Back

Dental aging offers new insights to the first epigenetic clock for common dolphins (Delphinus delphis)

Hanninger, E.-M. F.; Peters, K. J.; Gerber, L.; Barratclough, A.; Betty, E. L.; Palmer, E. I.; Horvath, S.; Stockin, K. A.

2025-07-23 genetics
10.1101/2025.07.20.665818 bioRxiv
Show abstract

Determining exact age in wild odontocetes is essential for understanding population dynamics, survival, and reproduction, yet remains logistically challenging. While epigenetic aging is emerging as a valuable approach, only nine species-specific clocks currently exist. Most have been calibrated using captive known-age animals or well-studied wild populations. Only two previous studies have used dental ages from stranded or bycaught individuals. This is due to concerns that dental age inaccuracies, especially in older animals, may affect epigenetic clock performance. To explore this, we developed the first species-specific epigenetic clock for common dolphins (Delphinus delphis), analysing DNA methylation at 37,492 cytosine-phosphate-guanine sites in skin samples from stranded and bycaught dolphins with estimated dental ages. Elastic net models with Leave-One-Out Cross-Validation were applied to three subsets: the relaxed subset (all individuals; n = 75, median absolute error (MAE) = 2.02, r = 0.81, R2 = 0.66), the strict subset (excluding individuals with minimum dental age estimates only; n = 73, MAE = 2.29, r = 0.81, R2 = 0.66), and the restricted subset (excluding outliers with prediction errors > 6 years; n = 63, MAE = 1.80, r = 0.91, R2 = 0.82) to compare performance. Our models consistently underestimated the age of dolphins >16 years, even when minimum dental ages were applied, suggesting absolute errors between dental and epigenetic estimates unlikely reflect dental aging error. Additionally, post-mortem decomposition condition code (DCC 1 to 3) did not affect age prediction, signalling promise for future epigenetic clocks calibrated with strandings and bycaught individuals.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.