Inferring pathway crosstalk networks using gene set co-expression signatures

Area of Science:

• Molecular Biology
• Systems Biology
• Bioinformatics

Background:

• Understanding cellular mechanisms requires constructing molecular interaction networks.
• Existing gene set methods often treat entire pathways, overlooking specific gene subset interactions.
• Pathway crosstalk is biologically known to involve specific gene subsets, not whole pathways.

Purpose of the Study:

• To develop a novel method for inferring pathway crosstalk networks using gene expression data.
• To identify specific subsets of genes (signatures) driving pathway co-expression.
• To provide a tool for simultaneous analysis at pathway and gene levels for complex diseases.

Main Methods:

• Developed signature-based gene set co-expression analysis (sGSCA).
• Applied sparse canonical correlation analysis (sCCA) to measure pathway co-expression and identify signature genes.
• Validated on simulated datasets and applied to hepatocellular and lung cancer gene expression data.

Main Results:

• sGSCA efficiently detected pathway crosstalk and relevant signature genes in simulated data.
• Applied to cancer datasets, sGSCA identified significant pathway crosstalks.
• Identified signature genes were highly enriched for cancer-related genes.

Conclusions:

• sGSCA offers a novel approach to infer pathway crosstalk networks from large-scale gene expression data.
• The method effectively identifies key gene subsets contributing to pathway interactions.
• sGSCA is a valuable tool for dissecting complex disease mechanisms at multiple molecular levels.