DNA methylation age prediction method and system based on uniAge model

Abstract

The invention belongs to but is not limited to the technical field of biology, and discloses a DNA methylation age prediction method and system based on a uniAge model, and the method comprises the following steps: screening out CpG sites which can be covered by RRBS and a methylation chip at the same time; screening sites from the common CpG sites and training an age prediction model by using DNAme chip data of human peripheral blood; and correcting the DNAme level obtained by sequencing data, aligning the DNAme level with chip data, or independently fitting a model coefficient on the sequencing data by using the screened CpG site. According to the invention, individual age is predicted by using DNAme levels of a plurality of CpG sites of peripheral blood; and training an age prediction model simultaneously applicable to methylation chip data and sequencing data.

Description

Technical Field

[0001] The present invention relates to the field of biotechnology, in particular to the analysis and interpretation of DNA methylation data, and is applicable to the technical field of detecting biological age through DNA methylation, and specifically to a DNA methylation age prediction method and system based on the uniAge model. Background Art

[0002] Changes in genomic DNA methylation are important drivers and molecular markers of aging. Biological age modeled by DNA methylation in specific tissues can not only be used to reflect an individual's actual age, but also reveal the degree of biological aging of an organism and even predict the risk of age-related diseases, which has significant scientific and clinical significance.

[0003] The Horvaths Clock in 2013 is one of the most famous epigenetic clocks and is applicable to a variety of tissues and cell types, including human blood, skin, and internal organs. The Horvath clock uses 353 CpG sites to predict age, and its results are often considered to be the most widely used and validated biological age indicator. The Hannums Clock in the same year is mainly based on the DNA methylation pattern of blood samples and uses 71 CpG sites to estimate age. Although the application range of the Hannum clock is more limited compared to the Horvath clock, it has higher accuracy in some specific scenarios, such as blood sample analysis. PhenoAge in 2018 combines methylation data and clinical risk factors (such as inflammatory markers, blood biomarkers) to predict biological age. It not only focuses on biological age, but also predicts disease risk and life span. This model is considered to be highly practical in health and disease research. The Skin & BloodClock in the same year is specifically designed for skin and blood samples and is suitable for biological age measurement of specific tissues, especially in cosmetics and skin aging research. GrimAge in 2019 is an improved epigenetic clock that combines DNA methylation with multiple clinical risk factors (such as smoking, inflammation, etc.). It is considered to be more accurate than earlier clocks in predicting lifespan and the occurrence of age-related diseases. DNAmTL (Telomere Length Clock) in 2020 estimates telomere length through methylation patterns and indirectly measures biological age. Telomere length is an important marker of aging, and DNAmTL provides an alternative that does not rely on traditional telomere measurement methods. PACE (PhenoAge Acceleration Clock) in 2021 focuses on the study of accelerated aging, measuring the rate of biological age growth rather than static biological age. This clock is more suitable for evaluating the effectiveness of anti-aging interventions. ImmuneAge in 2022 is specifically used to predict immune function and the degree of immune system aging, especially in vaccine research and geriatric research.

[0004] The existing DNA methylation age clock provides researchers with a powerful tool to understand the aging process, assess disease risk, and develop personalized health intervention programs. However, the existing methylation age clock is mainly developed based on CpG sites measured by DNA methylation chips (such as 450K and EPIC chips). Although this method has been widely used in scientific research and applications, it still has some limitations, especially with the diversification of application scenarios and data types, these limitations are particularly prominent.

[0005] First, the CpG sites used by chip technology are fixed and only cover part of the genome. This means that for DNA methylation age clocks developed using chip technology, the methylation information that can actually be used is limited, and it is difficult to fully reflect the genomic methylation status of an individual. Especially when conducting whole-genome or larger-scale studies, this lack of coverage may lead to an incomplete understanding of the aging process. More importantly, when researchers want to apply these chip-based models to other types of methylation data, such as studies based on sequencing data such as RRBS (Reduced-representation bisulfite sequencing) or WGBS (Whole genome bisulfite sequencing), many key CpG sites are often not covered in these sequencing data. This is because the set of CpG sites covered by reduced bisulfite sequencing (RRBS) and whole genome bisulfite sequencing (WGBS) technologies is not completely consistent with the set of chip technology, which limits the applicability of these methylation age clocks in different data types.

[0006] Secondly, the number and position of fixed sites on the chip are limited, making it difficult to screen out the most effective sites for more cost-effective targeted DNA methylation sequencing when conducting personalized or specific scenario methylation detection. In contrast, targeted DNA methylation sequencing technology has significant cost advantages in large-scale applications, especially when researchers hope to apply methylation clock technology to clinical or large-scale population studies. The high cost of the chip and the uncontrollable number of fixed sites make it unsuitable for these more economical and flexible detection schemes. This further shows that the current methylation clock model lacks sufficient flexibility to adapt to scenarios with different research and application needs. Summary of the invention

[0007] In view of the problems existing in the prior art, the present invention provides a DNA methylation age prediction method and system based on the uniAge model. Unlike traditional DNA methylation age clocks, our model can simultaneously process data from different technical platforms, including data from methylation chips (such as 450K and EPIC) and sequencing technologies (such as RRBS and WGBS). This compatibility ensures that our model can be used in more diverse research and application scenarios, not only covering more CpG sites, thereby improving the coverage of the genome, but also being able to flexibly convert between different methylation data types. In this way, we overcome the limitations of current DNA methylation age in data types and application scenarios, and create a more versatile and adaptable technology platform.

[0008] In addition, our method also lays the foundation for future targeted DNA methylation sequencing technology. This means that researchers can flexibly screen and design CpG site sets that are more suitable for specific scenarios according to different research needs, thereby significantly reducing sequencing costs and improving the accuracy and efficiency of research. The cost advantage of targeted DNA methylation sequencing makes it have great potential in large-scale population research and clinical applications, and our invention provides the underlying technical support for achieving this goal.

