Robust inference with GhostKnockoffs in genome-wide association studies.

Area of Science:

• Genetics and Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-Wide Association Studies (GWASs) are crucial for understanding complex traits' genetic basis.
• Controlling the false discovery rate (FDR) is effective for detecting small effect loci in GWASs.
• Existing FDR methods struggle with correlated genetic variants and related individuals in GWASs.

Purpose of the Study:

• To develop a robust knockoff-based method for FDR control in GWASs, accommodating related individuals.
• To enhance the discovery of small effect loci associated with complex polygenic traits.
• To generalize knockoff methods for broader GWAS applications, including meta-analyses and diverse association statistics.

Main Methods:

• Proposed a novel knockoff-based approach integrating GhostKnockoffs with advanced marginal association tests.
• Utilized GWAS Z-scores as input, ensuring robustness to arbitrary relatedness structures via generalized linear mixed models.
• Validated the method through simulation studies and a meta-analysis of Alzheimer's disease GWAS and sequencing data.

Main Results:

• The proposed method demonstrates robustness to relatedness structures when using Z-scores from valid generalized linear mixed models.
• Successfully applied the method to a meta-analysis of nine European ancestral GWASs and sequencing studies for Alzheimer's disease.
• The approach effectively enhances the discovery of small effect loci, improving feature selection in GWAS data.

Conclusions:

• The integrated GhostKnockoffs approach provides a flexible and statistically valid tool for FDR control in GWASs with related individuals.
• This method expands the applicability of knockoff-based procedures to complex genetic studies and meta-analyses.
• The findings support the enhanced discovery of genetic associations for complex traits, including neurodegenerative diseases like Alzheimer's.