Association between a prognostic gene signature and functional gene sets

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Expression-based gene signatures are crucial for prognosis and class prediction but require functional interpretation.
• Existing tools like Gene Set Enrichment have limitations, excluding information from non-annotated genes.

Purpose of the Study:

• To develop novel computational strategies for functional interpretation of gene signatures.
• To overcome limitations of existing methods by incorporating non-annotated genes and biological context.

Main Methods:

• Utilizing hierarchical variable selection and gene association networks to link signature genes with functional gene sets (pathways, Gene Ontology).
• Relating patient-specific risk scores to pathway expression patterns using global tests.
• Applying GlobalAncova to investigate gene expression effects from score-mutation interactions.

Main Results:

• Applied methods to an Acute Myeloid Leukemia (AML) survival prediction dataset.
• Demonstrated significant associations between the gene signature and cancer-related pathways.
• Identified links between the signature-based risk score and developmental biological processes.

Conclusions:

• Proposed methods offer a deeper functional interpretation of gene signatures beyond simple annotation.
• These strategies provide enhanced biological insights into mechanisms reflected by gene signatures.
• The approach moves past annotation restrictions, offering a more comprehensive understanding.