This 2020 Norwegian study investigated current epigenetic clock technology:
“Epigenetic clocks are based on CpGs from the Illumina HumanMethylation450 BeadChip (450 K) which has been replaced by the latest platform, Illumina MethylationEPIC BeadChip (EPIC). EPIC is a major improvement over its predecessor, 450 K (> 450,000 CpGs), in terms of number of probes (> 850,000 CpGs) and genomic coverage of regulatory elements.
The training set of other epigenetic clocks was mostly based on 450 K, except for Horvath Skin & Blood clock which used both 450 K and EPIC-derived DNAm data. Additional CpGs on EPIC do not enhance accuracy or precision of epigenetic clocks when the training set is reduced.
We validated epigenetic clocks in EPIC-derived blood-based DNAm data (n = 470; 305 European women and 165 South Asian women). eABEC showed that epigenetic age acceleration (EAA; residuals from regression of DNAm age on chronological age) was higher in South Asian women than in Norwegian women.
The reason for higher precision is likely due to the large training set (n = 2227) and wide age-span of samples (19 to 88 years for the training set of eABEC).
EPIC probes that are designed to cover regulatory regions did not increase precision. It is difficult to dismiss the possibility that other regulatory CpGs not currently included on EPIC might improve age prediction.”
https://bmcgenomics.biomedcentral.com/articles/10.1186/s12864-020-07168-8 “Blood-based epigenetic estimators of chronological age in human adults using DNA methylation data from the Illumina MethylationEPIC array”
This study’s main point was deficits in current technology. The above graphic demonstrated that epigenetic clocks could do better across different ethnicities.
These researchers repeated a point from An epigenetic clock review by committee about increasing training set size. They missed a point from Do epigenetic clocks measure causes or effects? that:
“The power of these measures as diagnostic and prognostic may stem from the use of longitudinal data in training them. Rather than continuing to train chronological age predictors using diverse data, it may be more advantageous to retrain some of the existing measures by predicting longitudinal outcomes.”
They also didn’t assign much relevance to coverage improvements of The epigenetic clock now includes skin:
“Although the skin-blood clock was derived from significantly less samples (~900) than Horvath’s clock (~8000 samples), it was found to more accurately predict chronological age, not only across fibroblasts and skin, but also across blood, buccal and saliva tissue.”
What I’d like to know about epigenetic clock measurements of biological age is: Why aren’t thousands of studies using them every year? How can we expect continuous improvements in their technologies or coverages or training sets without widespread use?