Machine learning-based geographical ancestry inference model for the Han Chinese population

Area of Science:

• Population Genetics
• Genomic Ancestry Analysis
• Machine Learning in Genomics

Background:

• The Han Chinese population displays complex, fine-scale genetic structure with regional variations.
• Existing ancestry inference models lack specificity for Han Chinese genetic diversity.
• Understanding this structure is crucial for evolutionary studies and accurate ancestry determination.

Purpose of the Study:

• To investigate the correlation between genetic variation and geographic distribution in Han Chinese.
• To develop and validate a machine learning model for regional Han Chinese ancestry prediction.
• To provide a robust tool for population and forensic genetics.

Main Methods:

• Analysis of high-density SNP data from 1,229 Han Chinese individuals across eight provinces.
• Application of Principal Component Analysis (PCA) and ADMIXTURE for genetic stratification analysis.
• Training and cross-validation of machine learning classifiers (XGBoost, random forest, k-NN) using PCA components.

Main Results:

• Identified seven distinct genetic clusters within the Han population, correlating with geography.
• The PCA-XGBoost model achieved high prediction accuracy (87.66% top-rank, 96.87% LR-based) in the reference set.
• The PCA-XGBoost model demonstrated excellent generalizability and stability on independent test sets (>85% top-rank, >95% LR-based accuracy).

Conclusions:

• The developed PCA-XGBoost model is efficient, robust, and accurate for regional Han Chinese ancestry prediction.
• This model serves as a reliable methodological tool for population genetics research.
• The findings contribute to a deeper understanding of Han Chinese population structure and biogeographical ancestry.