Age estimation of captive Asian elephants (Elephas maximus) based on DNA methylation: An exploratory analysis using methylation-sensitive high-resolution melting (MS-HRM)

# Estimation of captive Asian elephants (Elephas maximus) age based on DNA methylation: An exploratory analysis using methylation-sensitive high-resolution melting (MS-HRM) ## Description of the data and file structure 1. The raw methylation data of *RALYL* and *TET2* used in this analysis are in csv files. This is the dataset that we used to develop the age estimation model. The '**subject ID**' represents the same individuals. In contrast, '**sample**' represents the sample collections taken over time. In addition, cells containing '**n/a**' in our dataset within the 'sample' column are samples which were sampled recently and had no define number ID at the time (please refer to the Supplementary Information on the manuscript for more details on sampling date). '**sex**' represents the sex of the individual, where F: female and M: male. '**age**' represents the chronological age of the individual at the time of sampling. '**ralyl_methylationrate_ave**' and '**tet2_methylationrate_ave**' is the average methylation rate of two duplicate measure of each sample of *RALYL* and *TET2* regions respectively. '**ralyl_df_ave**' and '**tet2_df_ave**' is the Df value of two duplicate measure of each sample of *RALYL* and TET2 regions respectively. '**ralyl_before**', '**tet2_before**', and '**all_svm**' represents the predicted age output of the SVM model, before leave-one-individual-out cross-validation for the *RALYL* model, *TET2* model, and the combined model respectively. '**ralyl_loiocv_after**', '**tet2_loiocv**', and '**all_loiocv**' represents the output of the predicted age after leave-one-individual-out cross-validation for the *RALYL* model, *TET2* model, and the combined model respectively. 2. The R script is a pdf file. This is the R script for the statistical analysis and cross-validations examined in this study. The codes were created by modifying the R script of Qi et al. (2021) accordingly (). Cross-validation used to validate model performance was leave-one-individual-out cross-validation as samples from the same individuals were sampled over time.