Research

Built on peer reviewed science.

Alpha stands on the shoulders of a well developed scientific field. Our contribution is the unification, the machine learning architecture and the multi population scale at which we operate. Our validation pathway is designed to sit alongside the published literature, not to obscure it.

Medical validation

Five pillars of clinical evidence.

Alpha is validated against the same evidentiary standards accepted by the peer reviewed literature on second and third generation aging clocks. Every population model is evaluated on the following pillars before it is promoted to production.

Mortality prediction

All cause mortality hazard ratios reported per standard deviation of Alpha Age, benchmarked against chronological age and PhenoAge in each population.

Disease incidence

Prospective association with cardiovascular events, type 2 diabetes, cognitive decline and frailty over multi year follow up in reference cohorts.

Population calibration

Per population mean absolute error against chronological age reported separately for each of the seven Alpha models, not a single global metric.

Test retest reliability

Intraclass correlation coefficients on repeat blood draws within the pre analytical window agreed with clinical partners.

Intervention response

Alpha Pace sensitivity to caloric restriction, metformin and lifestyle intervention arms benchmarked against DunedinPACE.

C index 0.78

All cause mortality, reference EU cohort

MAE 4.2 yr

Alpha Age vs chronological, healthy adults

ICC 0.94

Test retest, 14 day repeat draw

r = 0.71

Alpha Pace vs DunedinPACE, methylation subset

Benchmark values reported here summarise internal validation on the current production models and are illustrative of the metrics disclosed in the full technical brief. Alpha outputs are decision support signals, not a medical diagnosis. See the medical disclaimer.

Methodology to evidence

Every step in the Alpha pipeline is tied to a peer reviewed precedent.

The table below maps each methodological choice in the Alpha architecture to the published evidence that motivates it and to the metric class we report against it. Footnote numbers refer to the sources listed at the end of this section.

Methodology stepAlpha approachPeer reviewed evidence baseMetric reported
Biomarker selectionPanel of routinely available blood biomarkers spanning eight physiological systems.PhenoAge (Levine 2018) established that nine standard chemistry markers outperform chronological age for mortality prediction.[1] [2]Feature stability index across cohorts
Model architectureUnified deep learning backbone with a per population calibration head.Deep neural network aging clocks (Putin 2016) and population specific calibration (Mamoshina 2018) demonstrate lower error than single global models.[3] [4]MAE vs chronological age per population
Mortality anchoringAlpha Age calibrated against time to death in reference cohorts with linkage.Meta analysis of DNA methylation age (Chen 2016) and Marioni 2015 show hazard ratios of ~1.05 per year of age acceleration for all cause mortality.[5] [6]Hazard ratio per SD, C index for all cause mortality
Pace of agingAlpha Pace trained to track longitudinal rate of change, not cumulative damage.DunedinPACE (Belsky 2022) validates a rate based estimator that is sensitive to intervention and correlates with functional decline.[7]Correlation with DunedinPACE, response to intervention arms
System level decompositionPer organ system sub scores reported alongside the composite Alpha Age.Systems Age (Sehgal 2023) and organ specific plasma proteome clocks (Oh 2023) show that organ level aging heterogeneity carries independent prognostic signal.[8] [9]Independent association of sub score with organ specific incident disease
Intervention sensitivityAlpha Pace evaluated in interventional cohorts (caloric restriction, metformin, lifestyle).CALERIE (Waziry 2023) reported a significant slowing of DunedinPACE under 25 percent caloric restriction over two years.[10]Between arm delta in pace of aging, effect size versus control
Population generalisationSeven regional models with region specific reference distributions.Cross population aging clock work (Mamoshina 2018) showed accuracy loss when a single model is applied across ancestries.[4]Per population MAE, calibration slope, subgroup C index
Pre analytical robustnessTest retest on repeat draws within a defined pre analytical window.Reliability of biological age estimators is an established prerequisite for clinical use (Higgins Chen et al., 2022 principal components approach).[11]Intraclass correlation coefficient (ICC) on 14 day repeat

Study outcome metrics

Alpha benchmarks alongside the published reference values.

Each row shows the outcome, the reference value reported in the peer reviewed literature, and the corresponding value observed for Alpha on internal validation cohorts. Full protocols, cohort definitions and statistical methods are available in the technical brief under NDA.

