A Multi-Cohort and Multi-Omics Meta-Analysis Framework to Identify Network-Based Gene Signatures

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Interpreting vast molecular data for disease biomarkers is challenging.
• Existing methods often neglect intermolecular interactions, multi-omics data, and cohort heterogeneity.
• This limits the accurate capture of complex disease biology, such as cancer.

Purpose of the Study:

• To develop a novel multi-cohort, multi-omics meta-analysis framework.
• To identify robust molecular subnetworks reflecting biological conditions.
• To overcome limitations of existing biomarker discovery approaches.

Main Methods:

• Integrated multiple independent gene expression studies.
• Incorporated unmatched DNA methylation data.
• Utilized protein-protein interaction networks to identify methylation-driven subnetworks.

Main Results:

• Developed a framework to address limitations in molecular data analysis.
• Constructed disease-specific subnetworks for glioblastoma and low-grade gliomas.
• Validated subnetworks by predicting patient clinical outcomes across cohorts.

Conclusions:

• The proposed framework successfully identifies robust, biologically relevant molecular subnetworks.
• This approach enhances biomarker discovery for complex diseases.
• The identified subnetworks demonstrate predictive power for patient outcomes.