Controlling false discoveries in multidimensional directional decisions, with applications to gene expression data on...

Area of Science:

• Statistical genomics
• Bioinformatics
• Gene expression analysis

Background:

• Microarray studies analyze gene expression patterns over ordered categories (e.g., time-course, dose-response).
• Identifying these patterns involves multiple testing, often focusing on controlling the false discovery rate (FDR).
• Existing methods primarily control the standard FDR, not the mixed directional FDR (mdFDR), which accounts for directional errors.

Purpose of the Study:

• To develop a statistical procedure for controlling the mixed directional false discovery rate (mdFDR) in multidimensional gene expression analysis.
• To extend existing Benjamini-Hochberg (BH) procedures to handle complex, multidimensional testing scenarios in gene expression data.

Main Methods:

• Developed a novel procedure extending the Benjamini-Yekutieli approach using Bonferroni tests for multidimensional mdFDR control.
• Provided a theoretical proof for mdFDR control under statistical independence of test statistics across genes.
• Conducted simulation studies to evaluate performance under both independent and dependent test statistics.

Main Results:

• The proposed method effectively controls mdFDR in multidimensional settings.
• Simulation results demonstrate the procedure's robustness under varying statistical dependencies.
• Application to time-course microarray data identified significant cell-cycle genes (e.g., MCM4, RFC2) missed by previous analyses.

Conclusions:

• The developed statistical procedure offers a robust approach for controlling mdFDR in complex gene expression studies.
• This methodology enhances the identification of biologically relevant genes, particularly in time-course experiments.
• The findings provide a valuable tool for re-analyzing existing microarray datasets and discovering novel biological insights.