[0009] Through this innovative method, we not only solved the technical bottleneck of the existing methylation age clock, but also provided a wider range of possibilities for future research and application. The development of this universal technology will promote the application of aging clocks in a variety of scenarios, thereby helping researchers to explore the aging mechanism more deeply, accurately assess individual health risks, and provide a more reliable basis for individualized health intervention measures.

[0010] The present invention is implemented as follows: a DNA methylation age prediction method based on the uniAge model comprises the following steps: Step 1: Screen out CpG sites that can be covered by both RRBS and methylation arrays; Step 2: Using DNAme chip data from human peripheral blood to screen sites in common CpG sites and train age prediction models; Step three, correct the DNAme levels obtained from the sequencing data and align them with the chip data, or use the screened CpG sites to fit the model coefficients separately on the sequencing data.

[0011] Furthermore, the data used include human whole blood sample 450K and EPIC chip data, sequencing data of human whole blood samples, sequencing and chip data of the same biological samples, and human RRBS data.

[0012] Furthermore, the CpG site initial screening screened the common and "reliable" CpG sites of RRBS and DNAme chips, which met the following conditions: Core sites of the chip: The intersection of sites (probes) covered by the 450K, 850K (EPICv1), and 935K (EPICv2) arrays; The base-converted probe source sequences (50 nt) are uniquely aligned to the end-to-end reference genome UCSC hs1 (T2T) according to the BS sequencing data alignment method; Core site of RRBS: most easily covered by sequencing. RRBS uses MspI endonuclease with a restriction site of C^CGG. Consistently methylated CpG sites: The difference in methylation levels between the positive and negative strands is minimal; The intersection of the microarray sites and the RRBS sites; Located on autosomes.

[0013] Furthermore, the training set and the test set are selected according to the age distribution of all data sets, so that the training set has a certain number of samples in each age range, and the training set and the test set have a relatively consistent age distribution.

[0014] Furthermore, the chip data was preprocessed: only the methylation values ​​(beta values) of the initial screening sites were retained, 65,221 sites; Remove samples with missing age and samples with age < 1 or age > 120; Sites with a deletion ratio >10% in all samples were removed; Samples with a site deletion ratio >10% were removed; For each site, missing values ​​of methylation were filled with random normally distributed values, and the mean and variance were calculated from the non-missing values ​​of that site.

[0015] Further, age-related CpG site screening: Age-related methylation sites were further screened from 65,221 initial screening sites. It is i Sample No. j Methylation level of each CpG site (beta value); Calculate the inner quarter range (IQR) of the methylation level of each site. ; Calculate the correlation between methylation at each site and age, Pearsons correlation and Spearmans correlation ; definition ; according to Sort the CpG sites from large to small, and take the 6000 CpG sites with the largest scores to enter the following model.

[0016] Furthermore, the LASSO model was used to screen CpG sites. The input data is 7149 chip training set samples (rows), including variables age, data set, gender, and methylation beta values ​​of 6000 CpG sites (columns). The LASSO model is fitted using the R package glmnet

[0017] . Use R's machine learning framework mlr3 to call glmnet. The learner parameters are lrn("regr.cv_glmnet",alpha = 1, lambda = 10^(seq(-5, 0, length.out=100)),s = 'lambda.1se',standardize = T, intercept = T, type.measure = 'mae'). You can also use glmnet training directly without using mlr3.

[0018] Further, the Ridge regression age prediction model: Ridge regression was used to fit age at the CpG sites selected by the LASSO model. The model was:

[0019] . The learner parameters of glmnet called by mlr3 are lrn("regr.cv_glmnet", alpha = 0, lambda = 10^(seq(-8, 0, length.out=100)), s = 'lambda.1se', standardize = F, intercept = T, type.measure = 'mae'). Among them, alpha = 0 indicates fitting ridge regression; This model is denoted as full model or uniAge-605, which contains 605 CpG sites and the estimated age is

[0020] in is the coefficient of the ridge regression estimate, It is j Methylation beta value of CpG sites, .

[0021] Furthermore, when the model is applied to sequencing data, quantile normalization is used to correct the distribution of methylation levels of 605 (100) sites in the sequencing data to be the same as the methylation distribution of the 605 (100) sites in the chip data training set. This correction can slightly improve the performance of the model on sequencing data.

[0022] Another object of the present invention is to provide a DNA methylation age prediction method based on the uniAge model and a DNA methylation age prediction system based on the uniAge model, comprising: CpG site screening module, used to screen out CpG sites that can be covered by both RRBS and methylation chips; Age prediction model training module, used to screen sites in common CpG sites and train age prediction models using DNAme chip data of human peripheral blood; The sequencing data correction module is used to correct the DNAme levels obtained from the sequencing data, align them with the chip data, or use the screened CpG sites to fit the model coefficients separately on the sequencing data.

[0023] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows: First, the present invention uses DNA methylation levels at several CpG sites in peripheral blood to predict individual age. Peripheral blood is an easily accessible human tissue with no (little) contamination, and peripheral blood has the most available data. These conditions ensure that the present invention can obtain a relatively accurate, robust, and easy-to-use DNA methylation age prediction model.

[0024] Second, the present invention uses the common CpG sites of the core sites of the DNA methylation chip (the probe intersection of the 450K and EPIC chips) and the core sites of the RRBS sequencing as the candidate feature set of the prediction model. The 100 CpG sites used by the uniAge model can be captured in both chip samples and sequencing samples. Among them, in actual RRBS samples, an average of 90% of the 100 CpG sites of the uniAge model are covered by sequencing, while only (on average) 30% of the CpG sites of the existing model are covered. This unified CpG site design makes the uniAge model (compared with existing models) widely applicable.

