An age classification model based on DNA methylation biomarkers of aging in human peripheral blood using random forest and artificial neural network
Ye Li,Yukun Yang,Jie He,Jinxiu Hu,Xiangqing Zhu,Chuan Tian,Mengdie Chen,Xiaojuan Zhao,Li Ye,Hang Pan,Xinghua Pan,Ye Li,Yukun Yang,Jie He,Jinxiu Hu,Xiangqing Zhu,Chuan Tian,Mengdie Chen,Li Ye,Hang Pan
DOI: https://doi.org/10.14715/cmb/2024.70.1.21
2024-02-21
Cellular and Molecular Biology
Abstract:Recent epigenetic studies have revealed a strong association between DNA methylation and aging and lifespan, which changes (increases or decreases) with age. Based on these, the construction of age prediction models associated with DNA methylation levels can be used to infer biological ages closer to the functional state of the organism. We downloaded methylation data from the Gene Expression Omnibus (GEO) public database for normal peripheral blood samples from people of different ages. We grouped the samples according to age (18-35 years and >50 years), screened the methylation sites that differed between the two groups, identified 44 differentially methylated sites, and subsequently obtained 11 age-related characteristic methylation sites using the random forest method. Then, we constructed an age classification model with these 11 characteristic methylation sites using an artificial neural network and evaluated its efficacy. The age classification model was constructed by an artificial neural network and its efficacy was evaluated. The model predicted an area under the curve (AUC) of 0.97 in the validation set and accurately distinguished between those aged 18-35 and >50 years. Furthermore, the levels of these 11 characteristic methylation sites also differed significantly between the two sets of samples in the validation set, including six newly identified age-related methylation sites ( P <0.001). Finally, we constructed a multifactor regulatory network based on the corresponding genes of age-related methylation sites to reveal the transcriptional and post-transcriptional regulation patterns. As a result of the increasing problem of aging, the age classification model we constructed allows us to accurately distinguish different age groups at the molecular level, which will be more predictive than chronological age for assessing individual aging and future health status.
cell biology,biochemistry & molecular biology