A robust cross-tissue DNA methylation model for forensic age estimation from oral samples

DNA methylation-based chronological age estimation is a powerful forensic tool, but its application to commonly encountered oral-derived samples (e.g., buccal swabs, saliva) is hampered by tissue specificity and inherent cellular heterogeneity, often leading to inaccurate predictions with existing models. This study aimed to overcome these limitations by developing and validating a robust cross-tissue DNA methylation model for forensic age estimation from such samples. We quantified DNA methylation at 18 CpG sites in 216 paired buccal swab and saliva samples (Han Chinese, 2–83 years) and systematically evaluated 32 model configurations—varying CpG marker panels, age transformation, and tissue variable inclusion—to identify markers with high cross-tissue stability and optimize predictive accuracy. An optimized 10-CpG quantile regression model achieved mean absolute errors (MAEs) of 3.19 years (buccal swabs), 3.44 years (saliva), and 3.45 years (combined dataset) in 10-fold cross-validation. Crucially, this model demonstrated excellent performance on an independent validation set of forensically relevant chewed gum samples (n = 25, aged 19–70 years; MAE = 3.29 years). The model also maintained reliable performance with bisulfite-converted DNA inputs as low as 5 ng and remained stable after 31 days of uncontrolled environmental storage. Our findings establish a methodologically sound and practically validated cross-tissue approach for forensic age estimation from diverse oral samples, offering a reliable solution to the challenges of tissue variability and cellular heterogeneity in real-world casework.