<![CDATA[Introduction: The capacity to predict an individual's age from biological evidence constitutes a significant advancement in forensic intelligence. DNA methylation, a stable epigenetic mark, provides a molecular basis for "epigenetic clocks." However, the operational reliability of these clocks necessitates rigorous validation across diverse biological samples and populations, particularly for challenging, low-template touch DNA evidence. Methods: Following approval from the Ethical Committee of CMHC Indonesia (No. 128/EC/CMHC/2023), we recruited 150 healthy Indonesian male volunteers aged 18-65. Semen, saliva, and high-yield standardized touch DNA samples were collected. DNA was extracted, quantified fluorometrically, and subjected to bisulfite conversion with efficiency controls. The methylation levels of a curated five-CpG panel (ELOVL2, FHL2, TRIM59, KCNQ1DN, C1orf132) were quantified using a rigorously controlled pyrosequencing workflow. Body-fluid-specific age prediction models were developed using multiple linear regression, validated with 10-fold cross-validation, and assessed for statistical assumptions including multicollinearity. Results: The models for semen and saliva demonstrated high predictive accuracy, yielding Mean Absolute Deviation (MAD) values of 3.19 years (R²=0.94) and 3.55 years (R²=0.92), respectively. The model developed from high-yield touch DNA was less precise but still highly informative, with a MAD of 5.49 years (R²=0.85). All models satisfied the assumptions of linear regression, with Variance Inflation Factors below 2.5 indicating low multicollinearity. The 95% prediction intervals were narrowest for semen, reflecting its superior precision. Conclusion: This study validates a robust, targeted epigenetic panel for age prediction in a Southeast Asian population. We present highly accurate, tissue-specific models for semen and saliva, suitable for immediate consideration in forensic casework. The touch DNA model, while requiring cautious interpretation, provides a valuable framework for generating investigative leads from trace evidence. Our findings underscore the critical importance of tissue-specific modeling and provide a detailed methodological and statistical blueprint for the responsible implementation of forensic age estimation.]]>
