FDR control with pseudo-gatekeeping based on a possibly data driven order of the hypotheses

Area of Science:

• Statistics
• Bioinformatics
• Computational Biology

Background:

• Controlling the false discovery rate (FDR) is crucial in high-dimensional data analysis.
• Existing multiple testing procedures may lack power or require strong assumptions about data distribution.
• Identifying significant biological markers requires robust statistical methods.

Purpose of the Study:

• To develop a novel multiple testing procedure that controls the false discovery rate (FDR).
• To enhance the identification of biologically relevant hypotheses by prioritizing those with larger effect sizes.
• To ensure the procedure's validity under both independent and arbitrarily dependent test statistics.

Main Methods:

• A data-driven ordering of hypotheses based on effect size.
• Testing hypotheses sequentially at an uncorrected significance level (q) until a stopping rule is met.
• A modification to ensure validity under arbitrary dependence structures.

Main Results:

• The proposed procedure effectively controls the false discovery rate (FDR).
• The method demonstrates superior performance compared to existing procedures, particularly in small sample size scenarios.
• The procedure successfully identified molecular signatures in intracranial ependymoma data.

Conclusions:

• The developed multiple testing procedure offers a powerful and flexible approach for FDR control.
• Its data-driven ordering enhances the discovery of significant hypotheses, especially in challenging small sample settings.
• The associated R package (someMTP) provides accessible implementation for researchers.