Population differentiation of polygenic score predictions under stabilizing selection

Abstract

Given the many loci uncovered by genome-wide association studies (GWAS), polygenic scores have become central to the drive for genomic medicine and have spread into various areas including evolutionary studies of adaptation. While promising, these scores are fraught with issues of portability across populations, due to the misestimation of effect sizes and missing causal loci across populations not represented in large-scale GWAS. The poor portability of polygenic scores at first seems at odds with the view that much of common genetic variation is shared among populations (Lewontin 1972). Here we investigate one potential cause of this discrepancy: phenotypic stabilizing selection drives the turnover of genetic variation shared between populations at causal loci. Somewhat counter-intuitively, while stabilizing selection to the same optimum phenotype leads to lower phenotypic differentiation among populations, it increases genetic differentiation at GWAS loci and reduces the portability of polygenic scores constructed for unrepresented populations. We also find that stabilizing selection can lead to potentially misleading signals of the differentiation of average polygenic scores among populations. We extend our baseline model to investigate the impact of pleiotropy, gene-by-environment interactions, and directional selection on polygenic score predictions. Our work emphasizes stabilizing selection as a null evolutionary model to understand patterns of allele frequency differentiation and its impact on polygenic score portability and differentiation.

The background and legacy of Lewontin's apportionment of human genetic diversity

“…That said, elevated differentiation between human groups in genetic variation underlying traits can arise for several reasons: if the loci underlying a trait have experienced geographically variable selective pressures (as seen with skin pigmentation loci), or less obviously, in various cases where there is stabilizing selection on a trait, even when there is a single optimum shared between species (e.g. [100,101]). Overall, the degree to which variable selective pressures have acted on variants underlying human traits is still unknown and is often difficult to understand rigorously using techniques developed to date [101–103].…”

Lottery, luck, or legacy. A review of “The Genetic Lottery: Why DNA matters for social equality”

“…To date, much of the discussion in human genetics about using PGS across ancestry groups has focused on the methodological limitations of existing PGS, which are based predominantly on studies of individuals of recent European ancestry (Martin et al., 2019; Mills & Rahal, 2020). Since ancestry groups differ in linkage disequilibrium patterns and allele frequencies, these factors alone lead PGS to be increasingly poor predictors of phenotypic differences in more distantly related ancestries, and they are compounded by differences in environmental effects and gene‐by‐environment interactions (Harpak & Przeworski, 2021; Martin et al., 2019; Mostafavi et al., 2020; Privé et al., 2022; Wang et al., 2020; Yair & Coop, 2021). As Harden points out, however, these are, at least in principle, surmountable difficulties, such that she “anticipate[s] a future in which scientists will have developed a polygenic score that is as strongly related, statistically, to academic achievement in Black students as it is in White Students” [p. 191].…”

Population differentiation of polygenic score predictions under stabilizing selection

“…Code used to generate simulations and process output can be found at https://github.com/SivanYair/SLiMsims_Stabili-zingSelection. The data are provided in electronic supplementary material [118].…”