Age-estimation methods based on DNA methylation hold promise as powerful tools for forensic investigations. Previously validated age-dependent methylation markers in blood DNA have demonstrated sufficient accuracy and precision for practical applications. These estimates are typically generated through regression analysis of pyrosequencing-derived methylation rates. The PyroMark series from Qiagen, the only platform offering pyrosequencing, is scheduled to discontinue all models except the current Q48 Autoprep. The Q48 system incorporates improvements such as automated template preparation and optimized chemistry, which may introduce systematic differences in methylation rates compared to the earlier Q24 and Q96 systems. Therefore, a comparative validation between the Q24/Q96 and Q48 is necessary to effectively utilize the reliability data accumulated with the earlier systems.

In this study, we performed parallel pyrosequencing of blood DNA from 88 individuals using the Q24 and Q48 systems, targeting four established markers (ELOVL2, ASPA, EDARADD, and PDE4C). Comparison of the data revealed a tendency for both peak heights and calculated methylation rates to decrease when measured on the Q48 system for all markers. The largest decline was observed for the EDARADD methylation rate, which weakened the correlation of this marker with chronological age, whereas the other markers retained their correlations with age.

When the obtained methylation rates were inserted into the published regression equations, distinct patterns emerged. Equation 1 (Bekaert et al., 2015), incorporating all four markers, generated age estimates that were on average about 4 years higher on the Q48 than on the Q24. In contrast, Equations 2 (Garali et al., 2020) and 3 (Yamagishi et al., 2024), which rely solely on ELOVL2, yielded estimates that were approximately 2 years lower on the Q48. Across all three regressions, mean absolute error (MAE) and root mean square error (RMSE) increased when using Q48 data, reflecting these equations were optimized for the Q24/Q96 outputs.

To mitigate these discrepancies, we propose two potential solutions: (1) developing additional conversion regressors to translate Q48 methylation rates to the Q24 scale, or (2) generating new age‑estimation equations based on accumulating Q48 data in the future. These adjustments will be crucial for preserving the forensic utility of methylation‑based age estimation during the transition to the Q48 system.

Authors

• Takayuki Yamagishi (National Research Institute of Police Science, Japan, Japan)
• Takuya Shimura (National Research Institute of Police Science, Japan, Japan)
• Kochi Toyomane (National Research Institute of Police Science, Japan, Japan)
• Ken Watanabe (National Research Institute of Police Science, Japan, Japan)
• Tomoko Akutsu (National Research Institute of Police Science, Japan, Japan)