Network-Based Enriched Gene Subnetwork Identification: A Game-Theoretic Approach.

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Identifying gene subsets for cancer etiology, progression, or therapy response is challenging due to biological complexity and limited sample sizes.
• Purely data-driven methods often yield high false positives and unclear biological relationships.
• Existing integrative models are limited by predefined pathway structures, missing novel interactions and crosstalk.

Purpose of the Study:

• To propose a novel coalition-based game-theoretic approach for identifying biologically relevant gene subnetworks associated with disease phenotypes.
• To overcome limitations of existing methods in discovering novel genomic interactions and pathway crosstalk.

Main Methods:

• A coalition-based game-theoretic algorithm is developed, starting from seed genes and traversing a protein-protein interaction network.
• Modulated subnetworks are identified using Shapley value, considering individual and collective gene utility.
• The approach is applied to ovarian cancer therapy response and triple-negative breast cancer immune infiltration.

Main Results:

• The proposed method demonstrates improved predictive power compared to state-of-the-art feature selection techniques.
• Novel potentially functional gene subnetworks associated with disease phenotypes were identified.
• The findings offer insights into the mechanisms underlying cancer progression.

Conclusions:

• The coalition-based game-theoretic approach effectively identifies biologically relevant gene subnetworks.
• This method enhances predictive accuracy and discovers novel subnetworks, advancing our understanding of cancer biology.
• The approach provides a powerful tool for cancer research, aiding in the discovery of new therapeutic targets and biomarkers.