Inflammation has long been considered one of the main drivers of aging, but its footprint in the epigenome had never been isolated as a separate biomarker. Existing epigenetic clocks, such as GrimAge or Horvath, rely on hundreds of CpG sites selected for predictive power rather than biological meaning. This work stands apart: it deliberately restricts the model to the promoters of just 127 inflammation genes to produce an interpretable tool. Nearby in the archive are papers on cellular senescence, immune passports, and blood analysis, but none of those neighbors directly linked an inflammatory gene set to epigenetic age.
From epigenetic data, the authors selected CpG sites in the promoter regions of 127 genes from an inflammatory signature (ISig). On these sites they trained two ElasticNet models: the first on 656 blood samples from Hannum et al., the second on 6039 samples from the ComputAgeBench training set. Performance was assessed using mean absolute error (MAE) and Pearson correlation. The clocks were then applied to 69 triples of cancer samples, adjacent normal tissue, and healthy controls from the EWAS Data Hub, as well as to fibroblasts after 49 days of OSKM reprogramming.
Both models predicted the age of healthy people accurately: the first with an MAE of 6.84 years and a correlation of 0.85, the second with an MAE of 5.9 years and a correlation of 0.94. In cancer tissues, age acceleration was significantly higher than in adjacent normal tissue, and higher there than in healthy controls, reproducing a gradient from normal to tumor. During OSKM reprogramming of fibroblasts, both models showed a decrease in biological age, that is, rejuvenation. On the independent ComputAgeBench set, the clocks detected accelerated aging in cardiovascular, immune, and musculoskeletal diseases.
Unlike neighboring studies that build clocks from thousands of sites or hunt for blood biomarkers, this work shows that a small, functionally coherent gene set is enough to track age dynamics. That opens the way to clocks focused on specific biological pathways rather than statistical optimization. The link between inflammation and epigenetic aging can now be studied mechanistically, not just correlationally.
The approach can be extended to other gene sets: metabolic, neuronal, immune. GeneClock Studio already allows training clocks on arbitrary gene lists, which will speed up the search for tissue-specific and pathogenetic signatures. We may see clocks sensitive to specific interventions, such as anti-inflammatory therapy.
The work will influence gerontology, oncology, and immunology by offering a tool to assess the inflammatory component of aging. Clinical trials of therapies targeting inflammaging will gain a quantitative biomarker of efficacy.
Test the clocks on longitudinal cohorts to separate cause from effect. Validate them in other tissues and in response to anti-inflammatory drugs.
For gerontology, the key question is whether inflammation is a cause or a marker of aging. These clocks provide a measurable tool but do not answer it. The problem of reversibility of epigenetic changes during rejuvenation also remains.
In the cancer set of 39 tumor types, samples from a 112-year-old woman had to be excluded to avoid skewing the statistics.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.