Correcting for Genomic Inflation Leads to Loss of Power in Large-Scale Genome-Wide Association Study Meta-Analysis

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genomic inflation in genome-wide association studies (GWAS) summary statistics is a significant challenge.
• Existing correction methods include genomic control (GC) using the lambda (λ) value and linkage disequilibrium score regression (LDSR) using the LDSR intercept.

Purpose of the Study:

• To investigate factors influencing λ-values in GWAS.
• To evaluate the impact of GC and LDSR corrections on association signals using type 2 diabetes (T2D) as a model.
• To compare the conservativeness of GC and LDSR methods in large meta-analyses.

Main Methods:

• Analysis of λ-values in relation to sample size and variant frequency in GWAS.
• Comparison of T2D genetic associations before and after GC and LDSR correction across meta-analyses of varying sample sizes.
• Quantification of the loss of independent loci due to GC and LDSR corrections.

Main Results:

• Larger sample sizes increase λ-values due to enhanced polygenicity capture; lower frequency variants decrease λ-values due to reduced power.
• GC correction reduces false positives but leads to substantial loss of robust associations, including up to 39.7% of independent loci in a large meta-analysis.
• LDSR intercept correction results in a smaller loss of independent loci (up to 25.2%), indicating it is less conservative than GC.

Conclusions:

• Both GC and LDSR intercept correction methods lead to power loss in large, well-powered GWAS meta-analyses of polygenic traits.
• The findings highlight the trade-off between controlling false positives and retaining true associations.
• There is a critical need for the development of improved genomic inflation correction methods for large-scale genetic studies.