RLasso-Cox

RLasso-Cox integrates gene interaction topological weights into a reweighted Lasso-Cox proportional hazards model to improve survival prediction and identify prognostic gene sets in cancer.

Key Features:

• Integration of Gene Interaction Information: Incorporates gene interaction network information to weight genes according to their topological importance.
• Reweighted Lasso-Cox Model: Implements a reweighted Lasso-Cox proportional hazards model (attributed to Wei Liu) that assigns differential penalties to genes based on network-derived weights.
• Random Walk Algorithm for Topological Weight Evaluation: Uses a random walk on the gene interaction network to evaluate topological significance of genes.
• Improved Prognostic Accuracy and Robustness: Demonstrates enhanced prognostic accuracy and robustness relative to conventional Lasso-Cox and other network-based methods on multiple cancer datasets.
• Identification of Small, High-Performance Gene Sets: Identifies compact gene sets with high prognostic performance on independent validation datasets.

Scientific Applications:

• Clinical cancer risk prediction: Improves survival risk stratification for cancer patients using gene expression and network information.
• Prognostic model development and validation: Facilitates construction of prognostic models that generalize to independent datasets.
• Biomarker discovery for personalized medicine: Enables identification of robust survival biomarkers and compact gene signatures for potential personalized treatment strategies.

Methodology:

Inputs high-dimensional gene expression data from cancer patients; applies a reweighted Lasso-Cox proportional hazards model that incorporates topological weights derived from a random walk on the gene interaction network; and performs extensive testing across multiple cancer datasets for validation.