Aging out of the blue: Region-specific epigenetic clock calibration for a blue zone with the DNAm SuperLearner
Epigenetic clocks estimate biological age from DNA methylation patterns at CpG sites, providing robust predictions of mortality and morbidity risk. "Blue zones," regions of exceptional longevity, offer a unique opportunity to investigate how biological aging diverges from chronological age. However, standard clocks are typically trained on large, heterogeneous datasets, reflecting average population trends rather than region-specific dynamics. Using data from the Costa Rican Longevity and Healthy Aging Study (CRELES), we profiled DNA methylation from residents of the Nicoya blue zone (n = 206) and a reference population in other parts of Costa Rica (n = 875). We propose training a SuperLearner, an ensemble machine learning approach, on the non-Nicoyan Costa Ricans to optimize predictive performance across existing clocks and flexible machine learners. By its Oracle property, the SuperLearner's weighted combination of algorithms performs asymptotically as well as the best possible combination of the candidate predictors. We then used this trained model to construct a calibrated hypothesis test comparing Epigenetic Age Acceleration (EAA) distributions between the blue zone region and the reference population. Before calibration or covariate adjustment, estimates for Nicoyans depended heavily on the chosen clock. The SuperLearner placed them about three years younger on average (R¯1=-3.05,95% CI [-3.64,-2.46]), whereas DNAmFitAge placed them almost four years older (R¯1=3.59,95% CI [3.02,4.17]). Mean calibration to non-Nicoyan Costa Ricans made estimates consistent across clocks, with an estimated EAA advantage to approximately two years (Δ^=-1.96,95% CI [-2.56,-1.37]). After age-stratified calibration within 10-year age groups, only a small EAA advantage (Δ^=-0.35,95% CI [-0.64,-0.05]) remained, concentrated among the oldest participants. This approach provides a robust and reproducible framework for estimating longevity in distinct regions when a relevant reference population is available.
Authors
- Alan Hubbard (ORCID: https://orcid.org/0000-0002-3769-0127)
- Julia L. MacIsaac
- Nicole Gladish (ORCID: https://orcid.org/0000-0001-6039-7001)
- William H. Dow (ORCID: https://orcid.org/0000-0002-4080-1668)
- Luis Rosero‐Bixby (ORCID: https://orcid.org/0000-0002-3063-3111)
- Andrés Cárdenas (ORCID: https://orcid.org/0000-0003-2284-3298)
- Kristy Dever (ORCID: https://orcid.org/0000-0003-3101-0685)
- David H. Rehkopf (ORCID: https://orcid.org/0000-0002-7597-6513)
- Michael Steffen Kobor (ORCID: https://orcid.org/0000-0003-4140-1743)
- Nolan David Gunter (ORCID: https://orcid.org/0009-0009-8113-4409)
Institutions
- University of British Columbia (CA)
- Universidad de Costa Rica (CR)
- University of California, Berkeley (US)
- Stanford University (US)
Publication Details
- Journal
- PLoS Computational Biology
- Published
- 2026-11-18
- DOI
- https://doi.org/10.1371/journal.pcbi.1014949
- Citations
- 1
- Primary Topic
- Epigenetics and DNA Methylation
- Type
- article
- Field-Weighted Citation Impact
- 1.12
Funders
- National Institute of Environmental Health Sciences