What is the age of your biological clock? Estimating age with a margin of error of just two years, the DNA clock delves into the black box of aging.

What is the age of your biological clock? Estimating age with a margin of error of just two years, the DNA clock delves into the black box of aging.

Beyond Birthdays, There's a "Time" Within Our Bodies

Age is determined by how many years have passed since birth. However, at the age of 60, some people can climb stairs with ease, while others struggle with multiple chronic conditions. Sleep, diet, exercise, smoking, stress, environment, illness, and genetic background—these differences accumulate over time, creating significant individual variations in the body's condition.

This is where the concept of "biological age" has gained attention in aging research. It is an attempt to quantify how much cells and tissues exhibit age-related characteristics, rather than relying solely on chronological age. One representative method is the "epigenetic clock," which reads chemical markers attached to DNA.

An international team, led by the Leibniz Institute on Aging - Fritz Lipmann Institute in Germany, has developed a new DNA clock called the "TFMethyl Clock." The aim of the research was not just to estimate age accurately. It sought to make the high accuracy of traditional clocks more explainable by understanding how the DNA locations used for predictions are connected to the biology of aging.


Reading "Post-it Notes" Instead of DNA Sequences

DNA methylation refers to the phenomenon where a small chemical group called a methyl group attaches to specific locations on DNA. It does not rewrite the DNA base sequence itself but is related to the regulation of whether genes are easily used or not. It's akin to "read" or "hold" post-it notes attached to the same blueprint.

The state of methylation is not fixed for life. It changes with aging, cell types, living environment, and some patterns correlate well with age. By inputting combinations of numerous measurement points into machine learning, one can estimate the age of a person from whom blood was drawn with considerable accuracy. This is the basic principle of the epigenetic clock.

However, a clock that accurately predicts age does not necessarily explain the mechanisms of aging. Just as AI gathers multiple small clues to approach the correct answer, even measurement points with weak age relationships and unclear biological roles can increase accuracy when combined in large numbers.

The research team experimentally demonstrated this issue. They created 100 models using 10,000 sites with weak age correlations and no overlap with transcription factor binding sites, resulting in an average median error of 3.77 years on independent data. Even features that could hardly explain the content of aging produced accuracy close to some existing clocks.

This result highlighted that the performance of "predicting age" and "explaining why a person is in that state" are separate. It's premature to think that understanding the cause of aging is achieved by only looking at high correlation coefficients.


A New Clock Created from 7,803 Samples

The study used a total of 7,803 blood samples collected from 15 datasets, with ages ranging from 0 to 101 years. Of these, 7,138 were used for model construction, and another 665 for independent validation.

The team narrowed down candidates from over 250,000 measurable CpG sites, focusing on two conditions. One was strong correlation with age, and the other was overlap with locations where transcription factors bind to DNA. Transcription factors are proteins involved in gene switch operations, and methylation changes at their binding sites are easier to biologically consider in relation to gene activity.

Candidates that passed multiple filters numbered 14,006. The final model, which combined sites showing similar changes and added measures to reduce measurement noise, consisted of 268 clusters and 548 CpG sites.

When validated with 665 independent blood samples, the correlation coefficient between estimated age and actual age was 0.97. The median error was 2.07 years, and the root mean square error was 3.47 years. Compared to existing clocks evaluated within the study, it showed favorable results not only in absolute error but also in biases common in younger and older age groups and resistance to measurement noise.

The figure of 2.07 years is indeed impressive. However, this does not mean that everyone's results fell within 2.07 years of their actual age. As a median, half had smaller errors, and the remaining half had larger ones. Also, the model primarily targeted chronological age and did not directly predict how many more years a person could live healthily.


What Emerged Were "Inflammation" and "Lipid Metabolism"

The true novelty of the TFMethyl Clock is not just in the accuracy competition. The research team extracted 661 CpG sites that were consistently selected even with repeated model construction and examined 267 related genes in the blood.

Two prominent pathways emerged as a result. One involved the production of interleukin-1β, an inflammatory cytokine, and immune responses, including NOD1, NOD2, STAT3, and TNF. The other involved pathways related to the metabolism of long-chain fatty acids, with ACSL5, ACSL6, and ELOVL2 being highlighted.

Chronic low-grade inflammation that continues with aging, known as "inflammaging," is being studied as a link to many age-related diseases such as cardiovascular disease, metabolic disorders, and neurodegeneration. Fatty acid metabolism is also deeply involved in maintaining cell membranes, energy utilization, and inflammatory responses. The fact that the new clock picked up these two suggests that age signs on DNA may not just be statistical patterns unrelated to known aging phenomena.

Among them, the site related to ELOVL2 had the most significant average impact on the model. ELOVL2 is involved in the metabolism of polyunsaturated fatty acids, and methylation around it has long been known as a strong age marker. When confirmed with other gene expression data, 200 out of 267 genes, about 75%, showed significant changes in activity with age.

