The cheese PGS signal is probably driven by higher cheese intake in more educated groups in the UK, where the GWAS sumstats come from.

We didn’t analyze personality PGSs, but based on onlinelibrary.wiley.com/doi/full/10.... and www.biorxiv.org/content/10.1..., I would expect weaker but similar patterns for Openness and Neuroticism, and little to no signal for the other traits.

7/7 Many thanks to the participants and the team at the Estonian Biobank @estbiobank.bsky.social, Vasili Pankratov, who led this study, and our amazing coauthors @ukuvainik.bsky.social, Mait Metspalu, Francesco Montinaro and Luca Pagani.

6/7 Caution is needed in interpretation: polygenic scores correlate with many traits, not just a single phenotype, and are influenced by various confounders. Further work is required to identify the specific phenotypic pathways linking genotype to migration decisions.

5/7 Differences in EA PGS between people born in major cities and other regions are already visible from the mid-20th century and increase over time despite social changes (panel A). Movers to Tallinn or Tartu have higher EA PGS than non-migrants even in the earliest observed cohorts (panel B).

4/7 Within-family effects are smaller than between-family effects, but are significant for PGS associated with certain socioeconomic traits (panel A). They disappear when conditioning on EA PGS (panel B).

3/7 The dominant migration pattern driving this structure is movement to Estonia’s two largest cities: Tallinn and Tartu. POB and POR denote place of birth and place of residence.

2/7 Migration creates regional PGS differences even among siblings born in the same region.

1/7 Internal migration intensifies regional differences in multiple polygenic scores (PGS), most strongly for educational attainment (EA, panel B). Adjusting for EA PGS largely removes regional differences in other PGSs (panel C).