How to Run Morgan Levine's Biological Age Test on Your Own Blood Work
There is no single “Morgan Levine biological age test” you can buy. There are two distinct things from her work, and they answer different questions. The first is PhenoAge, a mortality-calibrated regression on nine standard clinical chemistry and hematology values plus chronological age, published by Levine and colleagues in 2018. You can compute it today from any complete blood count and comprehensive metabolic panel with a high-sensitivity CRP added, for the cost of the blood draw. The second is DNAm PhenoAge, an epigenetic clock trained to predict that blood-chemistry PhenoAge from DNA methylation at 513 CpG sites, which requires a methylation array and costs more. If you want a number this week, compute PhenoAge from labs you probably already have. If you want a measurement that tracks something closer to underlying cellular aging, the methylation route is better, with caveats about reliability we will get to.
What PhenoAge is measuring
PhenoAge was built in two stages. First, a Cox proportional hazards model with elastic-net penalization was fit on NHANES III mortality data across 42 candidate clinical markers, selecting nine plus chronological age. Second, the resulting linear predictor was converted into an age-equivalent by asking: what chronological age, in the reference population, carries this same 10-year mortality risk? So a PhenoAge of 52 means your marker profile matches the average mortality hazard of a 52-year-old. It is a mortality-risk score expressed in years, which is why it is interpretable and also why it is dominated by markers of inflammation and red cell distribution.
That framing matters for how you read your result. PhenoAge is not measuring a cellular aging process directly. It is a compression of nine correlated clinical signals onto a single mortality axis. A high PhenoAge tells you the profile looks like that of an older person, and the useful next step is to look at which of the nine terms drove it.
The formula and the code
The linear predictor is a weighted sum with an intercept of −19.9067. The weights, in the units Levine et al. used:
| Marker | Units | Coefficient |
|---|---|---|
| Albumin | g/L | −0.0336 |
| Creatinine | µmol/L | 0.0095 |
| Glucose (serum) | mmol/L | 0.1953 |
| ln(C-reactive protein) | mg/dL | 0.0954 |
| Lymphocyte percent | % | −0.0120 |
| Mean cell volume | fL | 0.0268 |
| Red cell distribution width | % | 0.3306 |
| Alkaline phosphatase | U/L | 0.00188 |
| White blood cell count | 10³ cells/µL | 0.0554 |
| Chronological age | years | 0.0804 |
The linear predictor is then mapped through a Gompertz survival function over 120 months and inverted:
import numpy as np
COEF = dict(albumin_g_L=-0.0336, creatinine_umol_L=0.0095,
glucose_mmol_L=0.1953, ln_crp_mg_dL=0.0954,
lymph_pct=-0.0120, mcv_fL=0.0268, rdw_pct=0.3306,
alp_U_L=0.00188, wbc_1000_uL=0.0554, age_years=0.0804)
INTERCEPT = -19.9067
GAMMA = 0.0076927 # Gompertz shape, per month
def phenoage(**m):
xb = INTERCEPT + sum(COEF[k] * m[k] for k in COEF)
mort = 1 - np.exp(-np.exp(xb) * (np.exp(120 * GAMMA) - 1) / GAMMA)
return 141.50225 + np.log(-0.00553 * np.log(1 - mort)) / 0.090165Verify the constants against the paper’s supplementary material before you trust a number to a decimal place. Then compute the per-marker contribution, COEF[k] * m[k], and sort it. That decomposition is more informative than the scalar output, because it tells you whether your result is driven by CRP, by RDW, or simply by being 48 years old.
The unit conversions that silently break it
Most wrong PhenoAge numbers online are unit errors. US labs report glucose in mg/dL (divide by 18.016 for mmol/L), creatinine in mg/dL (multiply by 88.4 for µmol/L), albumin in g/dL (multiply by 10 for g/L), and CRP in mg/L (divide by 10 for mg/dL before taking the natural log). Getting creatinine wrong by a factor of 88 moves the linear predictor by roughly 0.8, which is about ten years of apparent age.
Three other failure modes are worth knowing. High-sensitivity CRP below the assay’s limit of detection, often reported as “<0.3 mg/L,” has no logarithm, so you must substitute a floor value and note that you did. Lymphocyte percent is the differential percentage, not the absolute count in 10³/µL, and the two differ by roughly an order of magnitude. Alkaline phosphatase assays vary by platform and reference interval, so serial values from different labs are not strictly comparable.
Finally, note the sensitivity structure. The RDW coefficient of 0.3306 per percentage point is about four times the annual age coefficient, so a 1-point RDW shift is worth roughly four years of PhenoAge. Analyzer RDW coefficients of variation are typically a few percent, and an acute infection will raise CRP and WBC together. A single PhenoAge reading during a cold can easily read five years older than the same person a month later. Compute it on at least three draws taken in a stable, fasted, non-acute state and look at the trend.
Where the methylation clocks fit
DNAm PhenoAge was trained on Illumina 450K methylation data to predict the blood-chemistry score, and it outperformed earlier clocks on mortality and healthspan endpoints. The broader family of methylation clocks now includes Horvath’s multi-tissue clock, GrimAge, and DunedinPACE, which estimates rate of aging rather than a cumulative age, and for which population norms have been published so you can place your value on a distribution 1. Levine’s own review of the field is the most useful thing to read before ordering one, because it is direct about what these measures do and do not establish 2. Horvath and Raj give the mechanistic background on why methylation tracks age at all 3.
