BISG

BISG identifies survival-related gene sets by applying biclustering with a rectified factor network to detect gene subsets whose expression correlates with patient survival.

Key Features:

• Rectified Factor Network (RFN)-based biclustering: Employs an RFN model to perform biclustering on gene subsets and reduce the combinatorial search space.
• Survival association testing: Correlates genes within each significant bicluster with patient survival using the log-rank test.
• Multi-sampling strategy: Implements a multi-sampling approach to robustly detect survival-related gene sets.
• Multiple biomarker detection: Detects multiple distinct survival-related gene sets that can serve as candidate biomarkers.
• Dataset scope: Applied analyses across three cancer types and systematically analyzed 12 cancer datasets.
• Gene family enrichment: Identifies enrichment of microRNA protein-coding host genes, zinc fingers C2H2-type, solute carriers, CD (cluster of differentiation) molecules, and ankyrin repeat domain containing genes.
• Pathway enrichment: Reports enrichment of identified gene sets in heme metabolism, apoptosis, hypoxia, and inflammatory response pathways.
• Comparative performance: Demonstrated superior performance in differentiating patient survival groups compared to GSAS and IPSOV.

Scientific Applications:

• Survival biomarker discovery: Identification of multi-gene biomarkers for differentiating patient survival groups in cancer.
• Cross-cancer survival analysis: Testing and validation of survival-related gene sets across multiple cancer types and datasets.
• Prognostic hypothesis generation: Generation of hypotheses for cancer prognosis and multi-gene targeted treatment strategies based on identified gene families and pathways.

Methodology:

Performs RFN-based biclustering on gene subsets and correlates genes within significant biclusters with patient survival using a log-rank test combined with a multi-sampling strategy.