Genome-Wide Assessment of Pleiotropy Across >1000 Traits from Global Biobanks

Michael G. Levin, Satoshi Koyama, Jakob Woerner, David Zhang, Alexis Rodriguez, Tarak Nandi, Buu Truong, Sarah Abramowitz, Hemant Gupta, Himani Kamineni, Whitney Hornsby, Zilinghan Li, Taylor Cohron, Jennifer E. Huffman, Patrick T. Ellinor, Dokyoon Kim, Katherine P. Liao, Ravi Madduri, Benjamin F. Voight, Anurag Verma, Scott M. Damrauer, Pradeep NatarajanmedRxiv 2025 · DOI

Introduction

The polygenic nature of complex human traits is well-established, with thousands of risk loci identified through genome-wide association studies (GWAS). However, a fundamental question remains: to what extent do these genetic regions influence multiple, seemingly unrelated phenotypes—a phenomenon known as pleiotropy? While previous research suggested that pleiotropy is common, it has been difficult to distinguish between true shared biological mechanisms and "apparent" pleiotropy caused by linkage disequilibrium (LD), where distinct causal variants for different traits are simply located near one another.

In a massive multi-ancestry effort, researchers analyzed data from nearly 1.8 million individuals across global biobanks to create a comprehensive map of genetic pleiotropy. By examining 1,167 clinically relevant traits, the study demonstrates that pleiotropy is the rule rather than the exception, providing new insights into the functional architecture of the human genome.

Bridging the Diversity Gap

Historically, genetic research has suffered from a significant "diversity gap," with the vast majority of participants being of European descent. This study addresses this limitation by integrating data from diverse populations, including individuals genetically similar to Admixed American (AMR), African (AFR), East Asian (EAS), European (EUR), and South Asian (SAS) reference populations.

The researchers utilized data from several major global biobanks:

  • Discovery Cohorts: VA Million Veteran Program (MVP), UK Biobank (UKB), FinnGen, Biobank Japan (BBJ), Tohoku Medical Megabank (ToMMo), and the Korean Genome and Epidemiology Study (KoGES).
  • Replication Cohort: The All of Us (AoU) Research Program, an independent cohort of over 200,000 U.S. participants.

Methodology: Untangling Shared Signals

To identify and characterize pleiotropic loci, the team employed several sophisticated statistical frameworks:

  • Multi-Ancestry Meta-Regression (MR-MEGA): This tool allowed for the detection of association signals while accounting for ancestry-specific differences in allele frequencies.
  • Fine-Mapping: Bayesian fine-mapping was used to calculate Posterior Inclusion Probabilities (PIP) to pinpoint specific causal variants within a locus.
  • Multi-Trait Colocalization (HyPrColoc): This was critical for distinguishing true pleiotropy. It assessed whether the same causal variant was responsible for the signals observed across multiple traits.
  • Polygenic Score (PGS) Development: Using PRS-CSx, the researchers developed predictive models that leverage multi-ancestry data to improve risk prediction across different populations.

Key Findings: Pleiotropy as the Norm

The study identified 27,193 genome-wide significant locus-trait pairs in within-population analyses and 29,139 in multi-population analyses. These signals were concentrated within 2,624 non-overlapping genomic windows, averaging approximately 600kb in size.

1. The Extent of Pleiotropy

The analysis revealed that 80% of the identified loci (2,110 out of 2,624) were associated with more than one trait. A median locus influenced six different traits, though some "super-pleiotropic" regions, such as the APOE locus, were associated with over 60 phenotypes. Crucially, multi-trait colocalization confirmed that 72% of these loci (1,902) showed high-confidence evidence of a shared causal variant across two or more traits, suggesting a common biological pathway rather than mere genetic proximity.

2. Functional and Evolutionary Insights

Pleiotropic variants were found to be significantly enriched for functional annotations. Compared to non-pleiotropic variants, lead pleiotropic signals were 19 times more likely to be "stop-gain" variants and 9.8 times more likely to be "missense" variants—changes that directly alter protein structure or function. Furthermore, these variants showed evidence of being under stronger natural selection, indicating that genes with multiple roles are "high-stakes" in human evolution.