[0025] Third, the present invention adopts a step-by-step screening and refinement of CpG site screening methods, and finally determines the uniAge model of 100 CpG sites. In the evaluation of a large amount of real data, the uniAge model of 100 CpG sites has the same level of prediction accuracy as the complete model containing 605 CpG sites without limiting the number of sites. Low-cost targeted DNA methylation sequencing can be easily applied to the prediction model (100) with a controllable number of CpG sites, which is extremely beneficial for large-scale research and application. The existing HorvathAge relies on the input of 353 CpG sites, and GrimAge relies on the input of 1030 CpG sites. Excessive reliance on the number of sites is not conducive to the application of targeted DNA methylation sequencing.

[0026] Fourth, as auxiliary evidence of the inventiveness of the claims of the present invention, the technical solution of the present invention has broad commercial prospects in the fields of health management, anti-aging, insurance and precision medicine after transformation: a) Precision health management and personalized medicine Health assessment and monitoring: Commercial DNA methylation testing services can provide users with personalized biological age assessments. Compared with traditional health checks, DNA methylation age can detect potential health problems earlier, thereby helping users take early intervention measures.

[0027] Anti-aging and longevity intervention: Many anti-aging companies use DNA methylation age data to design personalized lifestyle adjustments and supplement programs for users. By regularly monitoring changes in methylation age, the effectiveness of anti-aging measures can be evaluated.

[0028] Precision medicine: In the field of precision medicine, DNA methylation age can help doctors develop treatment plans that are more suitable for patients, especially in cancer treatment and chronic disease management.

[0029] b) Life insurance and health insurance Risk assessment: Insurance companies can use DNA methylation age as part of health risk assessment. Biological age reflects an individual's health status better than actual age, thus helping insurance companies to more accurately assess premiums and claims risks.

[0030] Personalized insurance products: By combining DNA methylation age, insurance companies can design more flexible personalized insurance products, such as offering preferential premiums to people in good health and with a lower biological age.

[0031] c) Enterprise employee health management Corporate health programs: Some companies have begun to introduce health management programs based on DNA methylation age to help employees understand their biological age and delay aging, improve work efficiency and happiness through health intervention programs.

[0032] Employee well-being and performance improvement: Through regular biological age monitoring, companies can optimize employee health programs, thereby reducing sick leave and improving work efficiency.

[0033] d) Anti-aging and beauty industry Marketing of anti-aging products: Beauty and skin care companies can develop personalized skin care solutions based on DNA methylation age test results and use this as a selling point to promote products. For example, a methylation age clock for skin aging can help optimize the design and effect evaluation of anti-aging products.

[0034] Personalized recommendations based on biological age: By detecting the biological age of the skin, the beauty industry can provide customers with targeted skin care and beauty recommendations, thereby improving customer experience and loyalty.

[0035] e) Nutrition and Supplement Market Personalized nutrition plans: Based on the test results of DNA methylation age, nutrition and health companies can customize personalized diet and supplement plans for users to help delay biological aging.

[0036] Effect evaluation: Supplement companies can use DNA methylation age to prove the effectiveness of their products, enhancing the scientific endorsement and market competitiveness of their products.

[0037] f) Data and analysis services Big data analysis and prediction: Genetic testing companies and health data analysis companies can use the large-scale DNA methylation data collected to analyze health trends and predict diseases. This type of data is of great value in the fields of pharmaceuticals, public health, and scientific research.

[0038] g) Consumer genetic testing services Direct-to-consumer testing services: Genetic testing companies can offer DNA methylation age testing as part of their health and genetic analysis products to provide consumers with more personalized health information.

[0039] Health and aging tracking applications: Some startups have developed DNA methylation age detection products that can be combined with mobile applications to help consumers track their aging process through data visualization. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of a DNA methylation age prediction method based on the uniAge model provided in an embodiment of the present invention; Figure 2 is a schematic diagram of age distribution of each data set provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of the correlation between methylation of CpG sites and age provided in an embodiment of the present invention; Figure 4 Schematic diagram of the distribution of IQR of CpG site methylation level provided by an embodiment of the present invention; Figure 5 is a structural diagram of a DNA methylation age prediction system based on the uniAge model provided in an embodiment of the present invention; Figure 6 It is a schematic diagram of the fitting value of uniAge on the training set provided by an embodiment of the present invention; Figure 7 It is a schematic diagram of the predicted value of uniAge on a chip data test set provided by an embodiment of the present invention; Figure 8 It is a schematic diagram of the predicted value of the existing model provided by an embodiment of the present invention on the chip data test set; Fig. 9 It is a schematic diagram of the prediction values ​​of different models provided by the embodiments of the present invention on the sequencing data test set; Fig.10 It is a schematic diagram of the missing ratio of sites in the sequencing data test set of different models provided in the embodiments of the present invention. DETAILED DESCRIPTION

[0041] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0042] like Figure 1 As shown, a DNA methylation age prediction method based on the uniAge model provided in an embodiment of the present invention comprises the following steps: Step 1: Screen out CpG sites that can be covered by both RRBS and methylation arrays; Step 2: Using DNAme chip data from human peripheral blood to screen sites in common CpG sites and train age prediction models; Step three, correct the DNAme levels obtained from the sequencing data and align them with the chip data, or use the screened CpG sites to fit the model coefficients separately on the sequencing data.

