Samuel Pattillo Smith

Error rates in Q_ST--F_ST comparisons depend on genetic architecture and estimation procedures Genetic and phenotypic variation among populations is one of the fundamental subjects of evolutionary genetics. One question that arises often in data on natural populations is whether differentiation...

New work from my lab, led by graduate student
Janis Liu. Janis studied type I error rates in Qst/Fst comparisons (1/n)
www.biorxiv.org/content/10.1...

We thank co-authors @oliviarxiv.bsky.social @hakha.bsky.social DanDan Peng and @jeremyjberg.bsky.social; also @gcbias.bsky.social for guidance and advice in developing this approach.

Finally, big thanks to some very generous colleagues for their feedback; we’d love to get yours as well! [n/n]

PGSUS is a demonstration that family-based designs can be useful here: Combining their articulation with the statistical power of population-based designs may pave the way forward in the interpretation and application of genomic predictors. [18/n]

…given how they integrate many small statistical associations with subtle potential biases. Even now, it feels like we are only scratching the surface! We need tools to better interpret genomic predictors.

This preprint is a culmination of over six years of work developing this approach. Following ideas we started developing in Mostafavi, Harpak et al. we became increasingly interested in what is baked into genomic predictors like polygenic scores… [16/n]

We were also able to see that different approaches for adjustment for population structure in GWASs (e.g., PCs as fixed effect covariates, LMMs) have distinct advantages with respect to mitigation of ancestry-axis-specific and isotropic SAD variance in PGS. [15/n]

In some instances, a given PGS appears to be stratified along a major axis of ancestry in one prediction sample but not in another (for example, in comparisons of prediction in samples from different countries, or in ancient DNA vs.~contemporary samples). [14/n]

[13/n] We also found evidence of stratification and isotropic inflation in PGSs constructed using the UK Biobank.

Applying PGSUS, we found evidence of stratification in PGSs constructed using large meta-analyses of height and educational attainment. [12/n]

In particular, PGSUS can detect stratification along major axes of ancestry as well as SAD variance that is “isotropic'' with respect to axes of ancestry. [11/n]

Our method, Partitioning Genetic Scores Using Siblings (PGSUS, pronounced ``Pegasus''), breaks down variance components further by axes of genetic ancestry, allowing for a nuanced interpretation of SAD effects.

github.com/harpak-lab/P... [10/n]

We leverage a comparison of a PGS of interest based on a standard GWAS with a PGS based on a sibling GWAS---which is largely immune to SAD effects---to quantify the relative contribution of each type of effect to variance in the PGS of interest. [9/n]

We developed a method that estimates the proportion of variance in a PGS (in a given sample) that is driven by direct effects, SAD effects, and their covariance. [7/n]

Our interpretation and application of PGSs may hinge on the relative impact of SAD effects, since they may often be environmentally or culturally mediated. [6/n]

However, because PGSs are constructed from population-level associations, they are influenced by factors other than direct genetic effects, including Stratification, Assortative mating, and Dynastic effects (“SAD effects'').

www.science.org/doi/10.1126/... [5/n]

PGS are often thought of as capturing the direct, causal genetic effect of one's genotype on their phenotype. [4/n]

Following these observations, attention has turned toward the construction of genomic predictors of traits, so-called “polygenic scores” (PGSs). [3/n]

An Expanded View of Complex Traits: From Polygenic to Omnigenic Many complex genetic traits arise from large numbers of variants, each with small effects. This Perspective argues that risk is ultimately driven by an even larger number of genes with no direct impac...

GWAS have revealed that the genetic basis of variation for many health conditions and other traits is highly polygenic, and that the joint effect of these variants is often well-captured by a simple linear combination, consistent with longstanding theoretical predictions. [2/n]

Think of a polygenic score you care about. Are direct genetic effects driving variation among people in this predictor? Or perhaps other, confounding factors? We at the @arbelharpak.bsky.social & @docedge.bsky.social Labs developed a method to tackle this question. [1/n]

[n/n] We thank @gcbias.bsky.social for guidance and advice in developing this approach. We also thank generous colleagues for their input and feedback; we’d love to get yours as well!

[18/n] PGSUS is a demonstration that family-based designs can be useful here: Combining their articulation with the statistical power of population-based designs may pave the way forward in the interpretation and application of genomic predictors.

[17/n] ...given how they integrate many small statistical associations with subtle potential biases. Even now, it feels like we are only scratching the surface! We need tools to better interpret genomic predictors.

[16/n] This preprint is a culmination of over six years of work developing this approach. Following ideas we started developing in Mostafavi, Harpak et al. we became increasingly interested in what is baked into genomic predictors like polygenic scores...

elifesciences.org/articles/483...

[14/n] In some instances, a given PGS appears to be stratified along a major axis of ancestry in one prediction sample but not in another (for example, in comparisons of prediction in samples from different countries, or in ancient DNA vs.~contemporary samples).

[14/n] In some instances, a given PGS appears to be stratified along a major axis of ancestry in one prediction sample but not in another (for example, in comparisons of prediction in samples from different countries, or in ancient DNA vs.~contemporary samples).

[13/n] We also found evidence of stratification and isotropic inflation in PGSs constructed using the UK Biobank.

[12/n] Applying PGSUS, we found evidence of stratification in PGSs constructed using large meta-analyses of height and educational attainment.

[11/n] In particular, PGSUS can detect stratification along major axes of ancestry as well as SAD variance that is “isotropic'' with respect to axes of ancestry.

[10/n] Our method, Partitioning Genetic Scores Using Siblings (PGSUS, pronounced ``Pegasus''), breaks down variance components further by axes of genetic ancestry, allowing for a nuanced interpretation of SAD effects.

github.com/harpak-lab/P...

[9/n] We leverage a comparison of a PGS of interest based on a standard GWAS with a PGS based on a sibling GWAS---which is largely immune to SAD effects---to quantify the relative contribution of each type of effect to variance in the PGS of interest.