Molecular “age clocks” promise to tell us how fast we’re aging biologically. But new research by Shyamalika Gopalan, an assistant professor in the Clemson University Department of Genetics and Biochemistry and a member of the University’s Institute for Human Genetics, along with collaborators in California and France, shows many of these clocks do not tell time well for some populations.
“None of these clocks are perfect. None of them are going to be,” Gopalan said. “But for some people, they can be way off because of genetic variation.”
Aging leaves a chemical signature on DNA in the form of methyl groups — tiny tags that get attached to specific spots along the DNA strand and help control gene activity. Over the past decade, scientists have discovered that DNA methylation patterns change in predictable ways as people get older. By analyzing these patterns, statistical models known as epigenetic clocks are able to estimate a person’s age with relatively high accuracy.
Clocks tell different times
However, the recently published work found that these clocks show differences in accuracy when applied to different populations from around the world. These differences could be explained by how these clocks were built.
Most established epigenetic clocks were trained largely on people of European ancestry and ignore genetic variants known as methylation quantitative trait loci, or meQTLs, that can push methylation levels up or down.
“Human genetics is very biased toward Western European ancestry samples. It is data from those populations that is the most widely available and mostly widely used,” Gopalan said.
In the study, the scientists analyzed saliva-derived DNA from three groups in Africa: central African Baka, southern African ‡Khomani San and southern African Himba. Each of these communities has its own complex evolutionary history and lives in distinct environments, contributing to rich variation in both genetic and epigenetic patterns, Gopalan said.
“African populations have more genetic diversity than the majority of populations that we tend to study in human genetics. Our hypothesis was that genetic variation can bias these predictive models in ways that haven’t been accounted for,” she said. “It essentially means that the same level of DNA methylation can translate to a very different age prediction depending on your genotype.”
The researchers applied eight of the most commonly used clocks to the African populations. Almost all of the clocks showed significantly higher errors in at least one of the populations compared to publicly available DNA methylation data from European and Hispanic/Latino individuals.
That’s because, depending on the specific clock, between one-fifth to nearly half of predictive sites are influenced by genetic variation. Many of these genetic variants are, in fact, only meaningfully variable in the African populations and are invariant in European populations.
When the researchers built new epigenetic clocks that deliberately avoided these genetically heritable sites, they found they could reduce error in the African cohorts while maintaining accuracy in the European and Hispanic/Latino samples.
“This study shows that we can’t necessarily take a model that was developed in one population and just apply it to another population and expect it to produce similar results,” Gopalan said. “Ideally, we would have better representation of global populations in our datasets so that we could build and train epigenetic clocks that perform better on everyone.”Detailed findings were published in the journal Communications Biology in an article titled, “Common DNA sequence variation influences epigenetic aging in African populations.”
