An Entropy-Based Directed Random Walk for Cancer Classification Using Gene Expression Data Based on Bi-Random Walk on...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Microarray technology and machine learning are crucial for disease prediction and risk gene discovery.
• Traditional methods overlook pathway network structure, treating pathways as simple gene sets.

Purpose of the Study:

• To propose an entropy-based directed random walk (e-DRW) method for inferring pathway activities.
• To enhance pathway analysis by incorporating network structure and improved gene weighting.

Main Methods:

• Developed an e-DRW method with two key enhancements: expanded human pathway information coverage and refined gene weighting using correlation coefficients and t-test scores.
• Utilized gene expression datasets as input and pathway datasets to construct directed graphs for analysis.

Main Results:

• The e-DRW method demonstrated robust and superior classification accuracy compared to existing methods.
• Experiments confirmed the robustness of predicted risk-active pathways.

Conclusions:

• The e-DRW method significantly improves prediction performance in disease pathology.
• Successfully identified topologically important pathways and genes specific to cancer types.