Symmetric Directional False Discovery Rate Control.

Area of Science:

• Bioinformatics
• Statistical Genetics
• Computational Biology

Background:

• Standard false discovery rate (FDR) methods are commonly used for identifying significant genes in gene expression studies.
• Analysis of real gene expression data revealed highly unbalanced results, with significantly more under-expressed than over-expressed genes.
• This imbalance suggested that the underlying two-sample t-statistics were skewed, potentially rendering standard FDR methods inappropriate.

Purpose of the Study:

• To address the limitations of standard FDR methods when applied to gene expression data with asymmetric distributions.
• To develop and evaluate a novel symmetric directional FDR control method for more balanced identification of significant genes.
• To improve the reliability of gene significance identification in biological studies with skewed expression patterns.

Main Methods:

• Proposed a symmetric directional FDR control method.
• Categorized genes into over-expressed and under-expressed groups.
• Paired over- and under-expressed genes and defined p-values for gene pairs using column permutations.
• Applied the standard FDR method to selected gene pairs instead of individual genes.

Main Results:

• The proposed symmetric directional FDR method demonstrated improved performance compared to standard FDR methods.
• Evaluation on simulated and real gene expression datasets showed more balanced identification of over- and under-expressed genes.
• The method effectively handles skewed t-statistics often encountered in gene expression analyses.

Conclusions:

• The symmetric directional FDR control method offers a more appropriate approach for analyzing gene expression data with asymmetric distributions.
• This method enhances the identification of biologically relevant genes by addressing the imbalance issue.
• The findings suggest a more robust statistical framework for gene expression data analysis in bioinformatics.