age dataset data and use the pandas module to convert the data into a data frame.

[0101] 2) Data splitting: Use the train test split function in the sklearn module to split the known age data set into two parts: a training data set and a validation data set. The training data set is used to train the model, and the validation data set is used to validate the obtained model. The validation data set accounts for 20% of the samples, and the parameters used are test_size = 0.2, random_state = 1; the training data set uses the ElasticNet function in the linear_model submodule of the sklearn module for data training.

[0102] 3) Optimal parameter selection: Before training the data, we first use GridSearchCV in the model_selection submodule under the sklearn module to optimally select the parameters of the elastic network function. The selected parameter pool is 'alpha':[0.0001,0.001,0.01,0.1,0,5,1], 'l1_ratio':[0.2,0.1,0.05,0.01].

[0103] 4) Obtaining the model: After the optimal parameters are selected, the training data set is trained using the ElasticNet function. After the training is completed, a methylation age assessment model, namely a methylation clock model, is obtained.

[0104] 1.3.2 Methylation clock model and validation evaluation

[0105] The obtained methylation age assessment model is used to analyze the validation data set to obtain the predicted methylation age results of the validation data set samples, and the correlation coefficient (R-square), mean absolute error (MAE), mean square error (MSE), root mean square error (RMSE), etc. are calculated between the obtained validation set prediction results and the known time age of the validation set samples. The calculation results can objectively reflect the prediction accuracy of the model and evaluate the quality of the model numerically. Among them, the closer R-square is to 1, the higher the degree of consistency between the methylation age of the validation set samples predicted by the model and the real time age of the samples; the closer MAE, MSE and / or RMSE are to 0, the higher the degree of consistency between the methylation age of the validation set samples predicted by the model and the real time age of the samples.

[0106] 1.4 Predicting the methylation age of the sample to be tested based on the methylation clock model

[0107] Select the unknown age sample obtained in step 1.2 as the sample to be tested, and substitute the unknown age data set corresponding to the sample to be tested into the methylation clock model to obtain the methylation age of the sample to be tested.

[0108] Example 2: Methylation clock age prediction for 12 samples

[0109] Sample information: 92 samples (samples in the database https: / / www.ncbi.nlm.nih.gov / geo / geo2r / ?acc=GSE184047 Download), where 80 samples are known in time and age, and 12 samples are unknown in time and age:

[0110] 2.1 Sequencing data quality control

[0111] The following is an integrated file of the methylation level result files and age data of 80 samples after quality control and removal of sites with missing values ​​(Table 1).

[0112] Table 1. Data quality control results statistics

[0113]

[0114]

[0115] Methy_site column (excluding the age row): methylation site name

[0116] Age row: the age of the sample

[0117] Sample columns: Except for the Methy_site column, the remaining columns are the corresponding columns of each sample

[0118] The numerical value corresponding to the methylation site in each sample column represents the methylation level of the site; in the chip data, if the value is greater than 0.6, it means that the site is in a methylated state, and if it is less than 0.2, it means that the site is in an unmethylated state.

[0119] Note: When age is unknown, it means that the age information of the sample needs to be predicted.

[0120] 2.2 Extracting methylation site information

[0121] Extract methylation site information from the integration file in step 2.1

[0122] The methylation sites include important methylation sites, which are methylation sites of genes reported in the literature or related to age changes. The screening correlation threshold is 0.6, including methylation-related sites in the gene regions of 12 genes in the human genome: ELOVL2, FHL2, SMARCB1, AEBP2, ZYG11A, OTUD7A, TRIM59, TTC7A, EVL, WBP1L, KLF14, and GRIN1. Specifically, the following probes on the Illumina methylation chip EPIC methylation array (annotation file path: https: / / support.illumina.com.cn / array / array_kits / infinium-methylationepic-beadchip-kit / downloads.html, annotation file update time Mar 13, 2020) target methylation sites: cg01912455, cg20793144, cg13868026, cg18071806, cg09038267, cg06784991, cg07955995, cg22454769, cg11667847, cg04875128, cg07553761 and cg16867657.

[0123] Among the 12 genes, the nucleotide sequence of the ELOVL2 gene corresponds to positions 10980992-11044538 of chromosome 6 of the human genome version GRCh37 (Gene ID: 54898, update date 17-Jan-2022);

[0124] The nucleotide sequence of the FHL2 gene corresponds to positions 105974169-106054986 of chromosome 2 of the human genome version GRCh37 (Gene ID: 2274, update date 5-Jan-2022);

[0125] The nucleotide sequence of the SMARCB1 gene corresponds to positions 24129153-24180196 of chromosome 22 of the human genome version GRCh37 (Gene ID: 6598, update date 17-Jan-2022);

[0126] The nucleotide sequence of the AEBP2 gene corresponds to positions 19592426-19675161 of chromosome 12 of the human genome version GRCh37 (Gene ID: 121536, update date 5-Jan-2022);

