Integrating mean and variance heterogeneities to identify differentially expressed genes

Area of Science:

• Functional genomics
• Statistical genetics
• Bioinformatics

Background:

• Traditional methods focus on mean expression heterogeneity, often ignoring variance heterogeneity.
• Condition changes can alter both mean and variance of gene expression levels.
• Variance heterogeneity provides additional biological insights into gene expression regulation.

Purpose of the Study:

• To propose the concept of mean-variance differentially expressed (MVDE) genes.
• To develop an integrative mean-variance test (IMVT) for detecting MVDE genes.
• To evaluate the performance of IMVT against existing methods.

Main Methods:

• Mathematical proof of null independence between mean and variance heterogeneity tests.
• Development of the integrative mean-variance test (IMVT).
• Comprehensive simulations under normality and Laplace distributions.
• Application to peripheral circulating B cell gene expression data.

Main Results:

• IMVT demonstrated superior performance and power compared to existing methods, especially when variance heterogeneity is present.
• IMVT controlled Type I error rates effectively in simulations.
• Analysis of B cell data revealed informative variance heterogeneity, with IMVT identifying novel significant genes.

Conclusions:

• Integrating variance heterogeneity offers significant gains in functional genomics analysis.
• The IMVT provides a more comprehensive summary of condition-induced expression changes.
• Explicitly exploiting variance heterogeneity is crucial for robust gene expression analysis.