Grouped FDR

Grouped FDR controls the false discovery rate or family-wise error rate in genome-wide association studies (GWAS) and other large-scale genetic analyses by incorporating prior information through grouped hypothesis weighting of single nucleotide polymorphisms (SNPs) and candidate gene sets.

Key Features:

• Weighted Multiple Testing Procedure: Grouped FDR employs a weighted multiple testing approach that permits input of prior knowledge as grouped tests, enabling organization of SNPs into categories such as candidate genes versus non-candidate SNPs.
• Dynamic Weight Estimation: Weights are dynamically estimated per group from observed test statistics within that group, providing data-dependent prioritization of groups more likely to contain true associations.
• Robustness and Flexibility: The methodology maintains effectiveness when many groupings are uncorrelated with signal, typically improving power if one or more groups cluster multiple tests with signals and causing minimal power loss when groupings are random.
• Adaptive Weight Adjustment: Groups showing no apparent signal are down-weighted relative to groups with several significant tests, while weights become approximately equal when no groups exhibit signals.
• Scalability and Efficiency: The approach is applicable when the number of groups is small relative to the total number of tests, making it scalable for large genomic datasets typical of GWAS.

Scientific Applications:

• Genome-wide association studies (GWAS): Enhances detection of true genetic associations in GWAS by integrating prior groupings and data-dependent weight adjustment to mitigate the multiple testing burden.
• Candidate gene and pathway discovery: Improves power to detect associations for identifying candidate genes or pathways involved in complex traits and diseases across large numbers of tests.

Methodology:

Hypotheses are grouped based on prior knowledge, group-specific weights are estimated from observed test statistics (data-dependent), and those weights are applied within a weighted multiple testing framework to control FDR or family-wise error rate.