Excited to share our preprint on PGI portability across ancestries—now on bioRxiv. With co-authors @aysuo.bsky.social, @paturley.bsky.social, @alextisyoung.bsky.social, and @Dan_J_Benjamin. Preprint: doi.org/10.1101/2025... Thread below for details.

In the next few weeks, we will start sharing the PGIs with the participating datasets, who will make them available as read-to-use variables through their own data-sharing procedures. Stay tuned! (9/9)

We have updated the repository’s User Guide to highlight interpretational issues we view as especially important:
www.biorxiv.org/content/bior...
We also provide updated FAQs that provide information on the Repository and PGIs:
www.biorxiv.org/content/bior... (8/9)

It is essential to recall that “causal” does not mean “immutable.” Genetic influences often operate through modifiable environmental pathways. To avoid overly deterministic interpretations, PGI associations should be analyzed and presented with appropriate nuance. (7/9)

We provide an analytic correction for these two sources of attenuation that enhances interpretability and cross-study comparability. (6/9)

We also develop a theoretical framework clarifying how to interpret coefficients from regressions that control for parental PGIs. We identify two distinct sources of attenuation: sampling error in the SNP weights, and bias due to divergence between the between- and within-family GWAS estimands.(5/9)

In 11 datasets that include first-degree relatives, and we infer missing parental PGIs using Mendelian imputation. We include these imputed parental PGIs in the new data release, increasing the number of datasets where parental PGIs can be controlled for in analyses to identify causal effects (4/9)

We now obtain the SNP weights using SBayesR (as opposed to LDpred) using a larger set of SNPs (~2.9 mil as opposed to ~1.2mil HapMap3 SNPs) resulting in gains in predictive power for most phenotypes. (3/9)

For 16 phenotypes from the first release, we integrate new GWAS data into the meta-analyses. For 45 out of 61 phenotypes, the meta-analyses include 23andMe GWAS, improving predictive power compared to what can be obtained with publicly available summary statistics. (2/9)

The repository now includes PGIs for 61 phenotypes (up from 47), spanning additional health, behavioral, and demographic traits. Participating datasets are up from 11 to 20. (1/9)

PGI Repository v2.0 preprint out! A 🧵 on the main results and updates @robel-alemu.bsky.social @paturley.bsky.social @alextisyoung.bsky.social

Preprint just out about the genetically heterogenous effects of an early childhood intervention in Colombia. Key findings below. Come see my talk at #ESSGN Bristol if you wanna hear more!

A great opportunity to work with @alextisyoung.bsky.social if you're interested in methods development!

So happy to finally see this paper out in @natcomms.nature.com. rdcu.be/ehQsd “Correcting for volunteer bias in GWAS increases SNP effect sizes and heritability estimates”. A thread on our findings!

🔊Next conference of the European Social Science Genetics Network (ESSGN) in Bristol, 22-23 May.

✅Keynote by @michellemeyer.bsky.social

✅Panel on the current frontier of MR w/ George Davey-Smith (Bristol), Dalton Conley (Princeton), and Alice Carter (Novo-Nordisk).

essgn.org/essgn-confer...

In the sense that an exogenous shock to the environment is found have heterogenous effects, yes, but that's about it :) We found, in Sweden, that among women from high SES families, an increase in compulsory schooling had a larger effect on the adult earnings of those with lower EA PGI