Identifying disease-associated pathways in one-phenotype data based on reversal gene expression orderings

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Tissue biopsy invasiveness often limits the availability of sufficient normal controls for comparative studies.
• Analyzing gene expression data to identify disease-disrupted pathways is crucial for understanding disease mechanisms.
• Existing methods may struggle with datasets lacking adequate normal controls.

Purpose of the Study:

• To develop and validate a novel pathway enrichment tool, DRFunc, for detecting significantly disrupted pathways.
• To enable the incorporation of normal controls from external experiments when internal controls are insufficient.
• To provide a robust method for analyzing gene expression data across different cancer types and experimental platforms.

Main Methods:

• Developed the DRFunc algorithm, a pathway enrichment tool that utilizes differentially ranked (DR) gene pairs.
• Identified concordant DR gene pairs between cases and controls across independent microarray and RNA-seq datasets.
• Applied the DRFunc algorithm to analyze gene expression data, combining controls from different studies.

Main Results:

• Validated the DRFunc method using diverse cancer datasets (microarray and RNA-seq).
• Demonstrated high concordance of DR gene pairs across independent datasets.
• Successfully identified significant pathways in glioblastoma samples.
• Showcased the algorithm's ability to detect altered pathways in datasets with weak expression signals, such as chemotherapy-treated breast cancer.

Conclusions:

• DRFunc effectively detects significantly disrupted pathways by integrating external normal controls, addressing a critical limitation in disease research.
• The algorithm is robust and applicable to various cancer types and gene expression data (microarray, RNA-seq), even with weak expression signals.
• DRFunc offers a valuable tool for pathway enrichment analysis, particularly in scenarios with limited or no internal normal control samples.