Two practical points. First, technical reliability is a real problem: first-generation clocks computed from single CpGs have test-retest variability large enough to swamp modest intervention effects, and principal-component and systems-based reformulations were developed specifically to fix this 4. If you are going to track a methylation clock over time, use a version with published reliability metrics and run technical replicates on the same DNA. Second, a single composite age number throws away structure. Systems Age decomposes one blood methylation assay into eleven physiological system scores, so you can see that, say, your inflammatory and metabolic axes diverge 5. That decomposition is where the actionable signal lives, and the same logic applies to PhenoAge: read the nine terms, not just the sum.
Clocks also have organ-specific behavior. Methylation age in prefrontal cortex tissue is associated with neuritic plaque load and amyloid burden, which is a reminder that a blood-derived clock is a blood-derived clock 6. And the exposures that move these measures include psychosocial ones: traumatic life experience has been associated with accelerated epigenetic aging 7.
What we would do
Order a standard CBC with differential, CMP, and hs-CRP fasted, compute PhenoAge with the code above, and store the raw nine values rather than just the output so you can recompute under any future formula. Repeat quarterly. Separately, run one methylation array and compute several clocks from the same IDAT files, preferring the reliability-corrected versions and a decomposed output over a single number. If any individual marker falls outside its reference interval, that is a clinical finding and belongs with a physician, not a calculator. PhenoAge is a research instrument for tracking your own trajectory, and it does not diagnose anything.
Questions people also ask
How much does a biological age test cost? The blood-chemistry PhenoAge costs whatever the panel costs, since the algorithm is published and free. Consumer methylation clock tests generally run a few hundred dollars, and the price mostly buys array processing plus a proprietary report.
Can a blood test tell your biological age? It can give you a mortality-calibrated age equivalent, which is what PhenoAge is. It cannot tell you a cellular age in any mechanistic sense.
Do we age rapidly at 34, 60, and 78? Those ages come from a cross-sectional plasma proteomics study that found population-level inflection points in protein abundance trajectories. They describe cohort averages, not scheduled events in an individual, and nothing in that work implies you should expect a step change on a birthday.
What is the best indicator of biological age? There is no single best one. For prognostic performance per dollar, PhenoAge from routine labs is hard to beat. For tracking change over months, a pace-of-aging measure such as DunedinPACE with known reliability characteristics is the better instrument 1. Reviews of clinical application remain cautious about using any of them for individual decisions 8.
Can I test my biological age at home? You can collect a saliva or blood-spot sample at home for a methylation assay, but tissue and collection method affect the result, and home-collected samples are harder to replicate. A venous draw at a lab gives cleaner serial data.
Oak builds longitudinal molecular profiles of individuals: whole-genome sequencing, RNA sequencing, proteomics, blood biomarkers, and continuous glucose data, integrated into one model of you. Build your profile.
Footnotes
-
Kyle J. Bourassa, Calen P. Ryan, Karen Sugden, et al. Age norms for DunedinPACE: An epigenetic pace of aging biomarker. 2026. https://doi.org/10.64898/2026.08.13.26360306 ↩ ↩2
-
Morgan E Levine. Assessment of Epigenetic Clocks as Biomarkers of Aging in Basic and Population Research. The Journals of Gerontology: Series A, 2020. https://doi.org/10.1093/gerona/glaa021 ↩
-
Steve Horvath, Kenneth Raj. DNA methylation-based biomarkers and the epigenetic clock theory of ageing. Nature Reviews Genetics, 2018. https://doi.org/10.1038/s41576-018-0004-3 ↩
-
Raghav Sehgal, Daniel S. Borrus, John Gonzalez, et al. Biological Versus Technical Reliability of Epigenetic Clocks and Implications for Disease Prognosis and Intervention Response. Aging Cell, 2026. https://doi.org/10.1111/acel.70635 ↩
-
Raghav Sehgal, Yaroslav Markov, Chenxi Qin, et al. Systems Age: a single blood methylation test to quantify aging heterogeneity across 11 physiological systems. Nature Aging, 2025. https://doi.org/10.1038/s43587-025-00958-3 ↩
-
Morgan E Levine, Ake T Lu, David A Bennett, et al. Epigenetic age of the pre-frontal cortex is associated with neuritic plaques, amyloid load, and Alzheimer’s disease related cognitive functioning. Aging, 2015. https://doi.org/10.18632/aging.100864 ↩
-
Morgan Levine. 230. Traumatic Life Experiences are Associated With Increases in Epigenetic Aging. Biological Psychiatry, 2018. https://doi.org/10.1016/j.biopsych.2018.02.249 ↩
-
A. Baki Yildirim, Duygu T. Yildirim, Hilal Akalin, et al. Epigenetic Clocks: Biological Aging, Disease, and Clinical Applications. The EuroBiotech Journal, 2026. https://doi.org/10.2478/ebtj-2026-0014 ↩