Aging is a complex process characterized by various phenotypic changes, including epigenetic modifications, inflammation, and metabolic dysregulation. Recent advancements in DNA methylation (DNAm) clocks have provided tools to estimate biological age and predict age-associated outcomes. However, most existing clocks rely on linear models, which may not adequately capture the complexity of survival-related DNAm signatures.
The study conducted by Muthukumar et al. (2026) introduces a novel heterogeneous stacked ensemble survival model that utilizes DNAm data from the Framingham Heart Study. The researchers identified 190 CpG loci through elastic net Cox regression, which were then used to develop a survival prediction model. This model integrates five complementary survival models using a neural network meta-learner.
In an external validation cohort, the performance of this survival model was evaluated, demonstrating strong predictive power for all-cause mortality. Notably, the model’s performance surpassed that of PhenoAge and was statistically comparable to GrimAge. However, it is essential to note that these performance estimates were derived from cohorts of European ancestry and validated in postmenopausal women aged 50-79 years, limiting the applicability of the findings to demographically similar populations.
What this study does not show
This study does not establish clinical utility across diverse populations beyond the specific cohorts studied, nor does it provide direct evidence of health outcomes resulting from the use of the DNA methylation clock.
Evidence scorecard
- Analytic validity: Established
- Clinical validity: Partial
- Clinical utility: Not established
The evidence is based on the primary research conducted by Muthukumar et al. (2026).