Epigenetic Clocks

1. Overview & Definition

Epigenetic clocks are statistical models that estimate biological age by measuring DNA methylation (DNAm) at specific CpG dinucleotides across the genome. Unlike chronological age, which measures time since birth, biological age reflects the cumulative molecular and cellular damage that drives functional decline and disease risk.

The first generation of epigenetic clocks, developed by Steve Horvath in 2013, demonstrated that DNAm patterns could predict chronological age with remarkable accuracy (correlation r > 0.90) across multiple tissues, cell types, and species. Subsequent generations have incorporated clinical biomarkers, mortality data, and pace-of-aging measures to improve predictive power for health outcomes beyond chronological age.

2. Biology of DNA Methylation

2.1 The Epigenetic Landscape

DNA methylation is a covalent modification of cytosine at CpG dinucleotides, catalyzed by DNA methyltransferases (DNMTs). In mammals, approximately 70–80% of CpG sites are methylated, with the remainder concentrated in CpG islands — regions of high CpG density typically associated with gene promoters.

Methylation patterns are established during development, maintained through cell division by DNMT1, and can be actively removed by the TET family of dioxygenases. Aberrant methylation patterns are a hallmark of aging and are associated with:

• Epigenetic drift: Progressive loss of methylation fidelity at CpG islands and increased methylation at non-island sites
• Hypermethylation: Increased methylation at promoter CpG islands, often silencing tumor suppressor genes and DNA repair genes
• Hypomethylation: Loss of methylation at repetitive elements and intergenic regions, contributing to genomic instability

2.2 Mechanistic Links to Aging

DNA methylation changes are mechanistically intertwined with multiple aging processes:

• Genomic instability: Hypomethylation of repetitive elements activates transposable elements, causing DNA damage
• Cellular senescence: Altered methylation at cell-cycle regulators (e.g., p16^INK4a locus) promotes senescence
• Stem cell exhaustion: Epigenetic changes impair stem cell self-renewal and differentiation capacity
• Loss of proteostasis: Methylation changes alter chaperone and autophagy gene expression
• Inflammation: Methylation changes at immune genes promote chronic inflammation

3. Major Epigenetic Clocks

3.1 First-Generation Clocks

Horvath Clock (2013)

The first pan-tissue epigenetic clock, developed by Steve Horvath at UCLA, uses 353 CpG sites selected through elastic net regression from 8,000 samples across 51 tissues and cell types. The Horvath clock estimates chronological age with a median error of 3.6 years and is highly correlated with chronological age (r = 0.96) across diverse tissues.

Key features:

• Multi-tissue applicability (blood, brain, liver, kidney, skin, etc.)
• Works across the entire lifespan (fetus to centenarian)
• Correlates with mitotic age in cell culture
• Accelerated in Down syndrome, HIV infection, and obesity

Hannum Clock (2013)

Developed by Gregory Hannum and colleagues, the Hannum clock uses 71 CpG sites from whole blood. It is optimized for blood samples and incorporates smoking-related CpG sites. The Hannum clock shows strong correlation with chronological age (r = 0.91) and has been validated in large cohorts.

3.2 Second-Generation Clocks

PhenoAge (2018)

Developed by Morgan Levine and colleagues, PhenoAge integrates clinical biomarkers (albumin, creatinine, glucose, CRP, lymphocyte percent, mean cell volume, red cell distribution width, alkaline phosphatase, white blood cell count) with DNA methylation data using 513 CpG sites. PhenoAge estimates a biological age associated with physiological decline and morbidity risk, rather than chronological age per se.

PhenoAge acceleration is predictive of:

• All-cause mortality (HR = 1.15 per year of acceleration)
• Cardiovascular disease incidence
• Cancer incidence
• Alzheimer's disease risk
• Physical functioning decline

GrimAge (2019)

Developed by Steve Horvath and Ake Lu, GrimAge is a composite clock that incorporates DNA methylation surrogates for plasma protein levels (plasminogen activator inhibitor-1, growth differentiation factor 15, tissue inhibitor of metalloproteinases-1, β-2 microglobulin, leptin, smoking pack-years) and chronological age. GrimAge is the strongest predictor of lifespan and time-to-death among current clocks.

