GlobalFiler™, one of the most widely used autosomal STR (aSTR) multiplexes in forensic casework, is routinely applied across Africa despite the limited availability of population-specific reference databases for the continent. At present, only a small number of African countries have published allele frequency data for the 21-locus GlobalFiler system, raising important questions about the adequacy of reference databases and the robustness of forensic statistical inference in regions characterized by exceptional genetic diversity.

We defined a GlobalFiler allele frequency database for Bantu-speaking populations from Southern Africa (Ndebele, Pedi, Phuthi, Tsonga, Sotho, Swati, and Xhosa), with STRidER accession numbers STR000439 (South Africa) and STR000442 (Lesotho).

The Southern Bantu (SB) database was evaluated across multiple forensic and population genetic analyses: assessment of genetic diversity, population structure, suitability of alternative reference databases, and intracontinental biogeographic ancestry inference. 

DAPC and STRUCTURE analyses identified four major African genetic clusters corresponding to Southern, Eastern, Western, and Horn of Africa regions. 

Subsequently we investigated (1) the adequacy of alternative reference databases and (2) intracontinental biogeographic inference. Alternative databases were inspected by reciprocal allelic coverage, Kullback–Leibler divergence and shifts in random match probabilities (RMPs).  Among these, STRidER Africa provided the closest available representation of SB populations, while intra-population analyses within the SB dataset indicated the need for a stratification correction (θ = 0.005). Intracontinental biogeographic ancestry inference was explored using an XGBoost machine-learning classification model across four African regions, achieving balanced accuracies ranging from 80% to 94%.
These results highlight both the potential and limitations of applying GlobalFiler data beyond their original population contexts and emphasize the critical need for regionally representative allele frequency databases to ensure accurate and equitable forensic inference.

Authors

• Maria Eugenia D'Amato (University of the Western Cape, South Africa)
• Mohaimin Kasu (University of the Western Cape, South Africa)
• Humphrey Brydon (University of the Western Cape, South Africa)