RandomWalkRestartMH

Multiplex-Heterogeneous Random Walk with Restart (RWR-MH): Network-based gene and disease association analysis

Multiplex-Heterogeneous Random Walk with Restart (RWR-MH) extends the Random Walk with Restart (RWR) algorithm to traverse multiplex and heterogeneous biological networks for integrated analysis of molecular interactions and disease relationships.

Key Features:

• Multiplex Network Capability: Performs RWR across multiple interaction layers, including protein-protein interactions and co-expression associations, enabling simultaneous multi-layer analysis.
• Heterogeneous Network Exploration: Integrates diverse node and edge types, including phenotype similarities between diseases and molecular interaction data, within a unified network framework.
• Leave-One-Out Cross-Validation: Implements systematic leave-one-out cross-validation to evaluate predictive performance for disease-associated gene identification.
• Comparative Performance Analysis: Demonstrates improved prediction of functional associations and disease-related genes relative to monoplex or single-layer heterogeneous RWR approaches.

Scientific Applications:

• Gene/Protein Function Inference: Infers gene or protein function by analyzing network proximity of seed nodes based on functional clustering principles.
• Disease Gene Prediction: Identifies candidate disease-associated genes, including applications to Wiedemann-Rautenstrauch syndrome and SHORT syndrome, through integrated multi-layer network analysis.

Methodology:

RWR-MH applies an adapted Random Walk with Restart algorithm that propagates probability across multiplex layers and heterogeneous node types, allowing inter-layer transitions to capture multi-source biological interactions within a unified probabilistic framework.