A close examination of double filtering with fold change and T test in microarray analysis

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• Researchers commonly use fold change and t-tests for identifying differentially expressed genes.
• The double filtering procedure is popular due to its simplicity, despite more advanced methods existing.

Purpose of the Study:

• To provide theoretical insight into the limitations of the double filtering procedure.
• To develop a more powerful statistical method for identifying differentially expressed genes.

Main Methods:

• Theoretical analysis of statistical assumptions for fold change and t-tests.
• Development of a likelihood ratio test statistic under a mixture gene variance model.
• Comparison with Bayesian mixture models and Significance Analysis of Microarrays (SAM).

Main Results:

• Fold change and t-tests rely on conflicting variance assumptions (common vs. gene-specific).
• A likelihood ratio test based on a mixture model is theoretically optimal.
• Bayesian inference and SAM are better approximations than double filtering.

Conclusions:

• Shrinkage testing methods, unified under a mixture gene variance assumption, significantly outperform the double filtering procedure.
• The findings are supported by hypothesis testing theory, simulations, and real data analysis.