Epigenetic Clocks Power Biological Age Data, But Why Do Results Differ Across Tests?

In a world increasingly focused on longevity, many companies now offer tests for biological age: the physiological condition of a person’s organs, cells, and tissues.¹ The most established methods for estimating biological age use epigenetic clocks, which use machine learning algorithms to analyze genome-wide DNA methylation data. However, while the science of epigenetic clocks is robust, a lack of standardized methods in the industry means the results for a sample can vary significantly. So, what does this mean for those who are keen to find out their biological age?

“There's no question that epigenetic clocks perform very well in epidemiological studies,” said Steve Horvath, a pioneer in epigenetic clock research and principal investigator at longevity startup Altos Labs, in a panel discussion on aging biomarkers at the Precision Medicine World Conference 2026. “But what the field has not yet achieved is really what one would call reference standards. It's not yet possible that you can go to different validated labs and get the same readout.”

Different Epigenetic Clocks, Different Age Results

Horvath has developed and validated several commonly used epigenetic clocks, including the Horvath clock, GrimAge, PhenoAge, and Dunedin PACE.^2–5 First-generation epigenetic clocks, such as the Horvath clock, were trained by comparing the DNA methylation signatures of young and old individuals, using this data to make predictions about chronological age.⁶ Second-generation clocks, including PhenoAge and GrimAge, built on these principles to model health phenotypes and mortality risk.⁶

The most recent iterations, such as Dunedin PACE, were trained on the differences in multi-organ system deterioration between individuals of the same age. These third-generation clocks have the potential to predict complex clinical outcomes, but incorporating broader biological information along with DNA methylation data means interpreting their results is even more complicated.⁶

Almost all epigenetic clocks used in commercial age prediction tests are trained on blood samples, which is the gold standard. However, some companies currently offer testing based on saliva—something Horvath said is an underappreciated issue.

“Blood is composed entirely of immune cells (leukocytes), with well-characterized methylation profiles,” Horvath explained in a follow-up interview. “Saliva, by contrast, is a mixture of approximately 65 percent immune cells and 35 percent buccal epithelial cells. Those two cell types have fundamentally different methylation signatures. When you apply a blood-trained clock to saliva without correction, the errors can be substantial.”

Epigenetic Clocks Differ in Measuring DNA Methylation

While Horvath said that researchers are aware of this potential pitfall, most companies offering biological age test kits generally don’t detail which clock they use, making it difficult to determine if the samples they are collecting will provide consistent results. Importantly, he said, it is blood samples that have provided evidence that epigenetic aging clocks like GrimAge can predict clinical outcomes, such as disease and mortality, in large-scale epidemiological studies. “Without that linkage, you cannot validate whether a saliva-based clock actually predicts anything meaningful,” he added.

How DNA methylation is measured can also affect the results; some groups use the high-throughput, lower-cost methylation array platform offered by Illumina, while others perform higher-resolution whole-genome bisulfite sequencing. “The Illumina array measurements are correlated with sequencing, but fundamentally, it's a different way of measuring,” said Horvath.

Many companies offer customer-specific plans that they claim can help reverse their biological age should it be higher than expected, yet most of the advice in these plans is lifestyle-based; dietary changes, avoiding smoking and alcohol, getting adequate sleep, and stress management are common recommendations. Other companies sell supplements, but their efficacy in reducing biological age is unclear.

Regardless of the sample type, how DNA methylation and aging is measured, or which clock is used, the fact remains that these tools were designed for epidemiological studies, not individuals. “The most direct use of methylation clocks is really for patient stratification purposes,” Horvath remarked.

According to panel member Mahdi Moqri of Harvard Medical School, other types of clocks may be more suitable. “A lot of groups are now working with [blood plasma], and they are using, for example, proteomics data,” Moqri explained. “It's a very powerful modality that captures residual proteins from many different tissues, and then we benchmark them against [epigenetic] clocks. So we know in some specific contexts, specific diseases, they outperform epigenetic clocks.” Combining these different modalities, while challenging, could provide the most comprehensive picture of biological age and disease risks, Moqri added.

Continue reading below...

Article

Using Site-Saturation Variant Libraries to Understand Viral Evolution

Discover how scientists bridge the gap between genomic detection and functional assessment to remain a step ahead of viral evolution.

Read More

Organ-Specific Aging Tests and Rejuvenation Therapy

In addition to overall biological age tests, several companies now offer tests that claim to measure the age of specific organs based on epigenetic clocks. According to Horvath, two approaches currently exist for organ-specific aging data. The first uses cell-free DNA released from organs that circulates in the peripheral blood; by detecting which organs are shedding more DNA, researchers can infer organ-specific damage.

The second, more established approach uses machine learning to analyze DNA methylation from blood, building organ-specific aging scores and predicting relevant diseases.⁷ While both methods are based on solid science, Horvath said organ-specific epigenetic age tests are currently transitioning from proof-of-concept to early validation. “I think the science behind these tests is real but still young,” he continued. “Comprehensive validation across diverse populations, with long-term outcomes, is still ongoing… Time will tell whether they can serve as diagnostic instruments.”

In his work at Altos Labs, Horvath is using artificial intelligence to further refine epigenetic clocks and is contributing to the development of rejuvenation-based therapies that can improve healthspan. Despite excitement in the field, he explained that a key challenge is to validate these interventions in the eyes of regulators like the Food and Drug Administration, using clocks as surrogate endpoints.

“We need biomarkers that detect a beneficial effect of a rejuvenation intervention within a year or two, or we need biomarkers that we apply to an animal model, a mouse, a rat, and then we have a pretty good chance that these readouts are actually informative for human drive that we can use in IND studies,” Horvath said. “…that's an important challenge. And I would say we're not quite there yet; that's the frontier of work.”