How to Test Your Biological Age and Which Numbers Are Worth Tracking

The short answer is that there is no single accurate biological age test, and the useful thing to measure is not an age at all but a rate. Second-generation DNA methylation clocks trained on mortality and morbidity outcomes, plus pace-of-aging measures like DunedinPACE, are the best-validated options available to an individual today. Consumer tests run roughly $200 to $500 and return a single number. That number is close to useless on its own, because its measurement error is often larger than the annual change you are trying to detect. What makes the measurement worth anything is raw data, repeated sampling on the same platform, and computing several clocks yourself.

What a biological age test measures

Nearly every commercial biological age test measures DNA methylation in whole blood. Methylation is a chemical mark on cytosines, usually at CpG dinucleotides, and the fraction of cells carrying that mark at a given site shifts predictably with age. An epigenetic clock is a regression model over a few dozen to a few hundred of those sites. The output is scaled in years because the training label was measured in years, not because the model has measured anything intrinsically temporal.

That distinction matters for interpreting your result. A clock trained to predict chronological age is, at best, an accurate age guesser. First-generation clocks such as Horvath’s multi-tissue predictor and Hannum’s blood predictor reach median absolute errors in the range of three to four years against known age ¹. A test that nails your birth date tells you nothing about your health. The signal people care about is the residual: the gap between predicted and chronological age. Training on chronological age actively removes the variance that produces that residual, which is why the field moved on.

The clocks worth computing, and what each was trained on

Second-generation clocks skip chronological age as the target. PhenoAge was built by first fitting a mortality model to nine clinical chemistry markers plus age, then training 513 CpG sites to predict that composite phenotype ². The resulting score outperformed chronological age for predicting all-cause mortality and healthspan-related outcomes across several cohorts ². GrimAge follows a similar structure using methylation surrogates for plasma proteins and smoking history.

DunedinPoAm took a different route. Instead of predicting a cross-sectional phenotype, Belsky and colleagues measured the within-person rate of change across 18 organ-system biomarkers in a birth cohort followed from age 26 to 38, then trained methylation to predict that slope ³. The output is a ratio: years of physiological decline per calendar year, centered near 1.0. Its successor, DunedinPACE, uses a more reliable biomarker set and has substantially better test-retest behavior. Pace measures answer a different question than age measures. PhenoAge asks how much damage has accumulated. DunedinPACE asks how fast it is accumulating now. If you are going to change something and measure the effect, the rate measure is the one that can move in under a year.

We would compute PhenoAge, GrimAge if you can access it, DunedinPACE, and one principal-component clock, then track all of them. Disagreement between them is information. They are not measuring the same underlying process, and the mechanistic basis of any of them remains open ⁴.

There is also a free version worth running tonight. The clinical-chemistry PhenoAge that the methylation clock was trained to imitate is computable from a standard blood panel plus chronological age. The panel values are listed below.²

• albumin
• creatinine
• glucose
• log C-reactive protein
• lymphocyte percent
• mean cell volume
• red cell distribution width
• alkaline phosphatase
• white blood cell count

The coefficients are published. If you have a recent comprehensive metabolic panel and CBC, you already have most of the inputs.

Getting data you can compute on

Buy the array. Do not buy the score. What you want back from a provider is the raw IDAT files from an Illumina methylation array, one red and one green channel per sample. Without IDATs you cannot rerun a clock when a better one is published, and you cannot apply your own normalization or quality control.

Watch the array version. The EPIC v1 array covered about 850,000 CpGs. EPIC v2 added content and dropped or renamed a substantial number of v1 probes, which breaks published clocks that depend on missing sites. Most clock implementations handle this by imputing absent probes to a reference mean, and that imputation is a silent source of error. Ask which array the lab runs and keep every longitudinal sample on the same one.

A workable local pipeline in R:

library(sesame)
sesameDataCache()
betas <- openSesame("idats/", prep = "QCDPB", func = getBetas)

library(methylclock)
ages <- DNAmAge(betas, clocks = c("Horvath", "Hannum", "Levine", "skinHorvath"))

# DunedinPACE (github.com/danbelsky/DunedinPACE)
pace <- DunedinPACE::PACEProjector(betas, proportionOfProbesRequired = 0.8)

The QCDPB preparation in sesame applies masking of low-quality and SNP-affected probes, dye bias correction, pOOBAH detection-p masking, and background subtraction. Note that different preprocessing choices shift clock output by a meaningful fraction of a year, so freeze your pipeline version and rerun old samples with it when you upgrade.

