Figure Legends
Figure 1. Map of Hokkaido, Japan, showing Noboribetsu Bear Park and the Rusha area of the Shiretoko Peninsula.
Figure 2. Scatter plots of age (year) versus DNA methylation level (%) in CpG sites that showed the strongest correlation. This figure includes one of the CpGs adjacent toSLC12A5 (a), POU4F2 (b), VGF (c), and SCGN(d).
Figure 3. Within-individual changes in DNA methylation levels with age. The same CpG sites as in Figure 2 are shown. Dotted line represents the straight line approximated from the samples used for model construction. This figure includes one of the CpGs adjacent to SLC12A5 (a), POU4F2(b), VGF (c), and SCGN (d).
Figure 4. Left: scatter plots of predicted age (year) and chronological age in the age estimation model after LOOCV. Right: scatter plots of predicted age (year) and chronological age for wild bears based on the methylation levels assigned to the left model. Solid line represents predicted age = actual age; distance between the dotted line and solid line represents the MAE of the model after LOOCV; single regression model (a, d), the best elastic net regression model (b, e), and the best SVR model (c, f).
Figure 5. Influences of the interactions among age, sex and growth environment on the model. (a) Scatter plots of Δage (year) and chronological age (year) in the female and male datasets. (b) Scatter plots of Δage (year) and chronological age (year) in the captive and wild datasets. The interaction between age and growth environment was significantly explanatory for Δage. (c) Scatter plots of |Δage| (year) and chronological age (year) in the female and male datasets. (d) Scatter plots of |Δage| (year) and chronological age (year) in the captive and wild datasets.
Figure 6. Influences of sex and growth environment on the model: (a) Δage according to sex: female and male. (b) Δage according to growth environment. (c) |Δage| according to sex. (d) |Δage| according to growth environment. The growth environment was significantly explanatory for |Δage|.