To explore the stability of LD Score across European populations, we estimated LD Scores using each of the 1000 Genomes EUR subpopulations separately (Utah Residents with Northern and Western European Ancestry (CEU), British in England and Scotland (GBR), Toscani in Italia (TSI) and Finnish in Finland (FIN)). The LD Scores from all four subpopulations were highly correlated, but mean LD Score increased with latitude (Supplementary Table 8), consistent with the observation that Southern European populations have gone through less severe bottlenecks than Northern European populations9. For example, in comparison to the combined EUR LD Score, the mean LD Score for FIN was 7% larger, and the mean LD Score for TSI was 8% smaller. We evaluated the impact of these differences on the behavior of the LD Score regression analysis and find that the EUR reference panel is adequate for studies in outbred populations of predominantly northern European ancestry, such as European American or UK populations (see Online methods). For other populations, a different reference panel should be used.
