Analyse multiple disease subtypes and build associated gene networks using genome-wide expression profiles

Area of Science:

• Bioinformatics
• Genomics
• Systems Biology

Background:

• Transcriptomic studies are increasing, but integrative approaches are needed to separate multiple pathological states.
• Robust gene marker selection for disease subtypes and understanding gene relationships remain challenging.

Purpose of the Study:

• To present a network-oriented, data-driven bioinformatic approach for gene-disease association analysis.
• To identify gene sets, minimal differentiating subsets, and their discriminant power for disease subtypes.
• To uncover gene relationships within specific disease subtypes.

Main Methods:

• Developed a network-oriented, data-driven bioinformatic approach using genome-wide expression data (microarrays, RNA-Seq).
• Implemented the approach in an R package named geNetClassifier, available on Bioconductor.
• Applied the tool to two independent leukemia datasets with different subtypes.

Main Results:

• Identified key deregulated genes associated with specific leukemia subtypes.
• Demonstrated the selection of similar genes across independent datasets and analysis methods (arrays vs. RNA-Seq).
• Validated the tool's ability to differentiate and classify disease subtypes using gene networks.

Conclusions:

• Gene networks incorporating gene-to-gene association, specificity, and discriminant power aid in creating gene-disease maps.
• The geNetClassifier tool offers an effective method for molecular characterization of diseases using genome-wide expression data.
• This approach facilitates unraveling molecular features characterizing specific pathological states.