Constructing gene co-functional and co-regulatory networks from public transcriptomes using condition-specific...

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Gene co-expression networks (GCNs) are crucial for understanding gene relationships.
• Existing GCN methods struggle with batch effects and sample composition in public RNA-seq data.
• This limits their utility for cross-species comparative studies.

Purpose of the Study:

• To develop a robust GCN construction method for large-scale, public RNA-seq data.
• To improve the accuracy of gene function prediction and regulatory network inference.
• To enhance cross-species GCN conservation for comparative genomics.

Main Methods:

• Introduced TEA-GCN (two-tier ensemble aggregation-GCN), a novel GCN construction approach.
• Utilized unsupervised transcriptomic dataset partitioning and multi-metric co-expression scoring.
• Leveraged natural language processing for biologically relevant dataset partitioning and explainability.

Main Results:

• TEA-GCN demonstrated superior performance over state-of-the-art methods across 12 species.
• Achieved enhanced accuracy in predicting gene functions and inferring gene regulatory networks.
• Identified tissue-/condition-specific co-expression patterns with high explainability.
• Showcased improved cross-species conservation of constructed GCNs.

Conclusions:

• TEA-GCN offers a robust and scalable solution for building high-quality GCNs from public RNA-seq data.
• The method enhances biological discovery through improved prediction and inference capabilities.
• TEA-GCN facilitates cross-species comparative transcriptomic analyses.