Random Matrix Analysis for Gene Interaction Networks in Cancer Cells

Area of Science:

• Computational Biology
• Systems Biology
• Network Science

Background:

• Understanding cancer requires investigating topological changes in cellular gene interaction networks.
• Current theoretical frameworks for analyzing these complex networks are insufficient.
• Predicting the behavior of large-scale networks from smaller subnetworks is a critical challenge.

Purpose of the Study:

• To investigate the topological uniqueness of gene interaction networks in human cancer cells.
• To apply random matrix theory to analyze gene network properties.
• To correlate network properties with cancer cell behavior.

Main Methods:

• Utilized random matrix theory to study the distribution of nearest neighbor level spacings (P(s)) of gene interaction matrices.
• Employed the Cancer Network Galaxy (TCNG) database, inferring interactions via a Bayesian network model.
• Analyzed 256 NCBI GEO entries of gene expression data from human cancer cells.

Main Results:

• Observed a Wigner distribution for P(s) in dense gene networks (over ~38,000 edges).
• Found a Poisson distribution for P(s) in sparse gene networks (fewer edges).
• Investigated the relationship between P(s), network sparseness, and edge frequency (reliance of inferred interactions).

Conclusions:

• The distribution of nearest neighbor level spacings in gene interaction networks is dependent on network density.
• This finding offers a theoretical approach to predict complex cancer network behavior based on network topology.
• The study highlights the importance of network structure in understanding cancer at a molecular level.