An Age Prediction Model Was Established Based On DNA Methylation Status Of Hair

Objective:This study aims to find the age-related DNA methylation sites in hair in northern Chinese Han population and construct an age prediction model.Methods:Based on the recent research results and literatures,a total of 21 methylation candidate sites that may be multi-tissue age prediction were screened.The candidate sites were further screened in a pretest by comparing 6 samples from a younger group and an older group based on SNaPshot technology.We further screened 10 CpG sites with age-related and normal shape profiles to construct a multiplex methylation SNaPshot assay.Species specificity and sensitivity of the assay were tested.The DNA methylation levels of130 hair follicle samples from northern Chinese Han were detected for 10 DNA methylation sites using this assay.Pearson correlation analysis was performed to evaluate the correlation between age and the methylation status of each CpG site.And we constructed four age prediction models,including multiple linear regression,backward stepwise regression,multi-layer perceptron and radial basis function.To determine whether sex affected age prediction precision,a Mann-Whitney U test was performed for each CpG site on all male and female samples.In order to determine whether color can affect age prediction,the methylation status of black and white hair of 6 volunteers was measured and analyzed by Mann-Whitney U test.To assess whether hair from different body parts would impact DNA methylation status,the age prediction from the scalp,calf,pubic area,and armpit of 6 volunteers were examined.The results were statistically analyzed by using Kruskal-Wallis H test.Results:In this study,10 DNA methylation sites with high correlation with age were selected to construct SNaPshot system.The 10 sites of CPG1-10 were CG01820374(LAG3 gene).Cg06493994(SCGN gene);Chr6:11044628(ELOVL2 gene);Cg14361627(KLF14 gene);Chr1:207823681(C1orf132 gene);Cg07547549(SLC12A5 gene);Cg24724428(ELOVL2gene);Cg25148589(GRIA2 gene);Chr7:130734357(KLF14 gene);Cg17861230(PDE4C gene),respectively.Groups were assigned according to the donor’s sex and age(1–19,20–39,40–59,and ≥60 years).According to the ratio of 7:3,130 samples were randomly stratified sampling and divided into the training group and the test group.Four methods of multiple linear regression,stepwise regression,multilayer perceptron and radial basis function were used to construct the age prediction model from the data of 90 training samples.The predictive power of the above 4 models was further verified in the 40 independent testing samples.The result showed that the multiple linear model based on 10 CpG sites had the best prediction performance.R2 was 0.917.Mean absolute deviation(MAD)of the training group was 3.68 years old,and MAD of the test group was 5.06 years old.The prediction accuracy of this model decreased in the groups with advanced age.The group of the youngest group(1-19 years)exhibited high age prediction accuracy,with a MAD of 3.25 years,while the group of the oldest group(≥ 60 years)had a MAD of4.68 years.In this study,CpG site methylation status and age prediction precision were not affected by sex,hair color or hair type.When our model is applied to predict age,these factors are not taken into account.ConclusionIn this study,we studied the methylation status of 10 CpG sites in hair of northern Chinese Han based on SNaPshot assay,and constructed the age prediction model.The age prediction model is not affected by sex,hair color or hair type.It can be used to infer the age of criminal cases,and narrow the scope of investigation.It has practical application value.In order to promote the application of the age prediction model in practice,we should further study the samples from more different ethnic groups or districts to improve the prediction performance of the model.