Uniform approximation is more appropriate for Wilcoxon Rank-Sum Test in gene set analysis

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene set analysis is crucial for interpreting high-throughput gene expression data.
• The Wilcoxon Rank-Sum (WRS) test is a common method for gene set enrichment analysis.
• Current WRS test implementations often rely on normal distribution approximations for null distributions.

Purpose of the Study:

• To evaluate the accuracy and power of normal versus uniform approximations for the WRS test in gene set analysis.
• To identify scenarios where the normal approximation is suboptimal.
• To propose and validate the uniform approximation as a superior alternative.

Main Methods:

• Comparative analysis of normal and uniform distribution approximations for the WRS test.
• Simulations were conducted to assess performance under various conditions.
• Real-world gene set analysis, specifically Gene Ontology (GO) term analysis, was used for validation.

Main Results:

• The normal approximation for the WRS test is shown to be problematic for gene sets with a relatively small number of genes.
• A uniform approximation demonstrated superior power, accuracy, and computational efficiency in these scenarios.
• The advantages of the uniform approximation were confirmed using both simulated and real gene expression datasets.

Conclusions:

• The normal approximation can lead to inaccurate results in gene set enrichment analysis, particularly with smaller gene sets.
• A uniform approximation provides a more robust and powerful statistical framework for the WRS test.
• The findings advocate for the adoption of uniform approximations in gene set analysis for improved biological interpretation.