The changes were not linear. Some genes showed gradual changes before age 50 but intensified after age 60, others decreased in activity between ages 40 and 60, and some suddenly dropped after age 60. Aging may not be a simple slope progressing at the same speed every year, and the periods of significant change may differ for each molecule.

However, caution is needed here as well. This analysis did not prove that changes in inflammation or lipid metabolism cause aging. It is also possible that aging caused these changes, or that another factor influenced both. Just because marks overlapped on the map does not mean the cause has been pinpointed.


Can We Conclude "Your Body Is Younger Than Your Actual Age"?

Headlines may be tempted to express "the real age of your body is revealed." However, the research results of the TFMethyl Clock cannot be directly translated into personal health assessments.

Firstly, this model is primarily trained to accurately estimate chronological age. Even if the estimated value is five years younger than the actual age, it does not provide grounds to conclude that disease risk is lower, lifespan is longer, or rejuvenation has occurred. A clock that perfectly predicts chronological age is useful in forensic science but is not necessarily the best for identifying health differences at the same age. This contradiction is discussed in aging clock research as the "biomarker paradox."

Secondly, the focus here is on blood. Methylation and gene regulation states differ in other tissues such as the brain, liver, muscles, and skin. More verification is needed to consider numbers obtained from blood as the overall age of the body.

Thirdly, the Illumina 450K array used measures only a small portion of the approximately 28 million CpG sites believed to exist in the human genome. As measurement technology advances, other important locations may be discovered. It is also necessary to verify whether the same accuracy is maintained when population groups, living backgrounds, and disease presence change, in more diverse external groups.

Moreover, it is important not to be swayed by single test results. Differences in blood collection conditions, cell composition, measurement devices, and analysis procedures can affect the numbers. The commercial "biological age tests" are not necessarily identical to the research model used in this study. The clocks used, verification targets, errors, and associations with disease or mortality differ by product.


On SNS, Focus Is on "Opening the Black Box" Rather Than "Accuracy"

Reactions on social media regarding this research are currently more centered on introductions by researchers, research institutions, and accounts dealing with longevity science, rather than large-scale discussions by general users.

 

Alena van Bömmel, one of the responsible authors, reported on LinkedIn that this was the first paper from the laboratory, emphasizing the construction of the TFMethyl Clock while examining the interaction between DNA methylation and transcription factor binding. Co-researcher Tushar Patel also introduced the incorporation of methylation's regulatory functions and the strengthening against technical noise present in real measurement data. The comments section showed a series of congratulations to the co-researchers, indicating a positive reception within the research community.

On X, an account introducing longevity research summarized that the combination of transcription factor binding site information and clustering for noise reduction demonstrated accuracy surpassing existing models. In the Japanese-speaking sphere, posts by aging researchers highlighted the construction of the clock from 548 CpG sites and the emergence of gene groups related to inflammatory cytokines and fatty acid metabolism.

The interest discernible from these posts is not just in the straightforward figure of "predicting age with a 2-year error." Experts find value in the step forward from traditional black-box models to models that can trace pathways potentially involved in aging.

However, the confirmed public reactions are skewed towards research-related individuals and specialized accounts, and it cannot be said that consumer evaluations or consensus in medical fields have been formed. The number of posts and "likes" on social media does not prove clinical effectiveness. A stance that separates topicality from scientific certainty is necessary.


A Future That This Research Might Change

The TFMethyl Clock will not immediately become a home service for aging assessment. Nonetheless, it holds great potential as a research tool.

For example, when tracking how DNA methylation changes before and after an intervention, it becomes easier to consider not just a "age score" but which regulatory regions and gene pathways might have been affected. In research on age-related diseases, narrowing down methylation sites linked to inflammation and lipid metabolism could serve as a starting point for verifying causal relationships in subsequent experiments. The research code is also publicly available, providing an environment where reproducibility can be verified with different populations and measurement data.

The future focus is whether we can advance from predicting chronological age to making meaningful future predictions for health. To what extent can the TFMethyl Clock's values distinguish disease onset, decline in physical function, and mortality risk among people of the same age? Do score changes due to lifestyle improvements or treatments truly lead to extended healthy lifespans? Can it be reproduced in tissues other than blood and in different ethnicities and regions? These need to be confirmed through long-term follow-up studies and intervention trials.

The question of "what is the true age of your body" is enticing. However, human aging is not simple enough to be expressed by a single number. The value of this achievement lies not in giving someone a verdict of "young" or "old," but in beginning to reveal the gears of inflammation, immunity, lipid metabolism, and gene regulation that move behind the hands of a highly accurate clock.

DNA clocks are not crystal balls that predict the future. However, they can serve as microscopes to measure, compare, and approach the mechanisms of the complex phenomenon of aging. This study can be seen as an achievement that has adjusted the focus of that microscope one step further.


Source URL

Related Articles