[0043] (Public) Datasets There are 4 parts of data involved: 1. Human whole blood sample 450K and EPIC chip data, used for CpG site screening and (chip) model training, a total of 29 data sets and 15,610 samples; 2. Sequencing data of human whole blood samples, used to evaluate the performance of the model on sequencing data, a total of 2 data sets and 102 samples; 3. Sequencing and chip data of the same biological sample, used to correct technical deviations between sequencing and chip, from two sources: GEO (Gene Expression Omnibus) and ENCODE (The Encyclopedia of DNA Elements). There are 4 data sets and 92 samples from GEO, and 155 samples from ENCODE; 4. Human RRBS data, used for initial site screening, totaling 4 data sets and 1,835 samples.

[0044] For data summary information, see "data / DNA methylation age dataset.xlsx" The phenotypic information of each dataset sample, such as age, gender, and disease, can be found in “data / {dataset}.tsv”. The script “data / pheno-data.R” for organizing the phenotypic information is provided for reference.

[0045] The processed data of some (15 data sets) chip data were downloaded for model training and evaluation. In order to facilitate data reading, some files were modified, such as the variable name in the first line. The modified processed DNAme data (beta value) is shown in "meth-data / *" Existing DNAme age models Name model Phenotype Platforms Samples Probes / CpGs probe pool PMID HorvathAge Linear model, age transformation Chronological age 450K / 27K 353 21,369,Intersectof 450K and27K 24138928,25968125 HannumAge Linear Model Chronological age 450K 656 71 450K 23177740 SkinBloodAge Linear model, age transformation Chronological age EPIC / 450K 2222 frommultipletissues / celltypes 391 Intersect of 450K and EPIC 30048243 ZhangAge Linear model, DNAme normalization Chronological age EPIC / 450K 13,661,most blood samples 514 31443728 PhenoAge Linear model, time of death to age transformation Mortality EPIC / 450K / 27K 513 29676998 GrimAge, (Online only) Mortality EPIC / 450K 2356 1030 450,161CpGs,Intersectof 450K andEPIC 30669119 The above models used for evaluation are compiled from the original article and supplementary materials. The supplementary materials of the article can be found in the corresponding folders of the "models / " directory. The CpG sites (chip probes) and corresponding coefficients of each model can be found in "models / {model}.tsv", and the program for calculating each model can be found in "models / DNAme-ages.R". GrimAge can only be used by online registration.

[0046] uniAge model Initial screening of CpG sites The purpose of the initial screening of the model is to screen the common and "reliable" CpG sites of RRBS and DNAme chips, which meet the following conditions 1. Core sites of the chip: the intersection of sites (probes) covered by the 450K, 850K (EPICv1), and 935K (EPICv2) arrays 2. The probe source sequences (50nt) after base conversion (according to the BS sequencing data alignment method) are uniquely aligned to the end-to-end reference genome UCSC hs1 (T2T) 3. RRBS core site: the easiest to be covered by sequencing, RRBS uses MspI endonuclease, and the restriction site is C^CGG 4. Consistently methylated CpG sites: The methylation levels of the positive and negative strands differ minimally 5. Intersection of microarray sites and RRBS sites 6. Located on autosomes Chip probe screening The intersection of the sites (probes) covered by the 450K, 850K (EPICv1), and 935K (EPICv2) arrays is nProbes = 369,845. The 50nt source sequence of the probes is mapped to the UCSC hs1 reference genome using bsmap (default parameters). Probes that are not uniquely mapped are removed, leaving 368,540 probes, which are recorded as set A. RRBS core site Download the processed files of the multi-tissue RRBS datasets GSE143673 (n = 451), GSE203169 (n = 82), GSE231984 (n = 781), GSE233417 (n = 521), nTotal = 1835.

[0047] By filter criteria GSE143673: DP>=10, sample proportion >= 0.4: number=1285147 GSE203169: DP>=5, sample proportion >= 0.9: number=855454 GSE231984: DP>=5, sample proportion >= 0.8: number=2532617 GSE233417: DP>=1, sample proportion >= 0.95: number=4921971 CpG sites that appear in at least three of the four sets are defined as RRBS core sites (denoted as set B1), nCpGs = 1,448,485 Discordant strand methylation CpG sites Using datasets GSE143673, GSE203169, GSE231984, nTotal = 1314. According to the screening criteria, the absolute value of the difference in methylation between the positive and negative strands of CpG is >= 0.2 in the three data sets. GSE143673: DP>=10, sample proportion >= 0.05: number=60204 GSE203169: DP>=5, sample proportion >= 0.1: number=33925 GSE231984: DP>=5, sample proportion >= 0.2: number=89610 The CpG sites that appear in any of the above sets (the union of the sites) are called inconsistent strand methylation CpG sites (denoted as set B2), nCpG = 174,072 The available CpG sites screened using RRBS data are set B = B1- B2, nCpGs = 1,315,096 The CpG sites on the autosomes are set C, and the CpG sites for the initial screening are A∩B∩C, nCpGs = 65,221. The information of the initial screening sites can be found in the file "uniAge / final-probe-autosomes.tsv" The scripts “uniAge / statCoverage.py” and “uniAge / count-consistent-CpGs.ipynb” used for preliminary screening of loci are provided for reference.

[0048] Training set and test set like Figure 2 As shown, the training set and test set are selected according to the age distribution of all data sets, so that the training set has a certain number of samples in each age range, and the training set and test set have relatively consistent age distribution.

