Finding differentially expressed genes in high dimensional data: Rank based test statistic via a distance measure

Area of Science:

• Bioinformatics
• Statistical genetics
• Computational biology

Background:

• Identifying differentially expressed genes is crucial for understanding biological processes.
• Existing methods often rely on normality assumptions or lack robustness to outliers.
• There is a need for powerful and robust statistical methods for gene expression analysis.

Purpose of the Study:

• To introduce a novel rank-based test statistic for identifying differentially expressed genes.
• To evaluate the robustness and power of the proposed test statistic.
• To compare the performance against established methods like paired t-test, Wilcoxon signed rank test, and Significance Analysis of Microarray (SAM).

Main Methods:

• Development of a rank-based test statistic utilizing a distance measure.
• Robustness evaluation against extreme values and distribution assumptions.
• Simulation studies to assess statistical power under various non-normal distributions.
• Derivation of the asymptotic distribution and p-value function.

Main Results:

• The proposed test statistic demonstrates high robustness to extreme values.
• It does not require assumptions about the parent population distribution.
• Simulation results indicate superior power compared to paired t-test, Wilcoxon signed rank test, and SAM under specific non-normal conditions.
• The method's applicability is validated on a real-world dataset.

Conclusions:

• The novel rank-based test statistic offers a robust and powerful alternative for identifying differentially expressed genes.
• This method is particularly advantageous when dealing with non-normal data distributions.
• The findings suggest improved gene expression analysis in biological research.