CheekAge: Next-gen epigenetic clock accurately predicts mortality risk

Using the next-generation CheekAge clock, scientists can now accurately predict mortality by analyzing cheek cells, offering new insights into aging and health risk assessment. This breakthrough could reshape the future of personalized health monitoring.

Study: CheekAge, a next-generation epigenetic buccal clock, is predictive of mortality in human blood. Image Credit: Shutterstock AI Generator / Shutterstock.com

In a recent study published in Frontiers in Aging, researchers utilize CheekAge, a next-generation epigenetic clock, to predict mortality risk.

About epigenetic clocks

An epigenetic aging clock is a machine-learning model that predicts chronological age based on DNA methylation. DNA methylation primarily occurs on dinucleotides formed by cytosines followed by guanine residues (CpG).

Epigenetic clocks can be trained to determine chronological age or predict health outcomes at the population level. First-generation clocks are often used in forensic investigations, as they have been trained to predict chronological age.

Comparatively, next-generation clocks utilize methylation data to provide important insights into health and lifestyle outcomes. However, these models require blood samples, which limits their utility for home use or in older adults.

What is CheekAge?

CheekAge was trained on the Infinium MethylationEPIC buccal dataset, which comprises over 200,000 DNA methylation sites from more than 8,000 adult buccal cell samples. CheekAge was developed to estimate epigenetic age and identify any associations that may exist between lifestyle and health/disease factors like weekly exercise, sleep quality, stress, and diet.

Previously, the researchers of the current study reported higher CheekAge in patients with conditions like meningioma or progeria, respiratory infections, and a history of childhood cancer that required radiation therapy. The present study seeks to examine the accuracy of CheekAge in predicting the risk of death since this is required of a high-quality aging biomarker.

About the study

The researchers used CheekAge to predict mortality in the Lothian Birth Cohorts of 1921 and 1936. These cohorts were initially established to monitor cognitive and brain aging and document lifestyle and psychosocial factors, in addition to biomedical, genetic, epigenetic, and neuroimaging data on all study participants.

The current study included 1,513 participants with methylation data. The chronological age ranged from 67.8 to 90.6 years.

Mortality data was obtained from central health registers and converted to age in days at the time of death, which formed the outcome measure. The last methylation data was used to predict CheekAge.

Initially, the researchers applied the blood methylation-based Infinium HumanMethylation450 CpGs to the original Infinium MethylationEPIC array buccal data on which CheekAge was trained. However, approximately 50% of the CpG inputs were missing due to differences between the datasets. Nevertheless, the researchers obtained results that were similar to those obtained with the full CheekAge model with all available CpGs.

Thereafter, Infinium HumanMethylation450 CpGs was applied to Lothian data. The standard delta age was then determined by dividing the delta age, which is calculated by subtracting chronological age from epigenetic age, by the standard deviation of all delta ages. This showed a significant correlation with mortality.

The study findings reflected a 21% increase in mortality for every standard deviation unit change. Survival curves predicted that 50% of individuals with the highest delta CheekAge would have died 7.8 years earlier than those with the lowest CheekAge.

This is the first time that a biomarker for aging developed for buccal tissue has predicted mortality from blood methylation data.

Comparison with other clocks

CheekAge consistently outperformed all first-generation clocks simultaneously used on the cohort data. In fact, the CheekAge outputs were comparable to the next-generation DNArn PhenoAge clock, which indicated a 23% increase in mortality for each unit standard deviation change. Importantly, DNArn PhenoAge was trained on blood samples rather than buccal samples used for CheekAge.

Mortality CpGs

Delta age was also re-examined in relation to each of the CpGs. With each iterative removal of a CpG, the significance of the mortality association was calculated. This led to the identification of ‘mortality and anti-mortality CpGs,’ defined as CpGs that increase and reduce the significance of the predicted mortality risk.

The largest effect was observed with the removal of the mortality CpG cg14386193, which is annotated to the gene ALPK2. This led to an almost three-fold rise in the false discovery rate (FDR) value. Other CpGs, like cg00991744 and cg20210051, both of which are annotated to PDZRN4, a possible tumor suppressor gene, and cg00664454, which is annotated to CPNE2, were also predictive of mortality risk.

These mortality CpGs have been implicated in aging or age-related diseases, such as survival, cancer, osteoporosis, or metabolic syndrome. Nevertheless, additional research is needed to confirm the transcriptional effects of the mortality CpGs. 

Enrichment analyses of the annotated CpGs with mortality associations were also performed to elucidate their biological roles. To this end, several genes associated with anti-mortality CpGs were also associated with developmental processes, morphogenesis, and proteostasis.

Conclusions

Taken together, we provide important validation for CheekAge and highlight novel CpGs that underlie a newly identified mortality association.”

Despite the limited availability of CpG inputs that were obtained from a different tissue, CheekAge identified significant associations with mortality. The longitudinal aspect of the dataset ensures that this model outperformed first-generation clocks trained on blood-derived datasets.

Most aging-related methylations are tissue-specific; therefore, a longitudinal follow-up study is needed to clarify which CpGs have the strongest associations with mortality.