Network-induced classification kernels for gene expression profile analysis.

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Gene expression profiling is crucial for classifying disease phenotypes.
• Current methods face limitations in robustness, accuracy, and biological interpretation.
• Integrating protein-protein interaction (PPI) networks with gene expression data may enhance classification.

Purpose of the Study:

• To investigate the relevance of PPI networks in gene expression-based disease classification.
• To evaluate existing methods that integrate expression and interaction data.
• To develop a novel, efficient method for integrating network and expression data for improved classification.

Main Methods:

• Analysis of co-expressed genes in relation to their proximity within PPI networks.
• Evaluation of two existing computational methods for integrating gene expression and PPI data.
• Development and application of a new kernel method, NICK (Network-Integrated Classification Kernel), for Support Vector Machine (SVM) classification.

Main Results:

• Co-expressed genes are indeed closer in PPI networks, confirming the relevance of interaction data.
• The performance improvement of one existing method was found to be independent of network topology, while another method's improvement was network-dependent.
• The NICK method demonstrated superior classification performance compared to extant methods.
• NICK achieved this improved performance with a computational speed two orders of magnitude faster than existing approaches.

Conclusions:

• Protein interaction networks provide valuable biological context for gene expression-based disease classification.
• The NICK method offers a computationally efficient and accurate approach for integrating gene expression and network data.
• NICK enhances the robustness and biological interpretability of disease phenotype classification.