Converging evidence of positive selection at height-associated loci in Europe

Valentin Hivert, Pierrick Wainschtein, Matthew R. Robinson, Evan K. Irving-Pease, Eske Willerslev, Michael E. Goddard, William Barrie, J. Sidorenko, Peter M. Visscher, Loic YengobioRxiv 2026 · DOI

TL;DR

This paper provides strong evidence that natural selection has played a significant role in shaping human height differences across Europe. By using advanced genetic analysis techniques that account for population structure and ancient admixture, the researchers identified specific genetic regions associated with height that show clear signs of recent positive selection.

What Problem Does This Paper Address?

Human height exhibits a noticeable north-south gradient across Europe, with individuals in northern regions generally being taller. This observation has long led to hypotheses about natural selection acting on genes influencing height, particularly in response to colder climates (aligning with Bergmann's rule). However, previous studies attempting to detect this selection using genome-wide association studies (GWAS) have been hampered by confounding factors, primarily population stratification. This means that observed genetic differences might be due to historical population movements and mixing rather than direct selection. This paper aims to resolve this by employing more robust methods to disentangle the effects of selection from population structure and ancient admixture.

Methods

The researchers employed several sophisticated genetic analysis techniques:

  • Trait-specific Singleton Density Score (tSDS): This method is designed to detect recent selection (within the last 2000-3000 years) by looking at the frequency of specific genetic variants (SNPs) that are associated with height. It focuses on the direction of the association (whether a variant increases or decreases height) rather than just the magnitude of its effect.
  • Qx Statistics: This is a more recent method that extends the concept of genetic differentiation (like Fst) to quantify polygenic adaptation across multiple populations. It compares the observed genetic variation in height among populations to what would be expected under random genetic drift. The Qx statistic can be decomposed into "Fst-like" (non-directional differentiation) and "LD-like" (directional covariance of allele frequency differences) components, providing insights into the nature of the evolutionary forces at play.
  • Accounting for Ancient Admixture: A key innovation in this study is the extension of the Qx framework to incorporate quantitative covariates representing proportions of ancient Eurasian lineages (like Western Hunter-Gatherers and Neolithic Farmers). This allows the researchers to explicitly model and account for the impact of ancient population movements and mixing on observed genetic patterns.
  • Within-family GWAS: These types of studies are less susceptible to population stratification because they compare siblings or other close relatives, where genetic background is very similar.

Datasets

The study utilized several large-scale genetic datasets:

  • 1000 Genomes Project (1KGP): This provided allele frequency data for individuals from five European subpopulations (Tuscans, Iberians, British, Finns, and Utah residents of Northern and Western European ancestry).
  • POPRES study: This dataset provided genetic information for individuals from Portugal, Spain, Italy, France, and Switzerland.
  • UK Biobank (UKB): A large dataset used for GWAS summary statistics and as an LD reference panel.
  • Estonian Biobank (EEB): Provided allele frequencies for Estonia.
  • Lifelines study: Provided allele frequencies for the Netherlands.
  • MoBa cohort: Provided GWAS summary statistics for Norway.
  • iPSYCH cohort: Provided GWAS summary statistics for Denmark.
  • SweGen project: Provided allele frequencies for Sweden.
  • Howe et al. (2022) within-family GWAS: Provided summary statistics for height.
  • Yengo et al. (2022) large GWAS: Provided summary statistics for height across European, African, East Asian, and Hispanic ancestry groups.

Key Insights and Ideas

  • Height is a prime example of polygenic adaptation: The study strongly supports the idea that complex traits like height are shaped by the cumulative effects of many genes, each with a small effect, and that natural selection has acted on this polygenic architecture.
  • Selection signals persist even after accounting for admixture: A crucial finding is that the evidence for selection on height-associated genes remains significant even after controlling for the effects of ancient population mixing, suggesting that recent natural selection is a genuine driver of height differences.
  • The Netherlands as a height outlier: The study consistently identified the Netherlands as a significant outlier, contributing a disproportionately large share to the observed genetic signal of local adaptation for height. This aligns with the known phenotypic extreme of tall stature in the Dutch population.
  • Admixture explains a large portion, but not all, of the signal: While ancient admixture is a major factor influencing genetic differences in height across Europe, the study demonstrates that a residual signal of selection independent of admixture also exists.

Key Results

  • The tSDS analysis revealed significant signals of positive selection for height-increasing alleles in the ancestors of contemporary UK populations, and this signal was even stronger when using larger GWAS datasets.
  • The Qx analysis detected a strong signal of polygenic adaptation across Europe, with the Netherlands contributing the most significantly (87%) to this signal. Portugal was the second largest contributor.
  • When accounting for ancient admixture, the Qx statistic showed that 43-87% of the signal was explained by differential admixture, but a significant portion remained, indicating ongoing selection.
  • A PALM analysis suggested a marginally significant selection gradient for height-increasing alleles over the last 2,700 years in European populations.

Limitations and Future Work

  • Residual confounding: While the study employed advanced methods to control for population stratification and admixture, the authors acknowledge that some residual confounding might still exist, particularly in the LD-like component of the Qx statistic.
  • Sex specificity: The authors suggest that future work should investigate the sex specificity of natural selection on height, as sexual selection hypotheses for male height exist. Examining sex-differentiated GWAS could provide further insights.
  • Environmental and social factors: The extreme height phenotype in the Dutch population is likely influenced by a complex interplay of genetic factors, assortative mating, environmental influences, and socioeconomic confounders, which are not fully captured by genetic analysis alone.
  • Richer datasets: The authors anticipate that as richer datasets from ancient and global populations become available, their ability to untangle the evolutionary and biomedical consequences of selection will improve.

Summary updated

Abstract

Abstract The contribution of natural selection to height differences across Europe remains elusive. Prior evidence of selection on height based on genome-wide association study (GWAS) data has been challenged because of residual confounding in earlier GWAS results. Here, we reassess this evidence using complementary approaches integrating multi-ancestry and within-family GWAS data with allele frequency estimates from 13 European ancestry populations. We detect signatures of positive selection at height-associated loci within and across populations, which remain significant after accounting for ancient admixture events and is largely driven by allele frequency differences between the Netherlands and other European ancestry populations. Altogether, our results provide new and converging evidence of polygenic adaptation on height in Europe.
BibTeX
@misc{hivert2026converging,
  author = {Valentin Hivert and Pierrick Wainschtein and Matthew R. Robinson and Evan K. Irving-Pease and Eske Willerslev and Michael E. Goddard and William Barrie and J. Sidorenko and Peter M. Visscher and Loic Yengo},
  title = {{Converging evidence of positive selection at height-associated loci in Europe}},
  year = {2026},
  doi = {10.64898/2026.01.27.702172},
  howpublished = {bioRxiv},
}
Figure