Abstract
Population labels in genetic studies are often assumed rather than empirically evaluated, a problem especially acute for the Middle East (MID), a heterogeneous region with no consistent population definition. We harmonized 1,911 public genomes from MID-assigned populations and developed visualization approaches that compare predefined labels with observed genetic structure, revealing substantial heterogeneity within MID boundaries, allele frequency differences up to 60% across genetic clusters, and limited overlap with the standard HGDP-MID reference. Guided by these findings, we expanded the HGDP-1kGP panel with 136 published whole-genome sequences from seven unrepresented MID regions. The expanded reference increased MID-like sample sizes by 103% in UK Biobank and 212% in All of Us while maintaining control of population stratification, and enabled detection of GWAS associations not found using the original reference. This framework provides a generalizable strategy for empirically evaluating population definitions and improving genomic analyses of populations poorly captured by existing reference panels.