Ensemble DNA methylation clock demonstrates Immune-metabolic Aging signatures associated with Mortality

Muthukumar Yugan Gogul, Karthikeyan Vijayakumar, and Gwang-Won Cho present a stacked survival model built from DNA-methylation data in the Framingham Heart Study. Five complementary survival models were fused with a neural-network meta-learner. The model used 190 CpG loci selected with elastic-net Cox regression and was externally validated in postmenopausal women aged 50 to 79 years. The reported performance exceeded PhenoAge and was statistically comparable to GrimAge. The fetched abstract does not give sample sizes, hazard ratios, confidence intervals, or discrimination statistics. This matters because a clock can predict mortality without being a modifiable cause of aging. The study is useful for biomarker monitoring and for understanding immune-metabolic signatures, not for proving that an intervention changes lifespan. The training and validation populations also limit transportability. Both were demographically narrower than the general population, and the abstract specifically limits interpretation to demographically similar groups of European ancestry. A stronger model is not automatically a better treatment endpoint. Any intervention claim would still require prospective trials showing that clock movement tracks functional health, disease events, or survival.

These performance estimates were derived in cohorts of European ancestry and externally validated in postmenopausal women aged 50-79 years, and should therefore be interpreted as applicable only to demographically similar populations.