DNA changes human age with significant accuracy

Time postmarks itself in our cells through different DNA changes patterns, allowing scientists to use only a small blood sample to determine a person’s age with unprecedented accuracy.

Researchers at the Hebrew University of Jerusalem have developed a groundbreaking method to determine a person’s age with a median error of only 1.36 years of age for those under 50. Their technology, called Magenet, uses AI to analyze how AI analyzes DNA methylation (how it is added to our genetic material), adding predictable patterns to our age.

“It turns out that the passage of time leaves measurable markers on our DNA,” explained Professor Tommy Kaplan, one of the senior authors of the study. “Our model decodes these markers with amazing accuracy.”

Reading time fingerprints at molecular level

The research team led by Bracha Ochana and Daniel Nudelman found that by focusing only on two specific regions of the genome, they were more accurate than previous methods. What sets their approach apart is analyzing methylation patterns at the level of individual DNA molecules, rather than average in many cells.

Using deep sulfate sequencing (which reveals the methylation state of multiple adjacent DNA sites), the team discovered two different ways in which our cells record time passes:

• Certain DNA regions accumulate methylation at each location and vary randomly and independently, such as individual particles falling into a sand grain of an hourglass.
• Other regions change in coordinated “blocks”, multiple adjacent sites switch their methylation state together
• These patterns produce molecular timestamps, which become more obvious every year

“We found that some age-related methylation changes may stem from cell neutral changes within a specific cell type, while others reflect changes in blood cell composition over time,” the research team noted. This finding helps explain how our bodies track the passage of time at the cell level.

Precision than all previous methods

The researchers tested blood samples from more than 300 healthy individuals aged 17-78 to achieve extraordinary accuracy. For individuals under 35 years of age, the median error is only 0.9 years, which is more accurate than any previous epigenetic clock.

What is particularly striking is the consistency of the method. Unlike other biological markers that may be affected by lifestyle or environmental impacts, Magenet’s age prediction remains stable regardless of factors such as smoking status, body mass index, or gender. The researchers confirmed this by analyzing samples from the Jerusalem Perinatal Study, which tracked individuals within 10 years.

When the team examined how their predictions changed over time, they found that if a person’s predicted age deviated from actual age in the first measurement, then ten years later, the same deviation persisted. This suggests that early life events may permanently change a person’s “epigenic age” and then faithfully record the passage of time.

Application from evidence collection to medical research

Perhaps most striking is that the method requires very little DNA. The researchers found that they could accurately predict 50 DNA molecules of age, which is equal to the genetic material of only a few cells. This is of great significance to forensic science, where researchers usually only get trace amounts of biological materials.

“The ability to determine age from a small sample like this can change how we conduct criminal investigations,” said Professor Ruth Shemer, another senior author of the study.

This technology also has potential applications in aging research and personalized medicine. By distinguishing between age (time of birth) and biological age (physiological state of the human body), doctors can tailor treatments based on individuals’ unique aging characteristics.

A new understanding of our age

In addition to its practical application, research provides basic insights into how our bodies measure time. The team found that age-dependent methylation changes occur randomly or in a coordinated block-like manner in clusters across CpG sites.

This mechanism seems to be a common feature of human aging, as researchers have successfully applied their model to urine samples (median error of 2.45 years), and to a lesser extent saliva samples.

The study, published in the Cell Report, represents our understanding of the biology of aging and provides powerful new tools for research and practical applications. As Professor Yuval Dor summed up: “This is a powerful example of what happens when biology encounters AI.”

Related