Estimating the effective sample size in association studies of quantitative traits

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• The effective sample size (ESS) quantifies sample correlation, crucial for predicting statistical power in genome-wide association studies (GWAS) employing linear mixed models.
• Accurate ESS estimation is vital for optimizing study design and interpreting results in genetic epidemiology.

Purpose of the Study:

• To introduce an analytical formula for ESS in mixed-model GWAS of quantitative traits.
• To provide a framework for approximating ESS in related and unrelated samples, covering both marginal genetic and gene-environment interaction tests.
• To quantitatively assess statistical power implications, including power loss/gain factors and the impact of family structure on gene-environment interaction GWAS.

Main Methods:

• Derivation of an analytical ESS formula for mixed-model GWAS.
• Development of ESS approximations for related and unrelated individuals.
• Simulation studies to validate ESS approximations and evaluate statistical power.
• Application of the developed framework to UK Biobank data for mixed-model GWAS.

Main Results:

• Validated ESS approximations through simulations, offering insights into statistical power under various conditions.
• Demonstrated that gene-environment interaction GWAS power in related individuals is sensitive to family structure and exposure distribution.
• Confirmed simulation findings in UK Biobank GWAS, revealing a negligible power drop due to family relatedness.

Conclusions:

• The proposed analytical ESS framework enhances power prediction for mixed-model GWAS.
• Family structure has a minimal impact on power in large-scale UK Biobank GWAS, contrary to expectations.
• The findings provide a quantitative basis for designing and interpreting genetic association studies, particularly those involving related individuals and interaction analyses.