It's a wrap: deriving distinct discoveries with FDR control after a GWAS pipeline

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Conditional independence testing with false discovery rate (FDR) control is crucial for analyzing large-scale genomic data, particularly genome-wide association studies (GWAS).
• Previous methods required pre-computed linkage disequilibrium patterns for specific populations (e.g., European genomes), limiting broader applicability.
• Existing tools like GhostKnockoffGWAS facilitate FDR-controlled secondary analyses on GWAS summary statistics but depend on population-specific negative control distributions.

Purpose of the Study:

• To introduce and release solveblock, a novel software pipeline that extends FDR-controlled conditional independence testing to diverse populations.
• To enable efficient estimation of high-dimensional correlation matrices from genotyped or reference datasets for improved genomic analyses.
• To provide a computational framework for defining appropriate resolutions for hypothesis testing and computing necessary negative control distributions.

Main Methods:

• solveblock efficiently estimates genome-wide correlation structures using sparsity assumptions from provided genotyped samples or reference data.
• The software identifies groups of highly correlated genetic variants to establish resolution for conditional independence testing.
• It computes the distribution of exchangeable negative controls, which are then used as input for downstream analyses like GhostKnockoffGWAS.

Main Results:

• The solveblock pipeline successfully estimates sample-specific correlation matrices and generates negative control distributions for diverse populations.
• In simulations, the method demonstrated effective control of the false discovery rate (FDR).
• Analysis of UK Biobank data for 26 phenotypes revealed an average of approximately 19 additional discoveries compared to standard marginal association testing.

Conclusions:

• solveblock significantly enhances the capability to perform FDR-controlled conditional independence testing across a wider range of genomic studies and populations.
• The software facilitates a two-step analysis procedure (solveblock followed by GhostKnockoffGWAS) for robust signal detection in GWAS.
• Open sharing of code, precompiled software, and processed files promotes accessibility and further research in genetic discovery.