Differential analysis of biological networks

Area of Science:

• Computational Biology
• Bioinformatics
• Network Analysis

Background:

• Comparing gene expression or DNA methylation networks between healthy and diseased individuals can reveal disease-associated biological pathways.
• Cancer progression involves localized rewiring of signaling and control networks, necessitating methods to detect disrupted interaction patterns.
• Current statistical procedures for two-network comparisons lack scalability for detecting localized topological differences.

Purpose of the Study:

• To introduce a scalable statistical methodology for detecting differential interaction patterns between two networks.
• To develop an algorithm that can identify localized topological differences indicative of disease states.
• To provide a tool for discovering novel molecular diagnostic and prognostic signatures.

Main Methods:

• Propose the dGHD algorithm, utilizing the Generalised Hamming Distance (GHD) statistic.
• Employ a non-parametric permutation testing framework for statistical significance assessment.
• Achieve computational efficiency through an asymptotic normal approximation.

Main Results:

• The GHD statistic detects more subtle topological differences than the standard Hamming distance.
• The dGHD algorithm demonstrates high performance in simulation studies, measured by sensitivity and specificity.
• The methodology successfully identified differential DNA co-methylation subnetworks associated with ovarian cancer.

Conclusions:

• The dGHD algorithm is effective in detecting subtle network topological differences.
• The proposed methodology shows potential for discovering network-derived biomarkers for traits of interest, such as cancer.
• This approach can advance the identification of molecular signatures for cancer diagnosis and prognosis.