GrimAge acceleration is associated with:

• Time-to-death (stronger association than any other clock)
• Time-to-coronary heart disease
• Time-to-cancer
• Time-to-menopause
• Frailty index and physical functioning

3.3 Third-Generation Clocks

DunedinPACE (2022)

Developed by Daniel Belsky and colleagues at Columbia University, DunedinPACE (Pace of Aging, Computed from the Epigenome) measures the rate of biological aging rather than cumulative biological age. It was trained on longitudinal data from the Dunedin Study, a 45-year cohort study that measured pace of aging using 18 biomarkers of organ system integrity.

DunedinPACE is conceptually distinct from earlier clocks:

• Measures "speedometer" (pace) rather than "odometer" (cumulative age)
• More sensitive to intervention effects than static biological age measures
• Validated against longitudinal decline in physical and cognitive function
• Associated with mortality risk, frailty, and dementia incidence

DunedinPoAm (2020)

A precursor to DunedinPACE, DunedinPoAm measures the rate of biological aging based on methylation patterns tied to longitudinal physiological changes. It has been validated in multiple independent cohorts.

CausAge (2024)

A newer model designed to estimate biological age by focusing on methylation changes that are causally linked to aging processes rather than merely correlated with age. CausAge uses Mendelian randomization to identify CpG sites with causal effects on aging-related traits.

ELOVL2 Clock

A simplified clock using methylation levels at CpG sites in the ELOVL2 gene, which shows one of the strongest correlations with chronological age. While less comprehensive than multi-CpG clocks, the ELOVL2 clock demonstrates that single-gene methylation can be highly predictive.

4. Clock Comparison & Performance

Clock	CpG Sites	Type	Best For	Correlation with Age
Horvath	353	Pan-tissue	General age estimation across tissues	r = 0.96
Hannum	71	Blood-specific	Blood-based age estimation	r = 0.91
PhenoAge	513	Clinical integration	Morbidity and mortality prediction	r = 0.94
GrimAge	1,030	Plasma protein surrogates	Lifespan and time-to-death	r = 0.92
DunedinPACE	173	Pace of aging	Intervention sensitivity; rate measurement	N/A (rate-based)
CheekAge	~350	Buccal-specific	Non-invasive sampling (saliva/cheek)	r = 0.92

5. Clinical Applications

5.1 Disease Risk Prediction

Epigenetic age acceleration is associated with increased risk of multiple age-related diseases:

• Cardiovascular disease: Each year of GrimAge acceleration increases CHD risk by ~15%
• Cancer: PhenoAge acceleration predicts lung, breast, and colorectal cancer incidence
• Neurodegeneration: Epigenetic age acceleration is associated with Alzheimer's disease and cognitive decline
• Diabetes: DNAm age acceleration predicts type 2 diabetes incidence independent of BMI

5.2 Transplantation Medicine

Epigenetic age of donor organs predicts graft survival and function. Older biological age (independent of chronological age) is associated with poorer outcomes in kidney, liver, and heart transplantation.

5.3 Forensic Applications

DNAm-based age estimation from trace biological samples (blood, saliva, semen) has been used in forensic investigations to narrow the age range of unidentified individuals or suspects.

5.4 Reproductive Medicine

Oocyte and sperm DNAm age may predict fertility outcomes and embryonic development potential. Epigenetic aging of the endometrium may influence implantation success.

6. Interventions & Clock Reversibility

Multiple interventions have been shown to modulate epigenetic clocks:

Intervention	Clock Effect	Evidence Level
Caloric restriction (25%)	Slowed DunedinPACE by 2–3%	RCT (CALERIE Phase 2)
Exercise	Reduced epigenetic age acceleration	Observational + small RCTs
Smoking cessation	Partial reversal of age acceleration	Longitudinal cohort
Bariatric surgery	Reduced epigenetic age by ~2–3 years	Before-after studies
Rapamycin	Modest reduction in some clocks	Observational; PEARL trial pending
NAD+ precursors	Mixed results; some studies show reduction	Small RCTs; inconsistent
Senolytics (D+Q)	Reduced epigenetic age in pilot study	Pilot trial (NCT04946383)

7. Limitations & Challenges

• Population bias: Most clocks were trained on European-ancestry cohorts and may perform less well in other ethnicities
• Tissue specificity: Blood-based clocks may not accurately reflect aging in the brain, liver, or other organs
• Batch effects: Technical variation between laboratories can confound age estimates
• Biological interpretation: The mechanisms linking specific CpG sites to aging remain largely unknown
• Commercial testing: Direct-to-consumer epigenetic age tests vary widely in accuracy and clinical validation
• Longitudinal performance: Most clocks have been validated cross-sectionally; longitudinal tracking within individuals is less established

8. References