Multiple testing in candidate gene situations: a comparison of classical, discrete, and resampling-based procedures

Area of Science:

• Genetics
• Statistical genetics
• Bioinformatics

Background:

• Candidate gene association studies often involve testing multiple hypotheses simultaneously on partially correlated SNP data.
• Accounting for multiple testing is crucial, with various procedures proposed to control error rates.
• The benefit of discrete-aware multiple testing procedures for correlated data remains an open question.

Purpose of the Study:

• To empirically evaluate the performance of different multiple testing procedures using simulated correlated SNP data.
• To compare False Discovery Rate (FDR) and Family-Wise Error Rate (FWER) controlling procedures, including those for discrete situations and minP-resampling.
• To investigate the impact of sample size and gene data scenarios on procedure performance.

Main Methods:

• Simulation studies were conducted using simulated SNP data with varying degrees of correlation.
• Multiple testing procedures evaluated include FDR controllers, FWER controllers, discrete situation procedures, and minP-resampling.
• Performance was assessed across a broad range of gene data scenarios and sample sizes.

Main Results:

• Sample size was the primary factor influencing the performance of multiple testing procedures.
• In small sample size scenarios, the minP-resampling procedure demonstrated higher statistical power than classical FDR procedures, despite controlling the stricter FWER.
• In larger sample size scenarios, FDR controlling procedures resulted in a greater number of significant rejections.

Conclusions:

• The choice of multiple testing procedure significantly impacts results in candidate gene association studies with correlated SNP data.
• The minP-resampling procedure is advantageous in small sample settings for its power, while FDR procedures are more potent with larger sample sizes.
• These findings highlight the importance of considering sample size when selecting multiple testing strategies for genetic association studies.