The Denominator We Ignored
Some cells age faster than others. Not by years or months. By methylation accumulation at polycomb CpG islands, measured in single cells, not averaged across millions, according to research on single-cell methylation patterns.
DNA methylation clocks are the standard tool in aging research. They measure the covalent addition of methyl groups to DNA bases, typically cytosine of CpG dinucleotides, according to foundational studies on DNA modification. The clocks predict biological age. They correlate with disease risk. They are built on bulk tissue samples.
Bulk means millions of cells homogenized into one number. The clock measures an average. The average hides the distribution.
What Single Cells Show
Methylation at polycomb CpG islands accumulates unevenly across single cells during aging, according to single-cell methylome data. Some cells carry high methylation loads. Others stay relatively clean. The variation exists within the same tissue, in the same person, at the same moment.
No one measured this before because no one looked at single cells. A DNA methylation atlas of normal human cell types exists at base pair resolution, according to recent mapping efforts. But cellular heterogeneity impacts efforts to accurately call differentially methylated regions, according to researchers studying methylation variation. The atlas shows cell types. It does not show the variation within each type.
The difference matters. Every clock built on bulk tissue treats methylation as a single number per sample. The single-cell data shows that number is an average of a distribution we have not measured.
The Stem Cell Explanation
A stem cell dynamics model explains how DNA methylation changes arise and propagate with age across mammals, according to research on aging mechanisms. Stem cells divide asymmetrically. One daughter cell stays in the stem pool. One differentiates into tissue.
DNA methylation is catalyzed by methyltransferase enzymes using S-adenosyl methionine as a donor, according to biochemical studies. The process can lead to mitotic propagation of the modified sequence, according to research on epigenetic inheritance. Methylation errors accumulate in the stem lineage. They propagate to daughter cells. The variation in single cells reflects stem cell division history.
Stem cell dynamics are identified as a key driver of aging across mammals, according to the dynamics model. The cells that divide more accumulate more errors. The cells that divide less stay cleaner. The tissue average depends on which cells you sample and when.
What This Breaks
DNA methylation-based biomarkers can be used for cancer detection and prognosis, according to clinical studies. The biomarkers are built on bulk tissue averages. If methylation accumulates unevenly, the average depends on which cells dominate the sample. A tumor with heterogeneous methylation could yield different biomarker values depending on biopsy location.
Cell-free DNA methylome analysis has been applied for early preeclampsia prediction, according to prenatal diagnostic research. Cell-free DNA comes from dying cells. If some cells age faster than others, the methylation signal in cell-free DNA depends on which cells are dying. The prediction assumes a homogeneous signal. For pregnant women screened using these tests, the difference between accurate and inaccurate prediction can determine whether high-risk cases receive early intervention or progress undetected to dangerous complications.
DNA methylation patterns are associated with anxiety and may affect stress and immune responses, according to studies on psychological stress. DNA methylation patterns may be linked to biological aging in the context of anxiety, according to research connecting mental health and epigenetics. Both findings are based on bulk tissue measurements. Both assume the average represents the population.
The Measurement Problem
DNA methylation affects the binding of regulatory proteins such as transcription factors, according to molecular studies. A cell with high methylation at polycomb islands behaves differently than a cell with low methylation. If half the cells in a tissue are highly methylated and half are clean, the bulk measurement shows moderate methylation. The tissue contains two populations. The measurement shows one average.
Maternal smoking in pregnancy is associated with DNA methylation changes in newborns, according to epidemiological research. In utero exposure to the Great Depression is reflected in late-life epigenetic aging signatures, according to historical cohort studies. Both findings measure methylation in bulk samples. Neither measures the distribution within individuals. For the children born to mothers who smoked during pregnancy, the bulk measurements show methylation changes but cannot reveal whether all cells are mildly affected or whether a subset carries the entire burden of epigenetic damage.
The question is not whether the associations are real. The question is whether the mechanism operates at the population level or the single-cell level. If methylation accumulates unevenly, the cells with the highest loads may drive the phenotype. The bulk measurement does not tell you how many such cells exist or where they are.
Who Decides What Gets Measured
Single-cell methylation sequencing can resolve the distribution. The technology exists. The barrier is not capability but adoption, and adoption follows funding.
Research funding agencies prioritize projects based on peer review panels that evaluate feasibility, cost, and expected impact. Single-cell methylation studies cost roughly ten to fifty times more per sample than bulk tissue analysis, depending on sequencing depth and cell number. A typical bulk methylation study might sequence 500 samples for $50,000. The same study at single-cell resolution could require $500,000 to $2.5 million.
Grant review panels weigh this cost against incremental knowledge gain. If bulk measurements already correlate with disease outcomes, reviewers question whether single-cell resolution justifies the expense. The decision calculus favors larger sample sizes at lower resolution over smaller samples at higher resolution. This creates a funding bottleneck: the studies that could demonstrate whether single-cell heterogeneity matters clinically are the same studies deemed too expensive to justify without prior evidence that it matters.
The computational burden adds a second barrier. Single-cell methylation data from one tissue sample can generate terabytes of sequencing data requiring specialized analysis pipelines and high-performance computing infrastructure. Most academic labs lack both the storage capacity and the bioinformatics expertise to process population-scale single-cell datasets. The work concentrates in well-funded genome centers, which prioritize projects with clearer paths to publication and impact.
What We Measure Next
Complete genomic and epigenetic maps of human centromeres have been generated, according to recent sequencing efforts. The maps show methylation patterns at base pair resolution. They do not show single-cell variation at those positions.
The atlas of normal human cell types exists at base pair resolution. The cellular heterogeneity that impacts differentially methylated region calls remains unmeasured in single cells at population scale. The bulk measurement remains the standard.
Until funding agencies and genome centers prioritize single-cell methylation studies at population scale, we will continue to interpret averages as mechanisms, mistaking a summary statistic for a biological truth.
