Stratified false discovery control for large-scale hypothesis testing with application to genome-wide association...

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• High-throughput genotyping increases the multiplicity problem in genetic studies.
• False Discovery Rate (FDR) control is crucial for managing false positives in large-scale genetic analyses.
• Genetic studies often exhibit natural stratification among hypotheses.

Purpose of the Study:

• Investigate the performance of stratified FDR control compared to aggregated FDR control.
• Analyze the impact of stratification on false positive and true positive rates in genetic studies.
• Evaluate the feasibility of low FDR control rates in Genome-Wide Association studies.

Main Methods:

• Compared stratified FDR control (separate strata) with aggregated FDR control (single stratum).
• Analyzed performance under fixed rejection region and fixed FDR frameworks.
• Applied methods to a Genome-Wide Association study dataset.

Main Results:

• Aggregated FDR is a weighted average of stratum-specific FDRs under the fixed rejection region framework.
• Identified conditions for stratified FDR to yield equal or greater true positives than aggregated FDR.
• Demonstrated potential advantages of stratified FDR in a GWA study.
• Found that controlling FDR at low rates (5-10%) may be infeasible for some GWA studies.

Conclusions:

• Stratified FDR control can be more effective than aggregated methods in genetic studies with natural stratification.
• The choice of FDR control method impacts the power and accuracy of genetic association studies.
• Careful consideration of FDR control strategies is necessary for interpreting results from large-scale genetic analyses.