PhenoAge was developed by Morgan Levine and colleagues at Yale, published in Aging (2018). It uses 9 blood biomarkers plus chronological age to estimate a "phenotypic age" that predicts all-cause mortality better than chronological age alone. A one-year increase in PhenoAge (relative to chronological age) is associated with a 9% increase in all-cause mortality risk.
The 9 biomarkers are available on a standard CBC (complete blood count) and CMP (comprehensive metabolic panel), with the exception of C-reactive protein (CRP), which must be ordered separately. All inputs are from a routine blood draw.
The model was trained on NHANES III (1988–1994) and validated on NHANES IV (1999–2018) with up to 12 years of mortality follow-up.
Citation: Levine ME, Lu AT, Quach A, et al. "An epigenetic biomarker of aging for lifespan and healthspan." Aging 2018;10(4):573–591. DOI: 10.18632/aging.101414
PhenoAge is one of several approaches to estimating biological age or mortality risk from clinical data. Here is how it compares to other notable models:
| Model | Inputs | Type | Status |
|---|---|---|---|
| PhenoAge Levine et al. 2018 |
9 blood markers + age | Gompertz mortality → biological age (years) | Open Published coefficients |
| Bortz Blood Age Humanity Inc. |
22 blood markers + age | Cox PH → biological age acceleration (years) | Open Coefficients public. Try it → |
| KDM Biological Age Klemera & Doubal 2006 |
Flexible biomarker set | Linear regression ensemble → biological age | Open Algorithm published; requires training data |
| Aging.AI 1.0 Putin et al. 2016 |
41 blood markers | Ensemble of 21 DNNs → age | Offline Architecture published, weights not released. Site gone. |
| Aging.AI 3.0 Mamoshina et al. 2018 |
19 blood markers + sex + ethnicity | Single DNN → age | Offline Trained on Korean, Canadian, E. European data. Site gone. |
| LifeScore Med360 Maier et al. 2020 (MassMutual) |
~60 lab + health history inputs | Random survival forest → 0–100 score | Offline Proprietary. API discontinued. |
Aging.AI was a free online tool by Insilico Medicine that predicted chronological age from blood tests using deep neural networks. Three versions were released:
The site (aging.ai) went offline circa 2024. The model weights were never published — only the architecture descriptions appeared in the papers. Archived versions of the site are available on the Wayback Machine.
v1.0: Putin E et al. "Deep biomarkers of human aging." Aging 2016;8(5):1021–1033.
DOI: 10.18632/aging.100968
v3.0: Mamoshina P et al. "Population specific biomarkers of human aging."
J Gerontol A 2018;73(11):1482–1490.
DOI: 10.1093/gerona/gly005
Developed at MassMutual using 20 years of life insurance application data (~1.5 million records, 23,000 observed deaths). The random survival forest model used ~60 inputs from lab tests and health questionnaires to produce a 0–100 life score representing mortality-risk percentile within age/sex/smoking cohorts. It yielded a 9% reduction in claims within the healthiest risk pool compared to traditional underwriting.
A consumer-facing tool (MyLifeScore) was publicly available but the API (api.lifescoremodels.com) has been discontinued.
Maier M et al. "Improving the Accuracy and Transparency of Underwriting with Artificial Intelligence." AI Magazine 2020;41(3):78–93.
The Klemera-Doubal method (2006) takes a different approach: it fits linear regressions of each biomarker against age, then combines them into a biological age estimate weighted by each marker's correlation with age. Unlike PhenoAge's fixed coefficients, KDM requires training on a reference population, making it harder to implement as a standalone calculator. It is available in the BioAge R package (160+ stars) alongside PhenoAge and homeostatic dysregulation measures.
Klemera P, Doubal S. "A new approach to the concept and computation
of biological age." Mech Ageing Dev 2006;127(3):240–248.
Kwon D, Belsky DW. "A toolkit for quantification of biological age
from blood chemistry and organ function test data: BioAge."
GeroScience 2021;43:2795–2808.
DOI: 10.1007/s11357-021-00480-5
Beyond blood biomarkers, physical activity is a strong predictor of mortality. Fishman et al. (2016) used NHANES accelerometer data (2003–2006, n=3,029 adults aged 50–79) to show that those in the highest tertile of total activity had one-fifth the mortality risk of those in the lowest (HR: 0.21). Replacing just 30 minutes of daily sedentary time with light activity was associated with a 20% reduction in mortality risk at 5 years (HR: 0.80).
Fishman EI et al. "Association between Objectively Measured Physical Activity and Mortality in NHANES." Med Sci Sports Exerc 2016;48(7):1303–1311. DOI: 10.1249/MSS.0000000000000885
The papers, spreadsheet, and analysis notes behind this calculator are available in the references folder.