OutcomeReference value (literature)Alpha value (internal)Sources
All cause mortality, C index0.72 (PhenoAge, NHANES IV)0.78 (Alpha Age, EU reference cohort)[2] [12]
All cause mortality, hazard ratio per SD1.09 per year of methylation age acceleration1.31 per SD of Alpha Age (age and sex adjusted)[6] [12]
MAE vs chronological age, healthy adults3.6 yr (Horvath 2013, multi tissue)4.2 yr (Alpha Age, pooled populations)[13] [12]
Test retest reliability, ICC0.96 (PC clocks, Higgins Chen 2022)0.94 (Alpha Age, 14 day repeat draw)[11] [12]
Correlation with pace of agingDunedinPACE reference distributionr = 0.71 (Alpha Pace vs DunedinPACE, methylation subset)[7] [12]
Intervention sensitivity, caloric restrictionSignificant slowing of DunedinPACE at 24 months (CALERIE)Directionally consistent slowing of Alpha Pace in re analysed CALERIE arm[10] [12]
Inflammatory ageing signal, cardiovasculariAge tracks multimorbidity and cardiovascular aging (Sayed 2021)Independent association of Alpha inflammatory sub score with incident CV events[14] [12]
Organ specific ageing signalPlasma proteome organ clocks predict organ specific disease (Oh 2023)Alpha system sub scores associate with organ specific incident disease at 5 yr[9] [12]

Sources

Footnote references.

  1. [1]Levine ME et al. (2018). Aging (Albany NY) 10(4):573 591. Source ↗
  2. [2]Liu Z et al. (2018). PLOS Medicine 15(12):e1002718. Source ↗
  3. [3]Putin E et al. (2016). Aging (Albany NY) 8(5):1021 1033. Source ↗
  4. [4]Mamoshina P et al. (2018). J Gerontol A Biol Sci Med Sci 73(11):1482 1490. Source ↗
  5. [5]Chen BH et al. (2016). Aging (Albany NY) 8(9):1844 1865. Source ↗
  6. [6]Marioni RE et al. (2015). Genome Biology 16:25. Source ↗
  7. [7]Belsky DW et al. (2022). eLife 11:e73420 (DunedinPACE). Source ↗
  8. [8]Sehgal R et al. (2023). bioRxiv 2023.07.13.548904 (Systems Age). Source ↗
  9. [9]Oh HS-H et al. (2023). Nature 624:164 172 (organ ageing proteome). Source ↗
  10. [10]Waziry R et al. (2023). Nature Aging 3:248 257 (CALERIE). Source ↗
  11. [11]Higgins Chen AT et al. (2022). Nature Aging 2:644 661 (PC clocks). Source ↗
  12. [12]Alpha Longevity internal validation on current production models. Full metrics available in the technical brief under NDA. Source ↗
  13. [13]Horvath S. (2013). Genome Biology 14:R115. Source ↗
  14. [14]Sayed N et al. (2021). Nature Aging 1:598 615 (iAge). Source ↗

Reference values are drawn from the peer reviewed publications cited above; Alpha values are current production benchmarks reported by Alpha Longevity. Alpha outputs remain decision support signals and are not a medical diagnosis. See the medical disclaimer.

Full citation list

Complete bibliography with DOIs.

Every peer reviewed reference cited across this page, in a single canonical list. Filter by clock type, intervention, cohort size or year to zero in on the studies most relevant to your programme. Each entry links to the publisher record via its DOI.

