Human age prediction based on DNA methylation of non-blood tissues. (April 2019)

• Record Type: Journal Article
• Authors: Xu, Yan; Li, Xingyan; Yang, Yingxi; Li, Chunhui; Shao, Xiaojian
• Is Part Of: Computer methods and programs in biomedicine. Volume 171(2019)
• Page Start: 11
• Page End: 18
• Publication Date: 2019-04
• Subjects: DNA methylation -- Age prediction -- Non-blood tissues -- Gradient boosting regressor; Medicine -- Computer programs -- Periodicals; Biology -- Computer programs -- Periodicals; Computers -- Periodicals; Medicine -- Periodicals; Médecine -- Logiciels -- Périodiques; Biologie -- Logiciels -- Périodiques; Biology -- Computer programs; Medicine -- Computer programs; Periodicals; Electronic journals; 610.28
• DOI: 10.1016/j.cmpb.2019.02.010
• Languages: English
• ISSNs: 0169-2607

Abstract

Highlights: Proposed a regression model based on Gradient Boosting Regressor. Selected thirteen age-related CpG sites through calculating Pearson correlation between age and DNA methylation. Function introduction of selection of thirteen age-related CpG sites. Abstract: Background and Objective: The study of human aging contributes to disease prevention, treatment and life extension. Recently, epigenetics studies have evidenced that there is a close association between DNA methylation and human ages. A quantitatively statistical modeling between DNA methylation and ages could predict the person's age more accurately. Methods: We propose a regression model to predict human age based on gradient boosting regressor (GBR). We collect a total of 1280 publicly available non-blood tissues samples with ages ranged from 0 to 90 years old. We calculate the Pearson correlation between CpG's DNA methylation level and age to select age-related CpGs. Results: Thirteen age-related CpG sites are selected. GBR has the smallest mean absolute deviation to the actual age comparing with other three different models including Bayesian ridge, multiple linear regression, and support vector regression. In the training datasets, the cross-validation results show that the correlation R 2 between predicted age and DNA methylation is 0.89, and the mean absolute deviation is 4.66 years. In an independent testing set with 262 samples, the GBR achieves the mean absolute deviation of 6.08 years. Meanwhile we also briefly describe the function of the selected thirteen CpG sites. Conclusions: We build an age predictor to study the association between ages and the DNA methylation of human non-blood tissues. Our new model provides a more accurate estimation of human ages which will be instrumental for understanding the regulation of DNA methylation on human aging and will accurately monitor the individual aging process.