Failure modes that move the number without your biology moving

The dominant confounder in whole blood is cell composition. Methylation differs sharply between neutrophils, monocytes, and lymphocyte subsets. A shift in your differential count therefore changes your methylome without any change in aging. An acute infection, a recent hard training block, or a blood draw at a different time of day can all do this. Estimate proportions and check them before you interpret a change:

library(FlowSorted.Blood.EPIC)
counts <- estimateCellCounts2(rgSet, referencePlatform = "IlluminaHumanMethylationEPIC")

The second failure mode is technical. Bisulfite conversion efficiency and DNA input mass both contribute variance. So do chip and row position and processing batch. Wagner has argued at length that translating methylation biomarkers into clinical use requires exactly this kind of standardization work, and that it is not yet done ⁵. First-generation clocks in particular show test-retest variability on the order of a year or more between replicate samples from the same blood draw. Reviews of the field flag reliability and the absence of standardized protocols as the main obstacles to practical use ⁶.

The practical response is to design around it. Run two technical replicates per timepoint from the same draw, and treat the spread between them as your noise floor. Sample no more often than every six months. Keep the lab and the array version constant. Hold the draw time and your fasting state constant as well. If a single unreplicated score changes by two years, the most likely explanation is the assay.

Nothing here should be read as a clinical result. Epigenetic clocks are research instruments with no regulatory clearance for diagnosis. If a blood marker feeding a chemistry-based score is out of range, that is a matter for a physician. A spreadsheet cannot settle it.

What else indicates biological age

Methylation is not the only readout, and it may not be the best one. Functional measures such as VO2 max, grip strength, and gait speed carry strong prognostic information and cost nothing but effort to collect. Newer approaches read chromatin organization directly from single-cell images rather than from bulk methylation arrays ⁷. Broader reviews of the theory behind epigenetic clocks are worth reading before you over-interpret any single score ⁸.

The framing is that biological age is a construct, and each test operationalizes it differently. Track several, track them over years, and treat the trend as the measurement.

Questions people also ask

How much does a biological age test cost? Consumer methylation tests run about $200 to $500 per sample. A research-grade EPIC array through a core facility costs roughly $300 to $400 including DNA extraction, and it returns IDAT files you own.

What is the most accurate biological age test? For predicting mortality and healthspan outcomes, second-generation clocks such as PhenoAge and GrimAge outperform chronological age and first-generation clocks ². For detecting change within a person over a short window, DunedinPACE is the better instrument because it was trained on longitudinal rate of change ³.

How accurate is a DNA test for age? Clocks trained on chronological age typically land within three to four years of true age ¹. That is accurate as forensic age estimation and largely irrelevant as a health measure.

Is there a free biological age test? Yes. The clinical-chemistry PhenoAge score uses nine routine blood markers plus your age, and the published coefficients let you compute it from a standard panel ².

What is the most accurate indicator of biological age? No single marker holds that title. A composite of a pace-of-aging methylation measure, a blood chemistry score, and functional capacity tests gives a more stable picture than any one of them ⁶.

Footnotes

1. Daniel J. Simpson, Tamir Chandra. Epigenetic age prediction. Aging Cell, 2021. https://doi.org/10.1111/acel.13452 ↩ ↩²

2. Morgan E. Levine, Ake T. Lu, Austin Quach, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging, 2018. https://doi.org/10.18632/aging.101414 ↩ ↩² ↩³ ↩⁴ ↩⁵

3. Daniel W Belsky, Avshalom Caspi, Louise Arseneault, et al. Quantification of the pace of biological aging in humans through a blood test, the DunedinPoAm DNA methylation algorithm. eLife, 2020. https://doi.org/10.7554/elife.54870 ↩ ↩²

4. Adam Li, Zane Koch, Trey Ideker. Epigenetic aging: Biological age prediction and informing a mechanistic theory of aging. Journal of Internal Medicine, 2022. https://doi.org/10.1111/joim.13533 ↩

5. Wolfgang Wagner. How to Translate DNA Methylation Biomarkers Into Clinical Practice. Frontiers in Cell and Developmental Biology, 2022. https://doi.org/10.3389/fcell.2022.854797 ↩

6. Ran Duan, Qiaoyu Fu, Yu Sun, et al. Epigenetic clock: A promising biomarker and practical tool in aging. Ageing Research Reviews, 2022. https://doi.org/10.1016/j.arr.2022.101743 ↩ ↩²

7. Alexey terskikh, Martin Alvarez-Kuglen, Delany Rodriguez, et al. Imaging-based chromatin and epigenetic age, ImAge, quantitates aging and rejuvenation. 2023. https://doi.org/10.21203/rs.3.rs-3479973/v1 ↩

8. Calen P. Ryan. “Epigenetic clocks”: Theory and applications in human biology. American Journal of Human Biology, 2020. https://doi.org/10.1002/ajhb.23488 ↩