But you shouldn’t use only 2 PCs to compute the distance 😉

Did you make it? 😁

Yes, but the sample sizes are quite small there.

Thanks!

I actually use a few people from the 1000G as well, to get a few more people from Finland, Japan, and South America.

In the future I would like to also add the Papuan from the SGDP.

✅ In this paper, I used the UK Biobank to define 18 reference worldwide populations, and provided a method to easily infer ancestry proportions of individuals from your study more accurately.

Please have it a try!

Paper: doi.org/10.1093/bioi...
Tutorial: privefl.github.io/bigsnpr/arti...

Not sure. I think someone provided some VCF -> bed -> BGEN conversion.
Not sure what is this DRAGEN.

Did they update the BGEN files they provide?
Initially they were missing 30% of variants.

Potential Main or Co-Supervisors include:

Naomi Wray (Oxford), Peter Visscher (Oxford), Erin Gabriel (KU), Preben B Mortensen (AU), @eagerbo.bsky.social
(AU), @privefl.bsky.social (AU), Thomas Gerds (KU), Julie Dreier (AU), and others.

Genetic PCs rarely provide large % of variation, but I don’t think it means they’re not useful.

Try to look at the loadings if possible.