A biobank-scale test of marginal epistasis reveals genome-wide signals of polygenic interaction effects

statistical genetics
epistasis
marginal epistasis
GWAS
complex traits
genetic architecture
  • Methodology: The paper introduces FAME (FAst Marginal Epistasis test), a new, computationally efficient statistical method designed to detect marginal epistasis—the aggregate interaction effect between a single SNP and the entire polygenic background—in large biobanks.
  • Key Finding: Applying FAME to 53 quantitative traits in the UK Biobank, the study identified 16 significant marginal epistasis signals across 12 traits, providing the first systematic, genome-wide evidence of polygenic interaction effects.
  • Implication: The findings confirm that genetic interactions are a measurable component of complex trait architecture, suggesting that current additive GWAS models are incomplete and that marginal epistasis may contribute to the “missing heritability.”
Published

23 January 2026

PubMed: 41366086 DOI: 10.1038/s41588-025-02411-y Overview generated by: Gemini 2.5 Flash, 10/12/2025

Research Goal and Methodology

This study addresses the challenge of identifying genetic interactions (epistasis) in complex human traits, which is often hindered by computational complexity and statistical power limitations in biobank-scale datasets. The authors introduce FAME (FAst Marginal Epistasis test), a novel, computationally efficient method designed to detect marginal epistasis for a single-nucleotide polymorphism (SNP) on a quantitative trait.

What is Marginal Epistasis?

Marginal epistasis, in this context, refers to whether the effect of a given SNP on a trait is significantly modulated by the individual’s overall genetic background. It tests for the aggregate effect of interaction between a focal SNP and all other polygenic effects, rather than requiring the testing of millions of pairwise SNP interactions.

FAME Algorithm

The FAME method is designed as an extension of standard linear mixed models (LMMs) used in GWAS. It estimates the interaction effect of a focal SNP with the polygenic background by:

  1. Estimating Polygenic Background: The effect of the entire genetic background is estimated using a standard LMM and captured by the individual-level polygenic scores (or ‘residual additive effect’).
  2. Testing Interaction: FAME then tests the interaction between the focal SNP’s genotype and this estimated polygenic background effect.

This formulation allows FAME to be highly efficient, enabling genome-wide testing of marginal epistasis, which was previously intractable for large biobanks.

Key Findings: Genome-Wide Marginal Epistasis Signals

The authors applied FAME to GWAS-significant trait-SNP associations across 53 quantitative traits in approximately 300,000 unrelated White British individuals from the UK Biobank (UKBB).

  • Significant Signals: FAME identified 16 significant marginal epistasis signals across 12 unique traits, with the most robust signals found for anthropometric traits (e.g., height, standing height) and blood cell phenotypes.
  • Trait-Specific Patterns:
    • Height: Two of the most significant epistasis signals were found for height, suggesting that the effect of specific height-associated SNPs is non-linearly dependent on the individual’s overall polygenic predisposition for height.
    • Blood Cell Traits: Strong signals were also detected for mean corpuscular volume and monocyte count, indicating that complex cell biology is also influenced by polygenic interaction effects.
  • Novel Findings: While previous studies have hinted at epistasis, the 16 signals identified by FAME are novel findings at a genome-wide scale, demonstrating the power of marginal epistasis testing to capture aggregate interaction effects.

Implications for Heritability and Genetic Architecture

The study provides evidence for the widespread presence of polygenic interaction effects, which has significant implications for understanding the total genetic architecture of complex traits:

  • Contribution to Variation: The authors estimated the proportion of phenotypic variance explained by these marginal epistasis effects, finding that, while individually small (typically \(<0.05\%\)), the cumulative contribution across the genome may be substantial, potentially helping to explain some of the “missing heritability” gap.
  • Additive Model Limitations: The presence of these polygenic interaction effects suggests that the common assumption in GWAS—that genetic effects are purely additive—is an oversimplification. FAME provides a scalable way to account for the systematic non-additive contributions to trait variation.

Conclusions

FAME is a powerful and scalable statistical method that has successfully identified genome-wide signals of marginal epistasis in a biobank-scale cohort. The identification of 16 robust marginal epistasis signals across diverse traits underscores that genetic interactions with the polygenic background are a fundamental component of the genetic architecture of human complex traits. This work opens the door for more comprehensive modeling of non-additive effects in future large-scale genetic studies.