DeRegNet

DeRegNet identifies maximally deregulated subnetworks in directed biological networks using deregulation scores from multi-omics data to reveal disease-relevant modules.

Key Features:

• De Novo Subnetwork Identification: Identifies deregulated subnetworks directly from large-scale biological networks without relying on predefined pathways or gene sets.
• Deregulation Score-Based Analysis: Uses deregulation scores derived from multi-omics data mapped onto directed graphs to localize regions of significant deregulation.
• Probabilistic Model and Optimization: Frames subgraph identification as a probabilistic model interpreted as maximum likelihood estimation and formulates the problem as a fractional integer programming combinatorial optimization.
• Performance and Validation: Comparative analyses on simulated datasets with known ground truths and application to a publicly available liver cancer dataset demonstrate improved identification of deregulated subnetworks and enable patient stratification.

Scientific Applications:

• Hypothesis Generation in Disease Research: Identifies candidate molecular subnetworks implicated in disease mechanisms, including cancer-related processes.
• Patient Stratification: Supports stratification of patients based on patterns of subnetwork deregulation identified from multi-omics data, as shown in liver cancer analyses.
• Contextualization of Multi-Omics Data: Places omics-derived deregulation scores into the context of biomolecular networks to prioritize functionally coherent modules.

Methodology:

Maps deregulation scores from multi-omics data onto directed graphs, models de novo subgraph identification via a probabilistic maximum likelihood framework, and solves the resulting fractional integer programming combinatorial optimization.