[0049] 1. Training set: a) Microarray data: GSE157131, GSE55763, GSE87648, GSE40279, GSE84727, GSE72680, GSE73103, GSE72775, GSE168739, nTotal = 7149 2. Test set: a) Microarray data: GSE125105, GSE42861, GSE80417, GSE147221, GSE72773, GSE61496, nTotal = 3323 b) Sequencing data: GSE117593, GSE85928, nTotal = 102 A total of 10,472 samples were used in the chip and 102 samples were sequenced. Chip data preprocessing Only the methylation value (beta value) of the initial screening site is retained, 65,221 sites 1. Remove samples with missing age and samples with age < 1 or age > 120 2. Remove sites with a missing ratio > 10% in all samples 3. Remove samples with site deletion ratio >10% 4. For each site, missing values ​​of methylation are filled with random normally distributed values, and the mean and variance are calculated from the non-missing values ​​of the site Age-related CpG site screening Age-related methylation sites were further screened from 65,221 initial screening sites. It is i Sample No. j Methylation level of CpG sites (beta value) 5. Calculate the inner quarter range (IQR) of the methylation level of each site.

[0050] 6. Calculate the correlation between methylation at each site and age, Pearsons correlation and Spearmans correlation

[0051] 7. Definitions

[0052] 8. Follow Sort the CpG sites from large to small, and get the 6000 CpG sites with the largest scores to enter the following model like Figure 3 , 4 As shown in the figure, this step screens the CpG sites with the greatest correlation with age and the largest range of methylation changes. Since the fluorescence value of the chip is continuously distributed, this screening removes those sites whose correlation with age depends on weak methylation changes. Tests show that this step can improve model performance.

[0053] The LASSO model was used to screen CpG sites. The input data was 7149 chip training set samples (rows), including variables age, data set, gender, and methylation beta values ​​of 6000 CpG sites (columns). The LASSO model was fitted using the R package glmnet.

[0054] . Use R's machine learning framework mlr3 to call glmnet. The learner parameters are lrn("regr.cv_glmnet",alpha = 1, lambda = 10^(seq(-5, 0, length.out=100)), s = 'lambda.1se',standardize = T, intercept = T, type.measure = 'mae'). You can also use glmnet training directly without using mlr3.

[0055] Cross Validation The model hyperparameter lambda was determined using 10-fold cross validation. Due to the randomness of cross validation, the 10-fold cross validation was repeated 5 times to fit the LASSO model. Each time the model was fitted, a variable (CpG site) could be selected, which is recorded as Finally, the sites that appear in at least 4 model selections are selected as model sites, including the CpG site set . dataset is not the variable of interest.

[0056] Batch effect. Since different data sets come from different studies and use different preprocessing methods, the batch effect between data sets is significant. The batch effect correction methods limma:removeBatchEffect(), sva:ComBat() (parameter method) and quantile normalization did not improve the model performance. Therefore, the batch effect of the training data was not corrected in advance, but added to the model. term to model batch effects.

[0057] Ridge regression age prediction model Ridge regression was used to fit age at the CpG sites selected by the LASSO model. The model is:

[0058]

[0059] The learner parameters of glmnet called by mlr3 are lrn("regr.cv_glmnet", alpha = 0,.lambda = 10^(seq(-8, 0, length.out=100)),s = 'lambda.1se', standardize = F,intercept = T, type.measure = 'mae'). Among them, alpha = 0 indicates fitting ridge regression.

[0060] The results show that the difference between genders is small and can be ignored. Using limma:removeBatchEffect(), sva:ComBat() (parametric method) and quantile normalization before coefficient estimation did not improve the model performance.

[0061] This model is denoted as full model or uniAge-605, which contains 605 CpG sites and the estimated age is

[0062] in is the coefficient of the ridge regression estimate, It is j Methylation beta value of CpG sites, . File List: 1.uniAge / clean-meth-data.R: script for reading methylation level data and phenotype data 2.uniAge / uniAge.R: scripts for chip data preprocessing, age-related CpG site screening, full model site selection and parameter estimation 3.uniAge / model-uniAge.csv: full model sites and coefficients 4.uniAge / uniAge-on-train-datasets.csv: full model fitted values ​​of age in the training set 5.uniAge / uniAge-on-test-datasets.csv: full model prediction value of chip test set age 100-site model (uniAge-100), in order to facilitate the application of targeted methylation sequencing, 100 sites are selected from the 605 sites of the full model, called the reduced model or uniAge-100. In theory, the full model without the restriction of the number of sites should perform better. The full model and the reduced model are retained to facilitate the comparison of the performance loss of the reduced model. For chip and whole genome sequencing data, there is no need to be restricted by the number of CpG sites, so there is no need to use a simplified model. The evaluation on the test set shows that the performance of the reduced model and the full model is almost the same.

[0063] The same process as the LASSO regression site screening and ridge regression coefficient estimation of the full model is used for the full model. Each CpG site continues to be screened using LASSO, and the appropriate lambda parameter is selected to make the number of selected sites exactly 100. Then, ridge regression is used to fit the age with the methylation beta values ​​of these 100 sites to obtain the reduced model, uniAge-100. File List: 1.uniAge / reduced-model.R: script for site selection and parameter estimation of the reduced model from the full model 2.uniAge / reduced-model-uniAge.csv: sites and coefficients of the reduced model 3.uniAge / reduced-uniAge-on-train-datasets.csv: reducedmodel fitted values ​​of age in the training set 4.uniAge / reduced-uniAge-on-test-datasets.csv: reducedmodel predicted value of chip test set age Nonlinear effects Consider whether adding the nonlinear effect of site methylation on age can significantly improve model performance. CpG sites at high, medium, and low methylation levels may have different effects (slopes) on age. For the methylation level variable of each site in the full model, construct piecewise linear B-splines. Use splines:bs(meth, degree = 1, knots =c(0.3, 0.7)) to construct splines, and use ridge regression to fit the parameters, that is,

[0064] in It is i Sample No. j The methylation beta value of each CpG site, is the parameter to be estimated, , .

[0065] The piecewise linear model did not improve model performance.

[0066] Sequencing data correction: When applying the model to sequencing data, quantile normalization is used to correct the distribution of methylation levels of 605 (100) sites in the sequencing data to be the same as the methylation distribution of the 605 (100) sites in the chip data training set. This correction can slightly improve the performance of the model on sequencing data. This method is ineffective on the test set of chip data.