[0127] The nucleotide sequence of the ZYG11A gene corresponds to positions 53308432-53360670 of chromosome 1 of the human genome version GRCh37 (Gene ID: 440590, update date 5-Jan-2022);

[0128] The nucleotide sequence of the OTUD7A gene corresponds to positions 31767601-32162876 of chromosome 15 of the human genome version GRCh37 (Gene ID: 161725, update date 5-Jan-2022);

[0129] The nucleotide sequence of the TRIM59 gene corresponds to positions 160153291-160167574 of chromosome 3 of the human genome version GRCh37 (Gene ID: 286827, update date 5-Jan-2022);

[0130] The nucleotide sequence of the TTC7A gene corresponds to positions 47143005-47303262 of chromosome 2 of the human genome version GRCh37 (Gene ID: 57217, update date 5-Jan-2022);

[0131] The nucleotide sequence of the EVL gene corresponds to positions 100437759-100610573 of chromosome 14 of the human genome version GRCh37 (Gene ID: 51466, update date 5-Jan-2022);

[0132] The nucleotide sequence of the WBP1L gene corresponds to positions 104503705-104576019 of chromosome 10 of the human genome version GRCh37 (Gene ID: 54838, update date 5-Jan-2022);

[0133] The nucleotide sequence of the KLF14 gene corresponds to positions 130415525-130418967 of chromosome 7 of the human genome version GRCh37 (Gene ID: 136259, update date 5-Jan-2022);

[0134] The nucleotide sequence of the GRIN1 gene corresponds to positions 140033606-140063208 of chromosome 9 of the human genome version GRCh37 (Gene ID: 2902, update date 23-Jan-2022).

[0135] The files obtained after extracting the methylation site information were split into samples according to the age data to obtain a known age data set (information of samples with known age, including 80 samples, Table 2) and an unknown age data set (information of samples with unknown age, including 12 samples, Table 3).

[0136] Table 2. Known age dataset

[0137]

[0138] Table 3. Unknown age dataset

[0139]

[0140] 2.3 Methylation clock model construction and evaluation results

[0141] 2.3.1 Construction of methylation clock model

[0142] The known age dataset obtained in step 2.2 was processed and analyzed using the ElasticNet function, and the analysis process was processed using Python 3.7.

[0143] 1) Data conversion: Read the known age dataset data and use the pandas module to convert the data into a data frame.

[0144] 2) Data splitting: Use the train_test_split function in the sklearn module to split the known age data set into two parts: a training data set and a validation data set. Use the parameters test_size = 0.2, random_state = 1, and obtain 64 training data set samples and 16 validation data set samples. The training data set will use the ElasticNet (elastic network) function in the linear_model submodule under the sklearn module for data training.

[0145] 3) Optimal parameter selection: Before training the data, we first use GridSearchCV in the model_selection submodule under the sklearn module to optimally select the parameters of the elastic network function. The selected parameter pool is 'alpha':[0.0001,0.001,0.01,0.1,0,5,1], 'l1_ratio':[0.2,0.1,0.05,0.01].

[0146] 4) Obtaining the model: After the optimal parameters are selected, the training data set is trained using the ElasticNet function. After the training is completed, a methylation age assessment model, namely a methylation clock model, is obtained.

[0147] 2.3.2 Evaluation of the methylation clock model

[0148] The obtained methylation age assessment model was used to analyze the validation data set to obtain the predicted methylation age results of the validation data set samples. The R-square, MAE, MSE, and RMSE were calculated between the obtained validation set predicted methylation age and the known time age of the validation set samples to evaluate the prediction accuracy of the model. The results are shown in Table 4, where the R-square is 0.965, close to 1, indicating that the methylation age of the validation set samples predicted by the model is highly consistent with the true time age of the samples; the MAE value is 3.575, indicating that the average absolute error between the predicted result and the true value is 3.57 years.

[0149] Table 4. Methylation clock model evaluation statistics

[0150]

[0151] Note: Calculated: calculated value; Ideal: optimal value.

[0152] 2.4 Methylation clock age prediction results

[0153] The unknown age samples (including 12 samples) obtained in step 2.2, i.e., the unknown age data set information of the samples to be tested, are substituted into the methylation age assessment model obtained in step 2.3 to obtain the methylation clock age prediction results of the samples to be tested (Table 5).

[0154] Table 5. Methylation clock age prediction results

[0155]

[0156]

[0157] Note: Sample: sample name, PredictAge: predicted age, Age: actual age

[0158] The present invention has been described in detail above. It will be apparent to those skilled in the art that the present invention may be implemented in a wide range under equivalent parameters, concentrations and conditions without departing from the spirit and scope of the present invention and without the need for unnecessary experimentation. Although the present invention provides specific embodiments, it should be understood that further improvements may be made to the present invention. In short, according to the principles of the present invention, this application intends to include any changes, uses or improvements to the present invention, including changes made by conventional techniques known in the art that depart from the scope disclosed in this application. Applications of some of the basic features may be made within the scope of the following appended claims.

