Robust normalization and transformation techniques for constructing gene coexpression networks from RNA-seq data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene coexpression networks are vital for analyzing high-throughput gene expression data.
• Optimal workflows exist for microarrays but not for RNA-seq data.
• Existing RNA-seq studies focus on differential gene expression, not network construction.

Purpose of the Study:

• To benchmark different workflows for RNA-seq coexpression network construction.
• To identify optimal normalization and network transformation methods for RNA-seq.
• To provide recommendations for accurate RNA-seq coexpression network analysis.

Main Methods:

• Benchmarking of 36 distinct workflows for RNA-seq coexpression networks.
• Testing workflows on diverse datasets (large/homogenous, small/heterogeneous).
• Analysis of aggregate performance, method choices, and experimental factor impacts.

Main Results:

• Between-sample normalization significantly impacts network accuracy.
• Size factor-adjusted counts yield networks that best recapitulate gene functional relationships.
• Identified key methods for robust RNA-seq coexpression network construction.

Conclusions:

• Provides concrete recommendations for building accurate RNA-seq coexpression networks.
• Offers detailed results and guidance for researchers to choose appropriate methods.
• Enhances the utility of RNA-seq data for gene coexpression analysis.