[0067] For the distribution correction method, see the function quantile_align() in "models / DNAme-ages.R", and for the average methylation level of the model site in the chip training set, see "uniAge / mean-meth-training-set-[reduced-]uniAge.tsv" like Figure 5 As shown, an embodiment of the present invention provides a DNA methylation age prediction method based on the uniAge model and a DNA methylation age prediction system based on the uniAge model, comprising: CpG site screening module, used to screen out CpG sites that can be covered by both RRBS and methylation chips; Age prediction model training module, used to screen sites in common CpG sites and train age prediction models using DNAme chip data of human peripheral blood; The sequencing data correction module is used to correct the DNAme levels obtained from the sequencing data, align them with the chip data, or use the screened CpG sites to fit the model coefficients separately on the sequencing data.

[0068] The specific application areas and related products of DNA methylation age cover health management, anti-aging, insurance, nutrition, and genetic testing. The following are some specific application areas and their representative products: a) Health management and biological age monitoring Application areas: Through DNA methylation age measurement, individuals can obtain biological age compared to their actual age. This information is helpful for personalized health management and early intervention.

[0069] Related Products: MyDNAge: Developed by Zymo Research, it provides an assessment of an individual's biological age by analyzing the DNA methylation status in blood or saliva samples.

[0070] Elysium Health Index: Provided by Elysium Health, which uses DNA methylation technology to assess biological age and combines it with lifestyle recommendations to optimize health management.

[0071] Chronomics: This is a service that provides comprehensive DNA methylation age testing, including health status analysis and lifestyle recommendations, mainly for personalized health management.

[0072] b) Anti-aging and beauty Application Areas: In the anti-aging field, DNA methylation age is used to evaluate the effectiveness of various anti-aging therapies, including skin care products, supplements and lifestyle interventions.

[0073] Related Products: EpiAging Clock by TruDiagnostic: TruDiagnostic offers tests that measure multiple epigenetic clocks to help users understand the biological aging of the skin and the whole body and provide validation for the effectiveness of anti-aging products and treatments.

[0074] OneSkin: This is a skin health company whose products combine DNA methylation data and claim to improve skin quality by reversing the biological age of the skin.

[0075] c) Personalized nutrition and dietary management Application areas: By analyzing DNA methylation age and other biomarkers, personalized diet and nutritional supplement plans can be developed to help delay the aging process.

[0076] Related Products: GlycanAge: This product not only analyzes DNA methylation, but also combines other biological markers to provide users with health and nutrition advice aimed at delaying biological aging.

[0077] InnerAge 2.0 by InsideTracker: InsideTracker’s product combines methylation age with other blood markers to provide users with personalized nutrition and lifestyle recommendations.

[0078] d) Insurance and risk assessment Application Areas: Insurance companies use DNA methylation age as part of risk assessment to more accurately customize premiums and products.

[0079] Related Products: Revolution Insurance by YouSurance: This is a service that incorporates biological age into health insurance assessments, providing customers with personalized insurance plans through DNA methylation age testing.

[0080] e) Scientific research and medical research Application areas: DNA methylation age is used to study the early diagnosis of diseases, aging mechanisms and the effectiveness of health interventions, and plays an important role in basic scientific research.

[0081] Related Products: Illumina EPIC Array: A high-throughput sequencing platform for large-scale epigenetic studies that can analyze hundreds of thousands of CpG sites associated with DNA methylation and is widely used in biological age research.

[0082] Horvaths Clock Toolkits: A variety of research tools and packages to help researchers analyze DNA methylation age in different tissues.

[0083] f) Consumer genetic testing and lifestyle optimization Application areas: Genetic testing companies provide direct-to-consumer products that combine DNA methylation age analysis to help individuals optimize their lifestyles.

[0084] Related Products: 23andMe + DNA Methylation Analysis: Although 23andMe’s main business is genetic testing, they also provide methylation age analysis in cooperation with third-party platforms.

[0085] EpigenCare: This product provides personalized recommendations for skin care by analyzing methylation patterns in skin samples.

[0086] g) Dynamic aging tracking and health optimization Application areas: By regularly tracking age changes in DNA methylation, users can dynamically monitor their aging process and evaluate the effectiveness of interventions.

[0087] Related Products: TrueAge by TruMe: Provides regular biological age monitoring services, combined with health advice, to help users optimize their lifestyle and adjust their health plans in real time.

[0088] EpiAging Monitoring by Novogenia: Providing long-term methylation age monitoring and trend analysis to support ongoing health and anti-aging interventions.

[0089] Datasets used for uniAge performance evaluation: like Figure 6As shown, training sets GSE157131, GSE55763, GSE87648, GSE40279, GSE84727, GSE72680, GSE73103, GSE72775, GSE168739, nTotal = 7149 Distribution of the absolute value of the deviation of the uniAge fitting value on the training set Mean absolute deviation uniAge-605 uniAge-100 50% (median) 2.095132 2.606567 75% 3.629820 4.580433 90% 5.273679 6.755359 95% 6.603948 8.357867 mean 2.59369 3.27759 Chip data test set like Figure 7 As shown, the test set is GSE125105, GSE42861, GSE80417, GSE147221, GSE72773, GSE61496, nTotal = 3323 UniAge performs well on chip datasets: like Figure 6 As shown in the figure, the 100-site uniAge model has good fitting accuracy in the chip data training set (n=7149), with a root mean square error (RMSE) of 3.44 (years) and a mean absolute error of 3.72 (years).

