Seed-Guided Subnetwork Discovery in Disease Mechanisms

Identifying disease-relevant subnetworks within large biological networks such as protein-protein interaction (PPI) graphs is fundamental to uncovering disease mechanisms and biomarkers. Traditional algorithms often suffer from local optima due to greedy search strategies or uninformative scoring, limiting their ability to robustly identify biologically meaningful modules.

In this study, we introduce a seed-guided subnetwork discovery algorithm, a novel method inspired by network propagation but optimized for targeted subnetwork expansion. The approach begins with a disease-relevant seed gene and iteratively incorporates neighboring genes based on their biological relevance, which can be informed by gene expression, mutation frequency, connectivity, or weighted edges.

This seed-based expansion provides multiple advantages: it bypasses local maxima, maintains computational tractability, and leads to more interpretable subnetworks by focusing analysis on high-relevance regions of the network. The method is flexible in its selection criteria, allowing integration of diverse omics data and biological constraints to tailor the subnetwork search to specific research questions.

Availability: Source code and supplementary materials will be made available upon publication.

Presented at: RECOMB CCB 2025 – Seoul, South Korea

Read the full paper: [coming soon]