DNA methylation-based multi-tissue age prediction model for blood, saliva, and buccal swab samples

DNA methylation is regarded as the most promising biomarker for forensic age prediction. However, the vast majority of forensic age prediction tools based on DNA methylation reported so far are tissue-specific, which limits their practical applicability. In our previous study, we successfully developed a cross-tissue 10-CpG quantile regression model for age prediction from saliva and buccal swab samples, achieving a mean absolute error (MAE) of 3.45 years. In order to develop a more widely applicable multi-tissue age predictor, in this study, we further quantified DNA methylation at 18 CpG sites in 216 blood samples (Han Chinese, 1-82 years) using two multiplex methylation SNaPshot assays and systematically evaluated 16 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 constructed with 648 samples including 216 blood, 216 saliva, and 216 buccal swab samples achieved MAEs of 3.32 years (blood) and 3.88 years (combined dataset) in 10-fold cross-validation. Specifically, this model demonstrated excellent performance on an independent validation set of forensically relevant chewed gum samples (n = 25, aged 19-70 years; MAE = 2.82 years). Although its prediction accuracies in each sample type were slightly lower than those obtained from each tissue-specific model, the newly developed multi-tissue model would be extremely useful in forensic analysis, especially when dealing with biological samples of uncertain tissue origin or complex cell compositions.