3 | Results
Fig 2 shows the standard curve of GRIA2 and CDKN2A and the
value of coefficient “a ” was calculated as 1.151 and 0.7838,
respectively.
The methylation rate of GRIA2 and CDKN2A ranged from
0.32–11.56% and1.32–83.20%, respectivly. The simple linear analysis
showed a significant correlation between age and methylation rate in
both genes (GRIA2 : R2 = 0.17, p =
0.01; CDKN2A : R2 = 0.34, p< 0.01; Fig 3). We examined the changes in methylation rate
over the years for five individuals for which samples were collected
across multiple times (Fig.3). Although overall, the methylation rate of
both genes increased with age, amongst these individuals, the
methylation rate of GRIA2 decreased with age for one individual
(individual number: #592MS) and that of CDKN2A for two
individuals (individual number: #592MS and #604FA).
No sex differences were observed in the correlation between methylation
rate and age in both genes (ANCOVA, GRIA2: p > 0.05,CDKN2A: p > 0.05; Table 2–3). On the other hand,
nursing females showed significantly lower methylation rates (ANCOVA,GRIA2 : p < 0.05, CDKN2A : p< 0.05; Table 2–3).
An age estimation model was developed using the methylation rates of
both genes, GRIA2 and CDKN2A as explanatory variables in
SVR (model 1). The model before LOOCV was performed, exhibited aR2 of 0.74 and a mean absolute error (MAE) of
2.63 years (Fig 4a). After LOOCV, the model exhibited aR2 of 0.33 and an MAE of 5.08 years (Fig 4b).
In comparison to model 2 which included sex as an explanatory variable,
and model 3 with female nursing states as an explanatory variable, model
1 showed the best precision and accuracy of estimation (Table 4).
The resulting model was used to estimate the ages of 19 individuals of
unknown age (Table S2). Among these individuals, we compared the
absolute difference between the age estimated from our model and the age
of first births or weaning for five individuals. The analysis revealed
an MAE of 9.41 years (Range = 21.56–4.58, Table S2).