Communications biologyApr 24, 2025

Differences in DNA methylation age between EPICv1 and EPICv2 arrays in blood and saliva samples

Longevity & Aging Weekly Brief ↗PubMed ↗DOI ↗OA ↗

Updated Mar 14, 2026

Abstract

High correlations (ρ > 0.8) were observed between Infinium MethylationEPIC (EPICv1) and EPICv2 arrays in estimating age across samples from 16 healthy individuals.

Eleven were used to estimate DNAm age from buffy coat, PBMC, and saliva samples.
Systematic biases were introduced by EPICv2, particularly affecting non-PC-based clocks and varying by biological sample type.
Saliva samples exhibited the greatest variability in DNAm age estimation between the two array versions.
PC-based clocks had lower systematic bias compared to non-PC-based clocks, with Mean Absolute Percentage Error (MAPE) ranging from 0.118% to 21.2%.

Simplified

This study aims to evaluate differences between Infinium MethylationEPIC (EPICv1) and Infinium MethylationEPICv2 (EPICv2) arrays in estimating age with eleven using buffy coat, peripheral blood mononuclear cell (PBMC), and saliva from 16 healthy middle-aged individuals. DNAm ages were estimated using six principal component-based (PC) clocks (PCHorvath1, PCHorvath2, PCHannum, PCPhenoAge, PCGrimAge, and PCDNAmTL) and five non-PC clocks (DunedinPACE, DNAmFit, YingCausAge, YingAdaptAge, and YingDamAge) across all biological samples. Agreement between arrays was assessed using Spearman correlation, Bland-Altman plots, and Wilcoxon Signed-Rank test. The 16 individuals with median age of 48 [43.5;53.8] years, were predominantly female, Chinese and non-smokers. High correlations (ρ > 0.8) were observed between EPICv1 and EPICv2 except for DunedinPACE, YingDamAge and YingAdaptAge. PC-based clocks showed lower systematic bias (MAPE:0.118-8.98%) compared to non-PC-based clocks (MAPE:5.31-21.2%). Saliva samples demonstrated greatest variability between arrays. EPICv2 introduces systematic biases especially in non-PC-based clocks and between different biological samples.

Key numbers

48 [43.5; 53.8] years

Median Age of Participants

Age range of 16 participants

-8.43 years

Mean Difference for YingCausAge

Largest absolute median difference across biological samples

> 0.8

Spearman Correlation for PC Clocks

Correlation level for most biological samples

Full Text

We can’t show the full text here under this license.

View Full Text ↗

Funding

Competing interests

Competing interests: B.K.K served on the Scientific Advisory Board for Ponce De Leon Health. S.H. and R.B. are co-founder of the non-profit Epigenetic Clock Development Foundation. J.G. is the informatics scientist of the non-profit Epigenetic Clock Development Foundation. Other authors have no conflict of interest to disclose.

PubMed

Differences in DNA methylation age between EPICv1 and EPICv2 arrays in blood and saliva samples

Abstract

Key numbers

Full Text

Funding

Competing interests

You found one interesting study. We’ll send the next 7.

What Lands in Your Inbox Each Week:

Recent issues from the longevity & aging brief

Abstract

Key numbers

Full Text

Funding

Competing interests

Related papers

You found one interesting study. We’ll send the next 7.

What Lands in Your Inbox Each Week:

Recent issues from the longevity & aging brief