Ancestry Bias in Genetic Causal Inference

1,102 concepts857 papers109 proposals

Mendelian randomization (MR) has emerged as a powerful tool for uncovering causal relationships in epidemiology. By using genetic variants—typically single-nucleotide polymorphisms (SNPs)—as instrumental variables, MR mimics the structure of a randomized controlled trial. Because these variants are randomly assorted at conception, they are largely free from the confounding factors and reverse causation that plague traditional observational studies.

However, the validity of an instrumental variable relies on a fundamental assumption: the genetic architecture linking the variant, the exposure, and the outcome must remain stable. When MR models are trained on data from one ancestral population and applied to another, this assumption frequently breaks down. The overwhelming reliance of modern genome-wide association studies (GWAS) on cohorts of European descent creates a critical population diversity gap, limiting the global generalizability of clinical insights.

Why Ancestry Breaks Instrumental Variables

To understand why ancestry bias invalidates causal inferences, we must look at the mechanics of genetic inheritance. SNPs chosen as instruments in an MR study are rarely the actual functional mutations causing a trait. Instead, they are "tags" that are inherited alongside the causal variant because they reside close to each other on the chromosome—a phenomenon known as linkage disequilibrium (LD).

LD patterns are shaped by evolutionary history, population bottlenecks, and migrations. Consequently, a tagging SNP that is tightly linked to an exposure-increasing variant in a European population may be completely unlinked from that variant in an East Asian or African population. When researchers attempt to apply European-derived instruments to diverse cohorts, three primary failure modes occur:

  1. Loss of Instrument Relevance: The statistical power of an instrument depends on its strong association with the exposure. If the correlation between the tag SNP and the causal variant disappears in a different ancestral background, the instrument becomes "weak," leading to biased causal estimates.
  2. Pleiotropic Shifts: An instrument must influence the outcome only through the exposure (the exclusion restriction). A variant may have a straightforward pathway in one population but trigger alternative, unexpected biological pathways (horizontal pleiotropy) in another due to differing genetic backgrounds.
  3. Gene-Environment Interactions: The phenotypic effect of a genetic variant is often contingent on environmental context, such as diet, lifestyle, or the gut microbiome.

The Generalizability Gap in Practice

The consequences of this ancestry gap span every domain of complex disease research, from metabolic syndromes to neuropsychiatric conditions.

Metabolic and Cardiovascular Pathways

In cardiovascular and metabolic research, large-scale frameworks like the Generalized Summary-data-based Mendelian Randomization (GSMR) method developed by Zhihong Zhu et al. have successfully mapped dozens of causal links, such as protective effects against type-2 diabetes. Yet, as the authors note in Causal associations between risk factors and common diseases inferred from GWAS summary data, these associations are fundamentally constrained by the European-centric datasets from which they were derived.

Similarly, when investigating the complex interplay of mineral metabolism, Qinghao Zhao et al. identified fibroblast growth factor 23 (FGF23) as a causal driver of calcific aortic valve stenosis in Genetic association of serum calcium, phosphate, vitamin D, parathyroid hormone, and FGF23 with the risk of aortic stenosis. However, because the underlying data lacks ancestral diversity, it remains unknown whether therapeutic interventions targeting this pathway would yield the same efficacy in non-European populations.

The Gut-Brain and Lung-Gut Axes

The challenge compounds when genetic architectures interact with highly variable environmental factors, such as the microbiome. In Association of gut microbiome with risk of intracranial aneurysm: a mendelian randomization study, Chencheng Ma et al. demonstrated that specific bacterial taxa causally influence the risk of unruptured intracranial aneurysms. However, because both human genomic architecture and microbiome compositions vary drastically across geographic and ethnic lines, these gut-brain axis insights cannot be reliably generalized globally without cross-ethnic validation.

The same limitation applies to the "gut-lung axis." While Ruiling Lu established a causal link between adult-onset asthma and ulcerative colitis mediated by interleukin-18 in Causal relationship between asthma and ulcerative colitis and the mediating role of interleukin-18: a bidirectional Mendelian study and mediation analysis, the homogeneity of the study population leaves open the question of whether this inflammatory pathway operates universally.

Neuropsychiatric and Behavioral Traits

Behavioral and psychiatric MR studies are uniquely sensitive to population stratification. For instance, Guorui Zhao et al. uncovered a bidirectional "vicious cycle" between physical inactivity and major depressive disorder in Dissecting the causal association between social or physical inactivity and depression: a bidirectional two-sample Mendelian Randomization study. Because social structures, cultural attitudes toward activity, and genetic backgrounds heavily influence behavioral traits, validating these mechanisms in non-European cohorts is essential to ensure that lifestyle interventions are globally applicable.

[European Discovery Cohort] ──> Identifies Tag SNP ──> Strong LD with Causal Variant ──> Valid Causal Inference
                                                                │
                                                   Transferred to Diverse Cohort
                                                                │
                                                                ▼
[Diverse Target Cohort]     ──> Same Tag SNP      ──> Weak/No LD with Causal Variant ──> Invalid/Biased Inference

Moving Toward Multi-Ancestry Validation

To resolve these biases, the field is shifting toward multi-ancestry study designs that explicitly test for consistency across divergent genetic backgrounds.

A clear example of this approach is found in Mendelian randomization study shows no causal relationship between psychiatric disorders and glaucoma in European and East Asian populations. Here, Yan Zhang et al. evaluated the relationship between depression, insomnia, and glaucoma across both European and East Asian cohorts. By demonstrating a consistent lack of causal genetic linkage in both distinct populations, the authors provided robust evidence that the clinical correlations seen worldwide are driven by environmental factors rather than shared biology.

When multi-ancestry data is absent, researchers must treat their findings as population-specific hypotheses rather than universal biological truths. Whether mapping the genetic architecture of multimodal human brain age, as done by Haochang Shou et al. in The genetic architecture of multimodal human brain age, or identifying metabolic biomarkers for postmenopausal osteoporosis, as explored by Yimin Liu et al. in A Mendelian randomization study investigating the causal relationship between 35 blood and urine metabolite biomarkers and postmenopausal osteoporosisShort Communication, the conclusion remains the same: equitable precision medicine requires diverse genomic data. Until the diversity gap in primary GWAS datasets is closed, the promises of genetic causal inference will remain bounded by the ancestry of the cohorts we study.

History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_population_generalizability_ancestry_bia_872ce77a,
  title = {population generalizability (ancestry bias)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/population-generalizability-ancestry-bias/rdr_872ce77af6e14ae8bc2361ce50100ca8},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_population_generalizability_ancestry_bia_872ce77a_v1,
  title = {population generalizability (ancestry bias)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/population-generalizability-ancestry-bias/rdr_872ce77af6e14ae8bc2361ce50100ca8/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake