Biogeographical ancestry inference (BGAI) of highly degraded samples remains one of the most formidable challenges in forensic practice. While microhaplotypes (MHs) have demonstrated substantial advantages for analyzing compromised DNA by integrating the merits of SNPs and STRs, the optimal analytical framework to maximize their efficacy in severely degraded contexts remains underexplored. This study aims to bridge this gap by establishing an integrated, artificial intelligence (AI)-driven framework that leverages genome-wide MH loci to maximize genetic information recovery and BGAI accuracy from challenging degraded evidence. Building upon prior genomic evaluations, we curated highly efficient, genome-wide MH marker sets optimized via distinct thresholds for Rosenberg informativeness (In) and genetic differentiation (Gst).  

To tackle severe DNA fragmentation, we adapted bioinformatics pipelines typically utilized for low-coverage ancient DNA (aDNA) whole-genome sequencing (WGS). This low-coverage genomic data was subsequently integrated into a comprehensive AI framework. The analytical pipeline sequentially incorporated principal component analysis (PCA), hierarchical soft clustering (tangleGen), machine learning (ML) classification, and likelihood ratio (LR) estimation to construct a robust ancestry predictive model. Our high-efficiency panel initially yielded 4,484 core MH loci that maintained high discriminative power despite simulated degradation. By applying the aDNA-adapted bioinformatics pipeline, we successfully recovered actionable genetic profiles from ultra-low coverage WGS data (down to 0.2x).  

Crucially, the AI-driven framework significantly outperformed traditional BGAI methods. In this preliminary study, the ML classifier, reinforced by tangleGen soft clustering and LR evaluation, achieved a striking predictive accuracy of 98% and an AUC of 0.999 in highly degraded cohorts. These findings demonstrate that synergizing aDNA bioinformatics with AI and MH markers comprehensively resolves the information-loss bottleneck in compromised samples. The proposed framework provides a robust, highly accurate reference protocol for forensic BGAI in casework involving severe DNA degradation. 

Authors

• Ranran Zhang (School of Forensic Medicine and Science, Fudan University; Institute of Forensic Science, Fudan University, China) 
• Jiao Luo (School of Forensic Medicine and Science, Fudan University, China) 
• Zhiqi Hua (School of Forensic Medicine and Science, Fudan University, China) 
• Pengfei Lu (School of Basic Medicine, Baotou Medical College, China) 
• Chengtao Li (School of Forensic Medicine and Science, Fudan University; Institute of Forensic Science, Fudan University, China) 
• Suhua Zhang (School of Forensic Medicine and Science, Fudan University; Institute of Forensic Science, Fudan University, China)