3. New Discoveries and Improved Prediction

The inclusion of diverse populations led to the discovery of thousands of new associations. Approximately 11.5% of locus-trait pairs in population-wise analyses and 6.4% in multi-population analyses had not reached significance in previously published GWAS.

The study also demonstrated the clinical utility of multi-ancestry data. The newly developed Polygenic Scores (PGS) generally outperformed existing scores in the PGS Catalog. The improvement was most pronounced in African populations, where the new scores showed better predictive performance for 97% of the phenotypes tested.

Limitations and Future Directions

Despite the unprecedented scale of this study, challenges remain. While diversity was improved, European participants still constituted the majority of the data (~1.2 million), which limits the power to detect rare, population-specific variants in other groups. Additionally, the reliance on Electronic Health Record (EHR) data introduces potential noise compared to standardized clinical assessments.

The identification of highly pleiotropic loci provides a roadmap for drug discovery, highlighting "core" biological nodes. However, the authors caution that targeting these genes may carry a higher risk of side effects due to their influence on multiple physiological systems. Future research will likely focus on the role of rare variants and the further refinement of these genetic maps as biobanks continue to grow in size and diversity.

Summary updated

Abstract

Large-scale genetic association studies have identified thousands of trait-associated risk loci, establishing the polygenic basis for common complex traits and diseases. Although prior studies suggest that many trait-associated loci are pleiotropic, the extent to which this pleiotropy reflects shared causal variants or confounding by linkage disequilibrium remains poorly characterized. To define a set of candidate loci with potentially pleiotropic associations, we performed genome-wide association study (GWAS) meta-analyses of up to 1,167 clinically relevant traits and diseases across 1,789,365 diverse individuals genetically similar to Admixed American (AMR, N<sub>Max</sub> = 60,756), African (AFR, N<sub>Max</sub> = 128,361), East Asian (EAS, N<sub>Max</sub> = 307,465), European (EUR, N<sub>Max</sub> = 1,283,907), and South Asian (SAS, N<sub>Max</sub> = 8,876) reference populations from the VA Million Veteran Program (MVP), UK Biobank (UKB), FinnGen, Biobank Japan (BBJ), Tohoku Medical Megabank (ToMMo), and Korean Genome and Epidemiology Study (KoGES). We identified 27,193 genome-wide significant locus-trait pairs (1MB region with P<sub>GWAMA</sub> < 5 × 10<sup>-8</sup>) in within-population analysis and 29,139 in multi-population analysis (P<sub>MR-MEGA</sub> < 5 × 10<sup>-8</sup>). Among these, 11.5% (n = 3,149) of locus-trait pairs in population-wise and 6.4% (n = 1,875) in multi-population analyses did not reach genome-wide significance in previously published GWAS. In aggregate, the genome-wide significant loci fell within 2,624 non-overlapping autosomal genomic windows on average ~600kb in size. Each locus contained genome-wide significant signals for a median of 6 traits (IQR 2 to 18), including 2,110 (80%) pleiotropic loci associated with >1 trait. Multi-trait colocalization identified 1,902 (72%) loci with high-confidence (posterior probability > 0.9) evidence of a shared causal variant across two or more traits. Variants in pleiotropic loci were significantly enriched for a broad spectrum of functional annotations compared to non-pleiotropic counterparts. Polygenic scores (PGS) developed from these data generally improved prediction compared to existing PGS, and were broadly associated with both primary and pleiotropic phenotypes. These results provide a contemporary map of genetic pleiotropy across the spectrum of human traits/diseases and diverse genetic backgrounds.
BibTeX
@misc{levin2025genome,
  author = {Michael G. Levin and Satoshi Koyama and Jakob Woerner and David Zhang and Alexis Rodriguez and Tarak Nandi and Buu Truong and Sarah Abramowitz and Hemant Gupta and Himani Kamineni and Whitney Hornsby and Zilinghan Li and Taylor Cohron and Jennifer E. Huffman and Patrick T. Ellinor and Dokyoon Kim and Katherine P. Liao and Ravi Madduri and Benjamin F. Voight and Anurag Verma and Scott M. Damrauer and Pradeep Natarajan},
  title = {{Genome-Wide Assessment of Pleiotropy Across \&gt;1000 Traits from Global Biobanks}},
  year = {2025},
  doi = {10.1101/2025.04.18.25326074},
  howpublished = {medRxiv},
}
Figure