A PYTHON TOOL FOR INTERROGATING POTENTIAL CONFOUNDING OF GWAS RESULTS OWING TO POPULATION STRATIFICATION

Abstract

This project develops a Python-based tool to detect residual population stratification in genome-wide association study (GWAS) summary statistics using publicly available reference data from the 1000 Genomes Project. The tool implements two complementary functions: a heatmap that visualises potential directional bias in effect sizes across allele frequency difference bins between GBR and TSI populations, and a regression framework that quantifies the variance in GWAS effect sizes explained by principal component loadings derived from LD-pruned reference genotypes. Applied to a demonstration GWAS of adult height, the pipeline reveals ancestry-related structure detectable even after standard PCA adjustment and provides a rapid, reproducible layer of post-GWAS quality control to support more robust and equitable genetic association analyses in health data science.