26 of 26

Clock type

Intervention

Cohort size

Year

  1. 2006Klemera P, Doubal S. (2006). A new approach to the concept and computation of biological age. Mechanisms of Ageing and Development 127(3):240 248.
    Framework / reviewMedium (1k-10k)
  2. 2013Horvath S. (2013). DNA methylation age of human tissues and cell types. Genome Biology 14:R115.
    MethylationLarge (>10k)
  3. 2013Hannum G et al. (2013). Genome wide methylation profiles reveal quantitative views of human aging rates. Molecular Cell 49(2):359 367.
    MethylationMedium (1k-10k)
  4. 2015Marioni RE et al. (2015). DNA methylation age of blood predicts all cause mortality in later life. Genome Biology 16:25.
    MethylationLarge (>10k)
  5. 2016Chen BH et al. (2016). DNA methylation based measures of biological age: meta analysis predicting time to death. Aging (Albany NY) 8(9):1844 1865.
    MethylationMeta analysis
  6. 2016Putin E et al. (2016). Deep biomarkers of human aging: application of deep neural networks to biomarker development. Aging (Albany NY) 8(5):1021 1033.
    Blood chemistryLarge (>10k)
  7. 2016Barzilai N et al. (2016). Metformin as a tool to target aging (TAME rationale). Cell Metabolism 23(6):1060 1065.
    Framework / reviewSmall (<1k)Metformin
  8. 2018Mamoshina P et al. (2018). Population specific biomarkers of human aging: a big data study using Korean, Canadian and Eastern European populations. J Gerontol A Biol Sci Med Sci 73(11):1482 1490.
    Blood chemistryLarge (>10k)
  9. 2018Levine ME et al. (2018). An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY) 10(4):573 591.
    Blood chemistryLarge (>10k)
  10. 2018Liu Z et al. (2018). A new aging measure captures morbidity and mortality risk across diverse subpopulations from NHANES IV. PLOS Medicine 15(12):e1002718.
    Blood chemistryLarge (>10k)
  11. 2018Justice JN et al. (2018). A framework for selection of blood based biomarkers for geroscience guided clinical trials: report from the TAME biomarkers workgroup. GeroScience 40:419 436.
    Framework / reviewMedium (1k-10k)
  12. 2018Ferrucci L, Fabbri E. (2018). Inflammageing: chronic inflammation in ageing, cardiovascular disease and frailty. Nature Reviews Cardiology 15:505 522.
    InflammationLarge (>10k)
  13. 2019Lu AT et al. (2019). DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging (Albany NY) 11(2):303 327.
    MethylationLarge (>10k)
  14. 2019Fahy GM et al. (2019). Reversal of epigenetic aging and immunosenescent trends in humans. Aging Cell 18(6):e13028.
    MethylationSmall (<1k)Reprogramming
  15. 2019Crimmins EM et al. (2019). Associations of age, sex, race/ethnicity and education with 13 epigenetic clocks in a nationally representative U.S. sample: HRS. J Gerontol A Biol Sci Med Sci 76(6):1117 1123.
    MethylationLarge (>10k)
  16. 2020Kirkland JL, Tchkonia T. (2020). Senolytic drugs: from discovery to translation. Journal of Internal Medicine 288(5):518 536.
    SenescenceMedium (1k-10k)Senolytics
  17. 2020Lu Y et al. (2020). Reprogramming to recover youthful epigenetic information and restore vision. Nature 588:124 129 (Sinclair lab).
    MethylationSmall (<1k)Reprogramming
  18. 2020Ahadi S et al. (2020). Personal aging markers and ageotypes revealed by deep longitudinal profiling. Nature Medicine 26:83 90 (Snyder lab).
    Multi omicSmall (<1k)
  19. 2021Galkin F et al. (2021). Human microbiome aging clocks based on deep learning and tandem of permutation feature importance and accumulated local effects. iScience 24(6):102550.
    MicrobiomeMedium (1k-10k)
  20. 2021Sayed N et al. (2021). An inflammatory aging clock (iAge) based on deep learning tracks multimorbidity, immunosenescence and cardiovascular aging. Nature Aging 1:598 615.
    InflammationMedium (1k-10k)
  21. 2022Belsky DW et al. (2022). DunedinPACE, a DNA methylation biomarker of the pace of aging. eLife 11:e73420.
    MethylationLarge (>10k)
  22. 2022Higgins Chen AT et al. (2022). A computational solution for bolstering reliability of epigenetic clocks: implications for clinical trials and longitudinal tracking. Nature Aging 2:644 661.
    MethylationMedium (1k-10k)
  23. 2022Rutledge J, Oh HS-H, Wyss-Coray T. (2022). Measuring biological age using omics data. Nature Reviews Genetics 23:715 727.
    Multi omicFramework / reviewMeta analysis
  24. 2023Waziry R et al. (2023). Effect of long term caloric restriction on DNA methylation measures of biological aging in healthy adults (CALERIE trial). Nature Aging 3:248 257.
    MethylationSmall (<1k)Caloric restriction
  25. 2023Sehgal R et al. (2023). Systems age: a single blood methylation test to quantify aging heterogeneity across 11 physiological systems. bioRxiv 2023.07.13.548904.
    MethylationMedium (1k-10k)
  26. 2023Oh HS-H et al. (2023). Organ aging signatures in the plasma proteome track health and disease. Nature 624:164 172.
    ProteomicLarge (>10k)

Foundational publications

The intellectual foundation of the modern aging clock field.

Levine ME et al. (2018). An epigenetic biomarker of aging for lifespan and healthspan.

Aging (Albany NY) 10(4):573 591

Liu Z et al. (2018). A new aging measure captures morbidity and mortality risk across diverse subpopulations from NHANES IV.

PLOS Medicine 15(12):e1002718

Horvath S. (2013). DNA methylation age of human tissues and cell types.

Genome Biology 14:R115

Belsky DW et al. (2022). DunedinPACE, a DNA methylation biomarker of the pace of aging.

eLife 11:e73420

Lu AT et al. (2019). DNA methylation GrimAge strongly predicts lifespan and healthspan.

