Data from: Epigenetic signatures of ageing in Asian elephants revealed by reduced representation bisulphite sequencing
Data files
Apr 15, 2026 version files 317.41 MB
-
all_methratios_imputedall.txt
104.06 MB
-
all_methratios_imputedwCpGsites.txt
107.13 MB
-
merged_methratio.txt
106.22 MB
-
README.md
1.97 KB
Abstract
Accurate age estimation is essential for understanding life-history variation, modelling population dynamics, and informing conservation strategies, yet remains challenging for long-lived species. Here, we developed a genome-wide, DNA methylation-based epigenetic clock for Asian elephants (Elephas maximus), an endangered species, using reduced representation bisulphite sequencing (RRBS). Genome-wide methylation profiles were generated from 91 blood samples, yielding 144,611 candidate CpG sites, of which 389 CpG sites were identified as strongly associated with chronological age. The final model predicted age with high accuracy (r = 0.96, MAE = 4.82 years), corresponding to a relative error of 6.06 %, comparable to epigenetic clocks developed for humans and other non-model species. Longitudinal analyses revealed heterogeneous epigenetic ageing trajectories, with most individuals showing increases in epigenetic age over time, while others exhibited relatively neutral or decelerated trajectories, indicating potential inter-individual variation in ageing dynamics. Functional enrichment analyses revealed that age-associated CpGs were enriched in genes and pathways related to development, neurogenesis, metabolism, and social or physiological regulation, including oxytocin, apelin, and melanogenesis signalling. Age-related methylation changes were characterised by predominant hypermethylation in CpG islands and genic regions, consistent with patterns reported across mammals. Together, these findings demonstrate that epigenetic clocks capture biologically meaningful features of ageing linked to life-history traits. As a long-lived mammal with extended reproductive capacity and complex social systems, the Asian elephant represents an informative comparative system for investigating the molecular architecture of ageing. This study contributes to the growing field of comparative epigenetic ageing and highlights the potential of DNA methylation–based approaches to inform evolutionary and conservation-relevant research when applied in appropriate ecological and methodological contexts.
Dataset DOI: 10.5061/dryad.ksn02v7kp
Description of the data and file structure
- merged_methratio.txt
Data filtered CpG methylation ratios (methylated reads divided by total reads) across all samples and chromosomes, where CpG sites were retained based on mean methylation (0.1–0.9) and average coverage (>5). This matrix includes missing values for sites not detected in all samples. Depending on downstream handling of methylation values and the desired use of these data, these cases can be excluded, ignored, or one can simply set "effective CT count" to the methylated count for those small subsets of sites.
- all_methratios_imputedall.txt
To prepare the data for downstream modelling, CpG sites with excessive missingness (>5 missing samples) were removed, and the remaining missing values were imputed using k-nearest neighbours (KNN) implemented in the impute R package. This is the resulting data, which contains a complete methylation matrix without missing values (CpG sites as row names).
- all_methratios_imputedwCpGsites.txt
Provides the same imputed data with CpG site identifiers retained as an explicit column for annotation and interpretation. This is the main data used to build the age estimation model in our paper.
Code/software
R version 4.3.1 was used.
2026_Arai_filtering_Rscript: This script processes CpG methylation data by calculating methylation ratios, filtering sites based on quality (coverage and variability), merging chromosome data, and imputing missing values to prepare a clean dataset for modeling.
2026_Arai_model_Rscript: This script builds an epigenetic age prediction model by selecting highly age-correlated CpG sites, normalizing the data, and training/testing an elastic net regression with cross-validation to optimize accuracy.
