Date Published: May 12, 2018
Publisher: Impact Journals
Author(s): Chunxiao Li, Wenjing Gao, Ying Gao, Canqing Yu, Jun Lv, Ruoran Lv, Jiali Duan, Ying Sun, Xianghui Guo, Weihua Cao, Liming Li.
The DNA methylation age, a good reflection of human aging process, has been used to predict chronological age of adults and newborns. However, the prediction model for children and adolescents was absent. In this study, we aimed to generate a prediction model of chronological age for children and adolescents aged 6-17 years by using age-specific DNA methylation patterns from 180 Chinese twin individuals. We identified 6,350 age-related CpGs from the epigenome-wide association analysis (N=179). 116 known age-related sites in children were confirmed. 83 novel CpGs were selected as predictors from all age-related loci by elastic net regression and they could accurately predict the chronological age of the pediatric population, with a correlation of 0.99 and the error of 0.23 years in the training dataset (N=90). The predictive accuracy in the testing dataset (N=89) was high (correlation=0.93, error=0.62 years). Among the 83 predictors, 49 sites were novel probes not existing on the Illumina 450K BeadChip. The top two predictors of age were on the PRKCB and REG4 genes, which are associated with diabetes and cancer, respectively. Our results suggest that the chronological age can be accurately predicted among children and adolescents aged 6-17 years by 83 newly identified CpG sites.
Epigenetics refers to the molecular mechanisms regulating gene expression without changing the DNA sequence . The mostly studied epigenetic marker is DNA methylation, the presence of methyl groups at CpG dinucleotides . Previous evidence suggested that global levels of DNA methylation increased over the first few years of life  and then decreased in late adulthood [4,5], suggesting that epigenetic modifications might play a vital role in the human’s aging process [6,7].
The basic characteristics of the participants are shown in Table 1. In the present study, 179 samples and 817,471 CpG sites passed quality control (QC) in the training and testing dataset (N=90 and N=89 respectively). In total, the study consisted of 101 male and 78 female singletons with an age range from 6 to 17 years (mean 10.7). The quality control results are provided in Figure S1 and Table S2 in supplements.
In this study, we identified 6,350 age-related CpG sites from the EWAS among 817,471 QC probes in 179 children (aged 6 to 17 years). In the training dataset, we finally selected 83 novel CpG sites predictive of age from all those age-related CpG sites by elastic net regression. Chronological age of the pediatric population could be accurately predicted by the DNA methylation values of the 83 CpG sites, which provided an accurate prediction of age with a correlation of 0.99 and an error of 0.23 years for the training dataset, with a robust correlation of 0.93 and an error of 0.62 years in the testing dataset.