[0090] Distribution of absolute value of uniAge prediction deviation on chip data test set Mean absolute deviation uniAge-605 uniAge-100 50% (median) 2.579847 3.023845 75% 4.498416 5.263006 90% 6.519476 7.748378 95% 8.365930 9.358484 mean 3.236782 3.72195 like Figure 7 As shown, the RMSE of the 100-site uniAge model in an independent chip data test (n=3323) is 4.81 (years), which is better than the existing prediction error of DNA methylation age ( Figure 8 ), among which the RMSE of HorvathAge is 26.15 (years), the RMSE of HannumAge is 26.68 (years), and the RMSE of Skip&BloodAge is 22.79 (years), while the systematic deviation of ZhangAge is more significant.

[0091] uniAge also outperforms existing methods on sequencing datasets: The performance of uniAge and existing methods was evaluated using bisulfite sequencing RRBS and WGBS samples: RRBS: GSE85928; WGBS: GSE117593, nTotal = 102 like Fig. 9As shown, for bisulfite sequencing samples (RRBS and WGBS, n=102), the root mean square error (RMSE) of uniAge is 15.72 (years), which is better than several existing commonly used methods, among which the RMSE of HorvathAge is 21.65 (years), the RMSE of HannumAge is 23.86 (years), and the RMSE of Skip&BloodAge is 17.44 (years).

[0092] on the other hand, Fig.10 The missing ratio of CpG sites required by different models in DNA methylation sequencing samples is shown. Through the preliminary screening of RRBS core sites, the site missing of the uniAge model is much lower than that of other models. In the evaluated RBBS samples, the CpG site missing ratio of the uniAge model is about 10% on average, while the average missing ratio of several existing methods is about 70%. Compared with existing methods, the missing ratio of uniAge has been significantly improved, which undoubtedly makes the prediction results of age more reliable.

[0093] Improvement work direction: Ridge model bias correction LASSO, Ridge, and Elastic net are all (regularized) biased estimates, which result in larger age predictions for young people and smaller age predictions for older people. Although this bias is a compromise to reduce the prediction variance (minimize the prediction squared loss), sometimes the bias is too large and can be corrected, considering the de-biased estimate of Ridge regression (Zhang and Politis, 2021)

[0094] and They are the classical and de-biased Ridge estimates respectively. Compare The bias of is smaller and is not an unbiased estimate. It is not clear which software package implements the de-biased Ridge estimate.

[0095] Correction of sequencing data Judging from the current results, the technical deviation of sequencing data and chip data has a great impact on the model. The quantile normalization that has been tried is effective but not sufficient. For each CpG site, the beta value distribution of the chip training set of the site is used as the prior, the read count and beta value in the sequencing sample are used as the likelihood, and the posterior mean of the beta value is used as the correction, so that the beta value of the CpG site obtained by the sequencing data is closer to the chip data.

[0096] Fit model coefficients on sequencing data alone: ​​Use the 605 or 100 CpG sites that have been determined to re-estimate the coefficients of the linear model on the sequencing samples. Since the current sequencing data is relatively small (sample number 102), consider an empirical Bayes linear model, and the prior distribution of the model coefficients is determined by the chip data. Specifically, randomly extract 102 chip training sets, fit the Ridge model, and obtain the coefficients of 605 (100) sites. Repeat the extraction 1000 times to obtain a distribution for each coefficient. This is the prior distribution of the coefficients, and the posterior distribution is calculated on the sequencing samples to obtain the model of the sequencing data. If there are about 500 sequencing samples, you can directly fit the sequencing data.

[0097] Example 1: DNA methylation age prediction using 450K chip and RRBS data 1. Sample preparation and data acquisition: Human whole blood samples were collected, and the methylation chip data were acquired using the Illumina HumanMethylation450K chip. RRBS sequencing was performed on the same samples at the same time.

[0098] Extract CpG site information from 450K chip data and RRBS data.

[0099] 2. Screening of CpG sites: According to step 1, CpG sites that can be covered by both 450K chip and RRBS were screened.

[0100] During the screening process, the intersection of the coverage sites of the 450K chip and RRBS was used, while ensuring that these sites were core sites on the chip and that the RRBS sequencing coverage was high.

[0101] Ensure that the selected CpG sites have minimal differences in methylation levels on the positive and negative strands and are located on autosomes.

[0102] 3. Model training: The common CpG sites in the 450K chip data obtained from peripheral blood samples were used to train an age prediction model based on the methylation levels of these sites.

[0103] Use linear regression or other suitable machine learning algorithms to fit the model coefficients on the training set to maximize prediction accuracy.

[0104] 4. Data correction and prediction: The DNA methylation levels of the RRBS sequencing data were corrected to align with the 450K chip data.

[0105] Age prediction was performed on RRBS data using a model trained on 450K chip data.

[0106] 5. Result verification: The predicted DNA methylation age was compared with the actual age to evaluate the accuracy and stability of the model.

[0107] The experiments were repeated multiple times to ensure the reliability of the results.

[0108] This embodiment is particularly suitable for scientific research experiments, which require accurate prediction of the age of biological samples from DNA methylation data from different sources, and is suitable for human genome research and disease prediction.

[0109] Example 2: DNA methylation age prediction using EPICv1 chip and RRBS data 1. Sample preparation and data acquisition: DNA methylation array analysis of human whole blood samples was performed using the Illumina MethylationEPIC v1 chip (850K), and RRBS sequencing was performed on the same samples simultaneously.

[0110] Extract relevant CpG site information from EPIC chip and RRBS sequencing data.

[0111] 2. Screening of CpG sites: According to steps one and three, “reliable” CpG sites covered by both the EPIC chip and RRBS were screened out.

[0112] During the screening process, the intersection of the EPIC chip core sites and the sites covered by RRBS sequencing was used, and it was ensured that these sites were uniquely mapped to the reference genome UCSC hs1 (T2T) on the chip.

