Explainable AI-driven marker selection and development of EastFocus AISNPplex 61 assay for fine-scale biogeographical ancestry inference

Summary: A hierarchical ancestry inference framework was developed to resolve continental and fine-scale East Asian population structures for forensic applications. Combining the shapley additive explanations (SHAP) method with recursive feature elimination, 58 ancestry informative single-nucleotide polymorphisms (AIM-SNPs) were selected. A self-developed 6-dye allele-specific PCR multiplex system (EastFocus AISNPplex 61) was established, integrating these AIM-SNPs, two Y-chromosomal InDels, and the Amelogenin locus. Validation demonstrated reliable single-source genotyping down to 250–500 pg and full minor contributor recovery in mixed DNA samples up to a 1:7 ratio (125 pg minor input). Population genetic and automated machine learning analyses confirmed that the panel distinguished five continental populations (with category-specific area under the curve (AUC) values ranging from 0.981 to 0.999) and differentiated intra-East Asian structures (with all cluster-specific AUC values exceeding 0.940). Integrating model probabilities with dimensionality reduction enabled two-dimensional spatial mapping to visualize genetic affinities and admixture patterns in studied populations. This workflow provides a practical, interpretable, and capillary electrophoresis-compatible solution for forensic biogeographical ancestry inference and micro-genetic structure mapping.