Epigenetic clocks estimate biological age from DNA methylation, and their divergence from passport age (age acceleration) is linked to disease risk, including cancer. However, previous studies were often limited to a single cancer type or a single clock, which prevented a broader picture. This work fills that gap by comparing seven clocks across eight cancer types within a unified framework, and it sits alongside research connecting epigenetics to oncology, but stands out for its scale and systematic approach.
The authors downloaded DNA methylation data from the TCGA database for 5528 samples, including tumor and normal tissues from eight cancer types. For each sample, they calculated epigenetic age using seven clocks: Horvath, Hannum, Levine (PhenoAge), BNN, Skin+Blood, BLUP, and EN. Age acceleration was defined as the difference between predicted and chronological age. To rule out the influence of age distribution, they performed subsampling, equalizing the proportions of tumors and normal samples across ten-year age groups. Then, for each CpG site from the clocks, they assessed its ability to distinguish tumor from normal using logistic regression, and to compare the informativeness of CpG sets from different clocks, they used elastic net classifiers with cross-validation.
In 44 of 56 combinations (8 cancer types × 7 clocks), tumors had higher age acceleration than normal tissues. The exception was uterine cancer, where normal tissues appeared "older" than tumors in six of the seven clocks. The distribution of age acceleration in tumors was wider and flatter than in normal tissue. Analysis of CpG sites revealed that many of those used by the clocks differ significantly in methylation between tumor and normal, with hypermethylation predominating. Some CpG sites, for example those linked to the genes DDR1, TBX5, RXFP3, NDRG2, TBR1, were consistently altered across all tissues and matched known epigenetic disruptions in cancer. Classifiers based on just 10–20 top CpG sites achieved an area under the ROC curve above 0.9 for most solid tumors, while blood and bone marrow cancers required more sites.
The work shows that age acceleration is a common feature of cancer, but with tissue-specific exceptions, which matters for biomarker development. Unlike many neighboring studies that focused on a single cancer type or a single clock, this one provides a systematic comparison of seven clocks across eight tissues, allowing an assessment of the universality and limitations of epigenetic clocks in oncology. The identified CpG sites that consistently distinguish tumor from normal could serve as the basis for minimal diagnostic panels.
The topic could move toward validating the found CpG sites in independent cohorts and other cancer types, as well as studying the mechanisms behind why uterine cancer behaves differently. It may be possible to create specialized epigenetic clocks optimized for specific tissues or clinical tasks. Research into how age acceleration changes during treatment and whether it is linked to prognosis is also promising.
The results could influence oncology, the epidemiology of aging, and the development of diagnostic tests based on DNA methylation. They are also useful for gerontology, since they link molecular mechanisms of aging to carcinogenesis.
The next step could be validating the discriminatory CpG sites on independent datasets and studying their functional role in tumor progression. It is also important to investigate the causal relationship between age acceleration and cancer.
The work directly addresses an unresolved problem in cancer epigenetics: why some tissues show accelerated aging in tumors while others do not, and which methylation changes are drivers versus passengers. It also touches on whether age acceleration can be used as a universal biomarker of cancer risk or whether it is tissue-specific.
The Horvath clock, created in 2013, uses just 353 CpG sites, yet it works in almost all human tissues, including brain, liver, and lungs.
In the series "Altered Carbon," biological age and body condition can be changed, and in reality epigenetic clocks allow us to estimate how much cells are "older" than their passport age, which vaguely resembles the idea of biological monitoring from science fiction.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.