Epigenetic aging models hold great promise for enhancing many aspects of wildlife research and management. However, their utility is limited by the need to train models using known-aged animals, which are rare among wildlife species. We present a novel approach to developing methylation-based age prediction models that enables us to train models using samples from individuals whose chronological age is estimated with uncertainty based on photo-identification catalogue data. Our approach incorporates this uncertainty into model training by representing the age of each individual with a probability distribution rather than a point estimate. We similarly represent the methylation profiles of individuals as binomial distributions and produce a distribution of predicted age for each sample that reflects the uncertainty in both its age and methylation profile. We compared age models trained using a wide range of parameterisations, training data sets and analytical methods to determine how well they predicted the catalogue-based age estimates. The resulting model has a median absolute error of 1.70 years, outperforming many published clocks trained with known-age samples. This approach significantly expands the range of species for which accurate methylation-based age models can be developed, particularly those of conservation concern where known-age samples are limited. By producing distributions of predicted age, it also enables researchers to accurately communicate the uncertainty in their age estimates to subsequent data users.
Citation:
Martien, K.K., R.W. Baird, K.M. Robertson, M.A. Kratofil, S.D. Mahaffy, K.L. West, S.J. Chivers, and F.I. Archer. 2026. Epigenetic Age Estimation for Hawaiian False Killer Whales (Pseudorca crassidens) in the Absence of ‘Known-Age’ Individuals. Molecular Ecology Resources 26(2): e70099. doi: 10.1111/1755-0998.70099
Link:
https://onlinelibrary.wiley.com/doi/10.1111/1755-0998.70099