Correcting for volunteer bias in GWAS uncovers novel genetic variants and increases heritability estimates

Abstract

The implications of selection bias due to volunteering (volunteer bias) for genetic association studies are poorly understood. Because of its large sample size and extensive phenotyping, the UK Biobank (UKB) is included in almost all large genomewide association studies (GWAS) to date, as it is one of the largest cohorts. Yet, it is known to be highly selected. We develop inverse probability weighted GWAS (WGWAS) to estimate GWAS summary statistics in the UKB that are corrected for volunteer bias. WGWAS decreases the effective sample size substantially compared to GWAS by an average of 61% (from 337,543 to 130,684) depending on the phenotype. The extent to which volunteer bias affects GWAS associations and downstream results is phenotype-specific. Through WGWAS we find 11 novel genomewide significant loci for type 1 diabetes and 3 for breast cancer. These loci were not identified previously in any prior GWAS. Further, genetic variant's effect sizes and heritability estimates become more predictive in WGWAS for certain phenotypes (e.g., educational attainment, drinks per week, breast cancer and type 1 diabetes). WGWAS also alters biological annotation relations in gene-set analyses. This suggests that not accounting for volunteer-based selection can result in GWASs that suffer from bias, which in turn may drive spurious associations. GWAS consortia may therefore wish to provide population weights for their data sets or rely more on population-representative samples.

Evaluating the causal relationship between educational attainment and mental health

“…In contrast, the MR is based on GWASs of international samples who are volunteers of European-ancestry. Both higher risk for mental illness and lower EA are known to increase non-participation and participant attrition [56][57][58][59] , a selection bias that could induce a collider bias 60 . In the withinsibship analysis, we rely on diagnoses by professionals in specialized care, therefore dependent on the current referral policies of the Dutch healthcare system.…”

GUANinE v1.0: Benchmark Datasets for Genomic AI Sequence-to-Function Models

“…ensure statistical significance while having limited or quantifiable confounders (e.g. batch effects or socioeconomic factors [79]), a requirement of any evaluatory dataset [25].…”

“…Our goal is to provide a set of benchmarking tasks for genomic AI that A) allow for direct, supervised training of high-complexity models from scratch (tabula rasa) for comparison to pretraining regimes for transfer learning, and B) ensure statistical significance while having limited or quantifiable confounders (e.g. batch effects or socioeconomic factors [79]), a requirement of any evaluatory dataset [25]. GUANinE v0.9 prioritizes functional genomic annotation and understanding on short-to-moderate length sequences (between 80 and 512 nucleotides), rather than exploring long sequences inputs or distal effects [39,21,37].…”

Intergenerational transmission of education and mental health

“…In contrast, the MR is based on GWASs of international samples who are volunteers of European-ancestry. Both higher risk for mental illness and lower EA are known to increase non-participation and participant attrition [261][262][263][264] , a selection bias that could induce a collider bias 265 . In the within-sibship analysis, we rely on diagnoses by professionals in specialized care, therefore dependent on the current referral policies of the Dutch healthcare system.…”

“…Correction for the non-representativity of these cohorts can be partially done, for example weights for correcting the important volunteer bias of UK Biobank have recently been made available recently 263,264,377 . Similar efforts should be undertaken for all large cohorts on which most of the current genetically-informed studies are taking place.…”