Pathway-based personalized analysis of cancer

Area of Science:

• Bioinformatics
• Computational Biology
• Cancer Genomics

Background:

• Gene expression data is crucial for understanding cancer biology.
• Existing methods may not effectively capture pathway-level alterations.
• A need exists for robust algorithms to translate gene data into actionable biological insights.

Purpose of the Study:

• To introduce Pathifier, an algorithm for inferring pathway deregulation scores from gene expression data.
• To provide a context-specific, sample-level representation of pathway activity.
• To demonstrate Pathifier's utility in cancer research.

Main Methods:

• Developed Pathifier algorithm to transform gene-level expression data into pathway-level deregulation scores.
• Applied the algorithm to multiple colorectal cancer and glioblastoma multiforme datasets.
• Validated the reproducibility and information preservation of the multipathway representation.

Main Results:

• Pathifier generates reproducible and informative pathway-based sample representations.
• Identified significant pathways associated with glioblastoma patient survival.
• Discovered two pathways (CXCR3-mediated signaling, oxidative phosphorylation) predictive of colorectal cancer survival.
• Identified novel glioblastoma and colon cancer subclasses with distinct survival outcomes.

Conclusions:

• Pathifier effectively infers pathway deregulation scores, offering a compact and biologically relevant representation of tumor samples.
• The algorithm facilitates the discovery of novel biomarkers and cancer subtypes linked to patient survival.
• Pathifier enhances our understanding of cancer heterogeneity and provides a foundation for personalized medicine approaches.