Epigenetic age predictors for non-invasive assessment of human skin

ABSTRACT

Skin is both the most visible and most environmentally exposed organ, with apparent aging phenotypes. DNA methylation clocks faithfully capture the progression of aging, but so far have been limited to training on abundant in vitro material or invasively collected samples to generate narrow methylomes using microarray platforms. Here, we demonstrate that skin biological age can be measured directly from a person's face with superior accuracy, using non-invasive tape-stripping. We developed two clocks, MitraSolo, based on single CpGs, and MitraCluster, on regions, trained on the largest enzymatic methyl-sequencing dataset of human epidermis (n = 462). Our models were validated on independent, longitudinal, and external datasets and were compared against established clocks. They predict age accurately, with an error of approximately 4 years, outperforming others on epidermal samples. They maintain high accuracy at low sequencing depths, enabling cost-effective scalability and show intra-individual prediction variation