SIEVEseq: Unified differential expression, variability, and skewness analyses using RNA-Seq data

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• RNA-Seq analysis often focuses on differential expression, potentially missing complex gene expression changes.
• Existing discrete count models struggle to independently parameterize mean, variance, and skewness.
• A unified statistical framework is needed to analyze multiple aspects of gene expression distributions.

Purpose of the Study:

• To introduce SIEVEseq, a novel statistical methodology for RNA-Seq data analysis.
• To provide a unified framework for simultaneously testing differential expression, variability, and skewness.
• To improve the comprehensive understanding of gene expression complexity.

Main Methods:

• Developed SIEVEseq, a statistical framework utilizing compositional data analysis.
• Transformed discrete RNA-Seq counts into continuous data fitted by a skew-normal distribution.
• Validated SIEVEseq through parametric and nonparametric simulations and analysis of the Mayo RNA-Seq dataset.

Main Results:

• SIEVEseq demonstrates superior control of false discovery rate and Type II error compared to existing methods.
• Analysis of Alzheimer's disease data revealed gene sets with significant differences in mean, variance, and skewness predict disease state.
• Incorporating differential variability and skewness identified additional disease-related biological aspects beyond differential expression.

Conclusions:

• SIEVEseq provides a robust framework for comprehensive RNA-Seq data analysis.
• The methodology enhances the detection of biological signals and improves disease state prediction.
• SIEVEseq reveals a broader spectrum of biological insights compared to methods focusing solely on differential expression.