Aging (Albany NY) 11(2):303 327

Putin E et al. (2016). Deep biomarkers of human aging: application of deep neural networks to biomarker development.

Aging (Albany NY) 8(5):1021 1033

Mamoshina P et al. (2018). Population specific biomarkers of human aging: a big data study using Korean, Canadian and Eastern European populations.

J Gerontol A Biol Sci Med Sci 73(11):1482 1490

Klemera P, Doubal S. (2006). A new approach to the concept and computation of biological age.

Mechanisms of Ageing and Development 127(3):240 248

Hannum G et al. (2013). Genome wide methylation profiles reveal quantitative views of human aging rates.

Molecular Cell 49(2):359 367

Chen BH et al. (2016). DNA methylation based measures of biological age: meta analysis predicting time to death.

Aging (Albany NY) 8(9):1844 1865

Marioni RE et al. (2015). DNA methylation age of blood predicts all cause mortality in later life.

Genome Biology 16:25

Sehgal R et al. (2023). Systems age: a single blood methylation test to quantify aging heterogeneity across 11 physiological systems.

bioRxiv 2023.07.13.548904

Galkin F et al. (2021). Human microbiome aging clocks based on deep learning and tandem of permutation feature importance and accumulated local effects.

iScience 24(6):102550

Fahy GM et al. (2019). Reversal of epigenetic aging and immunosenescent trends in humans.

Aging Cell 18(6):e13028

Waziry R et al. (2023). Effect of long term caloric restriction on DNA methylation measures of biological aging in healthy adults (CALERIE trial).

Nature Aging 3:248 257

Justice JN et al. (2018). A framework for selection of blood based biomarkers for geroscience guided clinical trials: report from the TAME biomarkers workgroup.

GeroScience 40:419 436

Sayed N et al. (2021). An inflammatory aging clock (iAge) based on deep learning tracks multimorbidity, immunosenescence and cardiovascular aging.

Nature Aging 1:598 615

Oh HS-H et al. (2023). Organ aging signatures in the plasma proteome track health and disease.

Nature 624:164 172

Foundational researchers

The scientific community Alpha builds upon.

Over the last decade, the following researchers and institutions have produced the foundational science that makes an instrument like Alpha possible. Portraits below link to each researcher's official institutional biography.

ML

Morgan E. Levine, PhD

Altos Labs (formerly Yale School of Medicine)

Creator of PhenoAge, the second generation blood based biological age standard.

View official bio →

SH

Steve Horvath, PhD, ScD

Altos Labs (formerly UCLA)

Creator of the Horvath methylation clock and co developer of GrimAge.

View official bio →

DB

Daniel W. Belsky, PhD

Columbia Mailman School of Public Health

Developer of DunedinPACE, the leading third generation pace of aging clock.

View official bio →

AL

Ake T. Lu, PhD

Altos Labs

Co developer of GrimAge, one of the highest performing mortality anchored clocks.

View official bio →

VG

Vadim N. Gladyshev, PhD

Harvard Medical School / Brigham and Women's

Leading laboratory in aging biomarker discovery and clock methodology.

View official bio →

LF

Luigi Ferrucci, MD, PhD

National Institute on Aging (NIH)

Scientific Director, principal architect of the Baltimore Longitudinal Study of Aging.

View official bio →

AZ

Alex Zhavoronkov, PhD

Insilico Medicine / Deep Longevity

Developer of deep neural network aging clocks trained on blood biochemistry and methylation data.

View official bio →

MS

Michael P. Snyder, PhD

Stanford University School of Medicine

Pioneer of multi omic longitudinal aging profiling and personal aging trajectories.

View official bio →

NB

Nir Barzilai, MD

Albert Einstein College of Medicine

Director, Institute for Aging Research; principal investigator, TAME trial.

View official bio →

EC

Eileen Crimmins, PhD

University of Southern California

Architect of Health and Retirement Study biomarker collection and international HRS family surveys.

View official bio →

JK

James P. Kirkland, MD, PhD

Mayo Clinic (Kogod Center on Aging)

Pioneer of senolytic clinical trials linking cellular senescence to biological age.

View official bio →

DS

David A. Sinclair, PhD

Harvard Medical School

Principal investigator on the biology of aging, NAD metabolism and epigenetic reprogramming.

View official bio →

AM

Andrea B. Maier, MD, PhD

National University of Singapore (Healthy Longevity)

Clinical translation of biological age biomarkers in Asian populations.

View official bio →