Models for Evaluation of Biological Age Based on Multi-omics

Abstract

Aging is a complex process of systemic degeneration at multiple cellular and tissue levels that has a complex mechanism. Individuals age at different rates, and there is a high degree of aging heterogeneity. Therefore, it is inaccurate to judge an individuals degree of aging by their chronological age. We performed transcriptome-focused multi-omics analyses (including transcriptomics, DNA methylation, proteomics, cytokine and telomere analysis, and single-cell transcriptome sequencing) on 139 healthy individuals aged 23 to 88 years. We systematically analyzed linear and non-linear changes in gene expression throughout aging. Genes with similar expression trajectories were enriched in similar biological pathways, including the PI3K-AKT signaling pathway and the inflammatory response. Through DNA methylation detection, we found that the expression of the top genes correlated with age was affected by methylation in the gene promoter region. These genes had no significant correlation with the expression of downstream proteins, but they were enriched in PI3K-AKT-associated proteins. Single-cell transcriptome sequencing showed that the expression of these genes did not significantly change in different cell subtypes, which proves that the gene expression changes were caused by internal age-related cellular changes rather than cell composition changes. We designed a transcriptome age clock and a methylation age clock using a set of 787 genes. Our models can accurately predict age with a mean absolute error (MAE) of 5.203 and 3.28, respectively, which is better than previously established aging models. The accuracy of our model was further verified by the detection of telomeres, which can identify accelerated aging of individuals.