Omics data provides a wealth of metrics that correlate with age, quite well in some cases. Weighted combinations of CpG site methylation status, protein levels, and RNA transcript levels have all been found to measure age, and new and improved versions of these aging clocks are introduced on a regular basis. For people with a greater burden of damage and age-related disease, measured age tends to be greater than chronological age. So there is some hope that these approaches are actually measuring biological age, and can thus be used to speed up the development of rejuvenation therapies. Unfortunately it is unclear as to how the measurements made by aging clocks connect to the underlying damage of aging. Perhaps they reflect all of it, but perhaps not. Thus at the present time any given clock must still be calibrated for each specific approach to rejuvenation before the results can be taken at face value.
Increasing evidence has pointed to the interactions between genetics, epigenetics, and environmental factors in the aging process. Over the last decade, there has been a growing body of research in identifying genetic and epigenetic biomarkers of aging to decipher the molecular mechanisms underpinning disease susceptibility. For example, the genome-wide association studies (GWAS) have identified genetic loci associated with longevity and several aging-related diseases. As aging is a multifactorial process determined by the dynamic nature of static genetics as well as stochastic epigenetic variation and transcriptomics regulation, both DNA methylation and gene expression have emerged as promising hallmark for understanding the aging process and its associated diseases.
Numerous estimators have been developed to predict human aging from DNA methylation data. While the first generation DNA methylation age estimators including Horvath’s clock and Hannum’s clock were developed based on chronological age, the second generation DNA methylation age estimators were obtained by optimizing the prediction error on phenotypic age derived from clinical attributes associated with mortality and morbidity. This includes PhenoAge and GrimAge which aim to improve prediction of aging related outcomes (e.g., time-to-death, time-to-disease for cancer, Alzheimer’s disease, and cardiovascular disease).
In addition to DNA methylation, changes in gene expression have been shown to be associated with aging and aging-related outcomes. Specifically, 56 consistently over-expressed and 17 genes consistently under-expressed with chronological age were identified by performing a meta-analysis on 27 microarray datasets from mice, rats, and human subjects. A closely related work was the development of the GenAge database of aging-related genes, including 307 genes potentially related to human aging.
Unlike DNA methylation in which several user-friendly software and computer programs are available for predicting epigenetic age across different tissues, there were limited transcriptional age predictors and the existing predictors have several pitfalls. First, most of the human transcriptional age predictors were developed based on microarray data and/or limited to only a few tissues. Second, the only predictor constructed using RNA-Seq data was derived based only on fibroblast data. To date, transcriptional studies on aging using RNA-Seq data across different human tissues was limited. Recognizing the gap in existing research of transcriptional aging based on RNA-Seq data, the aim of this study was twofold, first to identify common age-related genes across tissues; second to construct tissue-specific transcriptional age calculators for understanding how gene expression changed with age in different human tissues.
Based on our results, we introduce RNAAgeCalc, a versatile across-tissue and tissue-specific transcriptional age calculator. By performing a meta-analysis of transcriptional age signature across multi-tissues using the GTEx database, we identify 1,616 common age-related genes, as well as tissue-specific age-related genes. Based on these genes, we develop new across-tissue and tissue-specific age predictors. We show that our transcriptional age calculator outperforms other prior age related gene signatures as indicated by the higher correlation with chronological age as well as lower median and median error. Our results also indicate that both racial and tissue differences are associated with transcriptional age. Furthermore, we demonstrate that the transcriptional age acceleration computed from our within-tissue predictor is significantly correlated with mutation burden, mortality risk, and cancer stage in several types of cancer from the TCGA database, and offers complementary information to DNA methylation age.