Claims

1. A method for predicting methylation age, comprising the following steps: extracting methylation site information from a sample methylation level file to obtain a methylation site information file; splitting the methylation site information file into a known age data set and an unknown age data set according to age information; Using the elastic network function to construct the known age data set to obtain a model for predicting methylation age; Predicting the methylation age of the sample in the unknown age dataset using the model for predicting methylation age; The methylation site information is the methylation sites targeted by the following probes on the Illumina methylation chip EPIC methylation array: cg01912455, cg20793144, cg13868026, cg18071806, cg09038267, cg06784991, cg07955995, cg22454769, cg11667847, cg04875128, cg07553761 and cg16867657.

2. The method according to claim 1, characterized in that: The methylation level file is a file obtained by performing quality control on the original methylation data.

3. The method according to claim 1 or 2, characterized in that: Using an elastic network function to construct a model for predicting methylation age, including steps of model construction and model evaluation; The model construction includes: A1) Data format conversion: Use the pandas module to convert the known age dataset data into a known age dataset data frame; A2) Data splitting: Split the known age dataset data frame into a training dataset and a validation dataset using the train_test_split function in the sklearn module; the validation dataset accounts for 20%; A3) Optimal parameter selection: Use GridSearchCV in the model_selection submodule under the sklearn module to optimally select the parameters of the elastic network function, and the selected parameter pool is alpha: [0.0001, 0.001, 0.01, 0.1, 0.5, 1], l1_ratio: [0.2, 0.1, 0.05, 0.01]; A4) obtaining a model: using the elastic network function to train the training data set to obtain the model for predicting methylation age; The model evaluation includes using the obtained model for predicting methylation age to predict the methylation age of the validation dataset samples, and calculating the correlation coefficient, mean absolute error, mean square error and / or root mean square error between the methylation age and the chronological age of the validation dataset samples to evaluate the prediction accuracy of the model.

4. The method according to claim 3, characterized in that: The correlation coefficient is greater than 0.

80.

5. The method according to claim 1 or 2, characterized in that: The sample is an animal, a plant or a microorganism.

6. A device for predicting the methylation age of a sample to be tested, characterized in that: The device comprises: B1) Data identification module: used to identify whether the sample data is a standard methylation level file; when the sample data to be tested is the original data of methylation sequencing, execute B2) Data quality control module; when the sample data to be tested is the standard methylation level file, execute B3) Methylation site information extraction module; the sample includes a sample to be tested and a modeling sample; the sample data contains age information; the age information of the sample to be tested is unknown, and the age information of the modeling sample is a known real time age; B2) Data quality control module: perform quality control on data to obtain standard methylation level files; B3) Extracting methylation site information module: extracting the methylation site information of the standard methylation level file; splitting the methylation site information file into a known age data set and an unknown age data set according to age information; B4) Constructing a methylation age prediction model module: Selecting the known age data set in module B3) and using the elastic network function to construct a model for predicting methylation age; B5) a module for predicting the methylation age of the sample to be tested: using the unknown age data set in B3) and the model for predicting the methylation age in B4) to predict the methylation age of the sample to be tested; The methylation site information is the methylation sites targeted by the following probes on the Illumina methylation chip EPIC methylation array: cg01912455, cg20793144, cg13868026, cg18071806, cg09038267, cg06784991, cg07955995, cg22454769, cg11667847, cg04875128, cg07553761 and cg16867657.

7. The device according to claim 6, characterized in that: B4) constructing a methylation age prediction model module specifically includes the steps of model construction and model evaluation; The model construction includes: A1) Data format conversion: Use the pandas module to convert the known age dataset data into a known age dataset data frame; A2) Data splitting: Split the known age dataset data frame into a training dataset and a validation dataset using the train_test_split function in the sklearn module; the validation dataset accounts for 20%; A3) Optimal parameter selection: Use GridSearchCV in the model_selection submodule under the sklearn module to optimally select the parameters of the elastic network function, and the selected parameter pool is alpha: [0.0001, 0.001, 0.01, 0.1, 0.5, 1], l1_ratio: [0.2, 0.1, 0.05, 0.01]; A4) obtaining a model: using ElasticNet to train the training data set to obtain the model for predicting methylation age; The model evaluation includes using the obtained model for predicting methylation age to predict the methylation age of the validation dataset samples, and calculating the correlation coefficient, mean absolute error, mean square error and / or root mean square error between the methylation age and the chronological age of the validation dataset samples to evaluate the prediction accuracy of the model.

8. The device according to claim 6 or 7, characterized in that: The sample is an animal, a plant or a microorganism.

9. A computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the steps of the method according to any one of claims 1 to 4.

Images