Network subgraph-based approach for analyzing and comparing molecular networks

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Molecular networks, crucial for biological functions, are complex systems with feedback interactions.
• Existing methods for network comparison often rely on randomization or struggle with embedded substructures.
• Understanding molecular network similarity is key to deciphering biological processes and disease mechanisms.

Purpose of the Study:

• To introduce a novel, alignment-free network subgraph-based approach for measuring molecular network similarity.
• To quantify network similarity using subgraph frequency distributions and Jensen-Shannon entropy.
• To demonstrate the method's effectiveness in classifying and clustering diverse molecular networks.

Main Methods:

• Proposed a network subgraph-based approach, distinct from network motif and graphlet methods.
• Quantified network similarity by comparing subgraph frequency distributions using Jensen-Shannon entropy.
• Applied the method to cancer, signal transduction, and cellular process networks.

Main Results:

• The subgraph-based approach achieved 100% accuracy in classifying six network models.
• Identified common regulatory modules in cancers (e.g., AML, pancreatic, gastric, hepatocellular carcinoma).
• Demonstrated that irreducible subgraphs dominate the underlying substructures of molecular networks.

Conclusions:

• The proposed information-theoretic approach accurately determines structural similarity irrespective of network size or node identity.
• The method effectively clusters molecular networks with similar regulatory topologies.
• Provides a systematic framework for analyzing, comparing, and classifying molecular networks across various functionalities.