Forensic inference in Africa: Evaluating population structure, databases, and regional assignment accuracy

Area of Science:

• Population Genetics
• Forensic Science
• Human Ancestry

Background:

• Autosomal Short Tandem Repeats (aSTR) allele frequencies are crucial for forensic identification.
• Understanding population structure and diversity in Southern Africa is vital for accurate forensic analysis.
• Existing reference databases may not adequately represent diverse African populations.

Purpose of the Study:

• To report novel aSTR allele frequencies for seven Bantu-speaking groups in Southern Africa.
• To evaluate the adequacy of existing forensic reference databases for Southern Bantu populations.
• To assess the accuracy of biogeographical ancestry prediction models within Africa.

Main Methods:

• Collected aSTR allele frequency data from 538 individuals across seven Southern African Bantu-speaking groups.
• Utilized summary statistics, Multidimensional Scaling (MDS), Discriminant Analysis of Principal Components (DAPC), and Bayesian clustering to analyze population structure.
• Employed an XGBoost machine learning model for biogeographical ancestry prediction.

Main Results:

• Established a comprehensive Southern Bantu (SB) reference database with 21 aSTR loci and 11 triallelic profiles.
• Identified four major continental clusters: Southern, Eastern, Western, and Horn of Africa.
• STRidER Africa database showed the closest representation of SB populations; stratification correction is needed for SB populations.
• XGBoost model achieved high predictive balanced accuracy for intracontinental biogeographical assignment (80-94%).

Conclusions:

• The novel aSTR data enhance forensic capabilities in Southern Africa.
• STRidER Africa is the most suitable reference database for SB populations among those tested.
• Biogeographical ancestry prediction models show significant accuracy but require careful consideration of limitations and ethical implications.