Background

Massively parallel sequencing (MPS) enables high-resolution detection of mitochondrial DNA (mtDNA) heteroplasmy. To distinguish genuine inherited heteroplasmy from technical artifacts such as nuclear mitochondrial DNA segments (NUMTs), forensic guidelines recommend minor allele frequency (MAF) interpretation thresholds of 5–10%. However, because somatic mtDNA mutations accumulate with age, it remains unclear whether age-associated heteroplasmic divergence could compromise the temporal consistency of forensic identification.

Methods

Deep whole-mtGenome MPS data from an anonymized multi-regional adult cohort (20–60 years) were analyzed. Threshold sensitivity analysis compared heteroplasmy loads between young (≤35 years) and older (≥50 years) groups across multiple MAF cutoffs (1%, 2%, 5%, 10%, and 20%). Multivariate logistic regression adjusting for sex and geographic region was applied to identify independent age-associated heteroplasmic features. Machine-learning models (LightGBM) combined with SHAP (SHapley Additive exPlanations) analysis were used to evaluate the age-discriminative and predictive capacities of heteroplasmy and to characterize patterns of somatic variation.

Results

Age-related divergence in heteroplasmy was strongly dependent on the analytical threshold. At low detection thresholds (1–2% MAF), the elderly cohort showed significantly higher heteroplasmy burdens than the younger cohort (p < 0.01). Crucially, this difference was entirely neutralized at the operational thresholds of 5% and 10% (p = 0.141 and p = 0.957). Adjusted multivariate analysis showed that the age-associated signal was primarily driven by increased frequencies of A>G and C>T transitions and several site-specific hotspots (e.g., position 204), while aggregated heteroplasmy loads in the control and coding regions were not significant. Machine-learning classification demonstrated limited discrimination between age groups (AUC = 0.675; accuracy = 0.682), and regression analysis showed minimal predictive power for chronological age (R² = 0.036; MAE ≈ 8.76 years). SHAP interpretation confirmed that age-related signals were distributed across multiple low-impact features, consistent with stochastic somatic mutation processes.

Conclusion

Age-associated somatic heteroplasmy is largely stochastic, concentrated below 5% MAF, and lacks stable predictive signatures. Consequently, heteroplasmy patterns provide only limited discrimination between age groups and cannot reliably predict chronological age. The recommended 5–10% interpretation thresholds therefore act as an effective biological boundary, successfully filtering age-related somatic noise while preserving stable germline profiles. These findings provide robust empirical support for current forensic mtDNA interpretation guidelines, demonstrating that chronological age does not compromise the reliability of forensic mtDNA MPS analysis.

Authors

• Qi Yang (Institute of Forensic Science, Fudan University; School of Forensic Medicine and Science, Fudan University, China)
• Suhua Zhang (Institute of Forensic Science, Fudan University; School of Forensic Medicine and Science, Fudan University, China)
• Chengtao Li (Institute of Forensic Science, Fudan University; School of Forensic Medicine and Science, Fudan University, China)
• Anqi Chen (Institute of Forensic Science, Fudan University; School of Forensic Medicine and Science, Fudan University, China)

*Corresponding Authors: Anqi Chen, Email: anqi_chen@fudan.edu.cn.