2025-06-05 14:00:00 -0400 EDT
New work led by Joel Mefford demonstrating that polygenic score accuracy varies across the phenotypic distribution.
Beyond predictive R-squared: Quantile regression and non-equivalence tests reveal complex relationships of traits and polygenic scores
We quantify differences in the predictive value of polygenic scores (PGSs) across the phenotypic range using quantile regression and non-equivalence tests of quantile-specific effect sizes. In analyses from the UK Biobank, we show that a PGS's predictive accuracy depends on the quantile of the phenotypic distribution to which the PGS is being applied. Among 25 traits studied, only three showed consistent predictive value across the entire phenotypic range. The other 22 traits exhibited effect sizes that varied by at least 20% from the standard average estimate. The findings suggest that gene-by-environment interactions likely explain much of this heterogeneity.