[0113] Ensure that the selected sites have consistent methylation levels and are located on autosomes.

[0114] 3. Model training: Using CpG sites screened from the EPIC chip data, an age prediction model was trained using chip data from whole blood samples based on the methylation levels of these sites.

[0115] Apply the uniAge model for fitting and generate the coefficients of the prediction model.

[0116] 4. Data correction and prediction: Correct the RRBS data and align them with the EPIC chip data, or directly fit the model coefficients on the selected CpG sites.

[0117] Use the above trained model to predict age on RRBS data.

[0118] 5. Result verification: The predictions were compared with the actual age of the samples to assess the accuracy of the model.

[0119] Compare the prediction results of different batches of samples to verify the stability of the model on different data sets.

[0120] This embodiment is particularly suitable for DNA methylation age prediction research in large-scale populations, especially when both chip and sequencing data are available in the sample, and is more suitable for precision medicine research and biomarker discovery.

[0121] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A DNA methylation age prediction method based on the uniAge model, characterized in that: The following steps are involved: Step 1: Screen out CpG sites that can be covered by both RRBS and methylation arrays; Step 2: Using DNAme chip data from human peripheral blood to screen sites in common CpG sites and train age prediction models; Step three, correct the DNAme levels obtained from the sequencing data and align them with the chip data, or use the screened CpG sites to fit the model coefficients separately on the sequencing data.

2. The DNA methylation age prediction method based on the uniAge model according to claim 1, characterized in that: The data used include 450K and EPIC chip data of human whole blood samples, sequencing data of human whole blood samples, sequencing and chip data of the same biological samples, and human RRBS data.

3. The DNA methylation age prediction method based on the uniAge model according to claim 1, characterized in that: CpG site screening: Screen the common and "reliable" CpG sites of RRBS and DNAme chips, which meet the following conditions: Core sites of the chip: The intersection of sites covered by the 450K, 850K, and 935K arrays; The base-converted probe source sequences are uniquely aligned to the end-to-end reference genome UCSC hs1; Core site of RRBS: most easily covered by sequencing. RRBS uses MspI endonuclease with a restriction site of C^CGG. Consistently methylated CpG sites: The difference in methylation levels between the positive and negative strands is minimal; The intersection of the microarray sites and the RRBS sites; Located on autosomes.

4. The DNA methylation age prediction method based on the uniAge model according to claim 1, characterized in that: Training set and test set: The training set and test set are selected according to the age distribution of all data sets, so that the training set has a certain number of samples in each age range, and the training set and test set have relatively consistent age distribution.

5. The DNA methylation age prediction method based on the uniAge model according to claim 1, characterized in that: Chip data preprocessing: only retain the methylation values ​​of the primary screening sites, 65,221 sites; Remove samples with missing age and samples with age < 1 or age > 120; Sites with a deletion ratio >10% in all samples were removed; Samples with a site deletion ratio >10% were removed; For each site, missing values ​​of methylation were filled with random normally distributed values, and the mean and variance were calculated from the non-missing values ​​of that site.

6. The DNA methylation age prediction method based on the uniAge model according to claim 1, characterized in that: Age-related CpG site screening: Age-related methylation sites were further screened from 65,221 initial screening sites. It is i Sample No. j The methylation level of each CpG site; The interquartile range of the methylation level at each site was calculated. ; Calculate the correlation between methylation at each site and age, Pearsons correlation and Spearmans correlation ; definition ; according to Sort the CpG sites from large to small, and take the 6000 CpG sites with the largest scores to enter the following model.

7. The DNA methylation age prediction method based on the uniAge model according to claim 1, characterized in that: Screening CpG sites using LASSO model The input data is 7149 chip training set samples, including variables age, data set, gender, and methylation beta values ​​of 6000 CpG sites. The LASSO model is fitted using the R package glmnet. ； Use R's machine learning framework mlr3 to call glmnet with the learner parameters lrn("regr.cv_glmnet", alpha= 1, lambda = 10^(seq(-5, 0, length.out=100)), s = 'lambda.1se', standardize= T, intercept = T, type.measure = 'mae'), or do not use mlr3 and directly use glmnet for training.

8. The DNA methylation age prediction method based on the uniAge model according to claim 1, characterized in that: Ridge regression age prediction model: Ridge regression was used to fit age at the CpG sites selected by the LASSO model. The model was: ； The learner parameters of glmnet called by mlr3 are lrn("regr.cv_glmnet", alpha = 0, lambda =10^(seq(-8, 0, length.out=100)), s = 'lambda.1se', standardize = F, intercept= T, type.measure = 'mae'), where alpha = 0 indicates fitting ridge regression; This model is denoted as full model or uniAge-605, which contains 605 CpG sites and the estimated age is ； in is the coefficient of the ridge regression estimate, It is j Methylation beta value of CpG sites, .

9. The DNA methylation age prediction method based on the uniAge model according to claim 1, characterized in that: When the model is applied to sequencing data, quantile normalization is used to correct the distribution of methylation levels of 605 sites in the sequencing data to be the same as the methylation distribution of the 605 sites in the chip data training set. This correction can slightly improve the performance of the model on sequencing data.

10. A DNA methylation age prediction system based on the uniAge model that implements the DNA methylation age prediction method based on the uniAge model as described in any one of claims 1 to 9, characterized in that: include: CpG site screening module, used to screen out CpG sites that can be covered by both RRBS and methylation chips; Age prediction model training module, used to screen sites in common CpG sites and train age prediction models using DNAme chip data of human peripheral blood; The sequencing data correction module is used to correct the DNAme levels obtained from the sequencing data, align them with the chip data, or use the screened CpG sites to fit the model coefficients separately on the sequencing data.

Images