Genetic Prediction of Biological Age: A Meta-Analysis on Telomere Length, DNA Methylation, and All-Cause Mortality

Aging is a biological process during lifespan with accumulation
of mutations and damages, lowering fitness of at older ages and
increasing hazards to survival...


Introduction
Aging is a biological process during lifespan with accumulation of mutations and damages, lowering fitness of at older ages and increasing hazards to survival [1]. Aging is the most important risk factor associated with many diseases, such as cardiovascular disease, cancer, type 2 diabetes, hypertension, and Alzheimer's disease [2]. It accounts for about two thirds of death worldwide, and an even higher rate of 90% in developed countries.
Understanding the biological mechanism of aging will therefore lead to tremendous public health benefits. Aging process can be affected by both genetic and nongenetic factors. The nongenetic intervention on aging can be long-lasting, and potentially explained by epigenetic mechanisms. Epigenetics, which refers to changes to the genome that do not involve changes in DNA sequence, have gain popularity in recent research [3]. These epigenetic markers, e.g., telomere length and DNA methylation, can lead to a prediction of biological age or acceleration of aging, and age-related diseases [4]. Telomeres are repetitive noncoding DNA components located at the end of chromosomes to protect from degradation of coding sequences.
The telomeres shorten each time a cell divides because of the end replication problem, but also by oxidative stress, and lengthened by the enzyme telomerase and DNA exchange during mitosis [5,6].
Telomere attrition has been widely reported to be associated with increased morbidity and mortality of various age-related diseases [7]. DNA methylation is a process by which methyl groups are added to the DNA molecule. Methylation can change the activity of a DNA segment without changing the sequence [8,9]. Recently developed indices of cellular age based on DNA methylation data are being used to study factors that influence the rate of aging and the health correlates of these metrics of the epigenetic clock [10].
of association strength between telomere length and all-cause mortality. A similar search was conducted on the PubMed database to search for publications on DNA methylation markers and allcause mortality, using the keywords "epigenetic" and "all-cause mortality". A recent meta-analysis was found using 12 cohorts in a collaborative approach [17], where the involved research groups agreed to perform association tests between DNA methylation age and all-cause mortality using consistent modeling and share the results, either negative or positive. Three additional studies were carried out separately. Results from these 15 cohorts was combined using both fixed-and random-effect meta-analysis.
Multiple formulas have been developed, based on different sets of methylation markers, to calculate "DNA methylation age". It is typically compared to chronological age to obtain a measure of age acceleration. A formula derived by Horvath [18] was selected for the meta-analysis as much as possible (14 out of 15 cohorts).

Data
We introduce telomere length data and DNA methylation data that are used in the study, respectively.

Telomore Length Data
A total of 27 studies were included in the meta-analysis that reported association results between telomere length and all-cause mortality. Summary characteristics of individual studies were shown in Table 1. Individual association results were included in Figure 1, except Igari, et al. [16], which only reported association p-value instead of estimated effect size (hazard ratio, HR). When effect sizes (HRs) and their 95% confidence intervals (CIs) were available, we first converted the HR and CI into natural log, and then calculated z-score as log(HR)/SE(HR), where the standard error (SE) was estimated from log-transformed CI. The p-value reported in Igari, et al. [16] was converted into z-score from the inversequantile function of standard normal distribution.

DNA Methylation Data
A total of 15 studies were included in the meta-analysis that reported association results between DNA methylation and allcause mortality. Summary characteristics of individual studies were shown in Table 2. There were a total of 16,939 participants with 3,634 deaths. Most of the studies were on whites, except two on blacks and one on Hispanic. Results from these 15 cohorts was combined using both fixed-and random-effect meta-analysis.

Multiple formulas have been developed, based on different sets
of methylation markers, to calculate "DNA methylation age". It is typically compared to chronological age to obtain a measure of age acceleration. A formula derived by Horvath [18] was selected for the meta-analysis as much as possible (14 out of 15 cohorts).

Results
We report telomere length results and DNA methylation results, respectively.

Telomore Length Results
Twenty-six of the 27 studies reported estimated effect size as odds ratio or hazard ratio between reduced telomere length and all-cause mortality. Given heterogeneity among these studies, a pooled estimate did not have a direct interpretation. A forest plot ( Figure 1) was generated to show individual association results, but only for demonstration purpose. A weighted z-score method was used for meta-analysis. A significant association was observed between telomere length and all-cause mortality (combined z = 7.49, p = 6.75E-14). We further restricted analysis into different ethnic groups.
There were 23 studies conducted in European countries or United States. The combined z-score was 6.85 (p = 7.34E-12). Another 3 studies were based on Eastern Asian population (specifically, China and Japan). Although the sample sizes were much smaller in these 3 studies compared to many Europe-or US-based studies, their results still yielded a combined z-score of 3.56 (p-value = 0.0004).
Therefore, the telomere length was significantly associated with all-cause mortality in both of the two ethnic groups. Although not appropriate due to the inconsistent effect size estimates, we still applied a funnel analysis to diagnosis potential publication bias, in which negative/insignificant findings would be less likely to publish. The funnel plot ( Figure 2) did not suggest any obvious publication bias, with most of the studies being inside of the confidence interval (the "funnel").

Conclusion and Discussion
In this project, we carried out an extensive literature review on epigenetic markers and ageing. Using a meta-analysis approach, we observed significant association between telomere length, a al. [19] showed significant association between DNA methylation age and all-cause mortality after controlling for telomere length.
Because DNA methylation is prevalent on the DNA sequence with millions of methylation markers, it potentially provides a stronger prediction tool than telomere length for aging. DNA methylation data are also easily accessible through public databases, allowing more complicated modeling approaches than meta-analysis. My next step will focus on DNA methylation on its relationship with the ageing process.

Future Work
In this preliminary analysis, we confirmed association between telomere length and all-cause mortality. It suggests the role of epigenetic markers in the ageing process. In future study, we will explore other epigenetic markers, e.g., DNA methylation, and develop an efficient prediction model for ageing and age acceleration using data mining tools.

Data Availability Statement
The data presented in this study are searched from https:// www.ncbi.nlm.nih.gov/pubmed and is available upon request to the author.