CPR

CPR identifies prognostic genes by clustering heterogeneous patient samples with K-means and applying a modified PageRank on functional interaction (FI) networks weighted by cluster-specific gene expression to prioritize hub genes for cancer prognosis.

Key Features:

• Clustering for Heterogeneity Reduction: Uses K-means clustering to group data samples and reduce sample heterogeneity.
• Modified PageRank Algorithm: Applies a modified PageRank within each cluster to prioritize genes on functional interaction (FI) networks weighted by gene expression values specific to samples in each cluster.
• Hub Gene Selection: Identifies hub genes from the PageRank-prioritized lists as candidate prognostic biomarkers.
• Performance Superiority: Demonstrates improved performance over conventional feature selection techniques and other network-based prognostic gene selection methods when integrated with Random Forest algorithms.
• Cluster-Specific Prognostic Genes: Detects cluster-specific prognostic genes and associated enriched biological processes reflecting distinct aspects of tumor progression or oncogenesis per cluster.

Scientific Applications:

• Molecular mechanism elucidation: Identifies heterogeneous prognostic genes to support investigation of molecular mechanisms underlying cancer progression.
• Personalized therapeutics: Facilitates identification of cluster-specific biomarkers that can inform personalized therapeutic strategies and targeted treatments.
• Functional pathway analysis: Supports functional studies to explore distinct biological pathways and processes involved in oncogenesis across patient clusters.

Methodology:

K-means clustering of samples; weighting of functional interaction (FI) networks by cluster-specific gene expression; application of a modified PageRank within each cluster to prioritize genes; selection of hub genes as prognostic biomarkers; evaluation integrated with Random Forest algorithms.