MultiVERSE

MultiVERSE extends the VERSE framework to embed multiplex and multiplex-heterogeneous networks using Random Walks with Restart (RWR-M and RWR-MH) to generate node embeddings for downstream tasks such as link prediction, community detection, network reconstruction, and rare disease–gene association discovery.

Key Features:

• VERSE extension: Extends the VERSE framework to support embedding of multiplex networks composed of layers with shared nodes and distinct edge types.
• Multiplex-heterogeneous embedding: Embeds multiplex-heterogeneous networks containing multiple node and edge types to capture richer network representations.
• Random Walks with Restart (RWR): Implements Random Walks with Restart on Multiplex (RWR-M) and Multiplex-Heterogeneous (RWR-MH) networks to learn node embeddings.
• Scalability and speed: Designed for fast and scalable computation to handle large biological and social networks.

Scientific Applications:

• Community Detection and Node Classification: Produces lower-dimensional embeddings that facilitate community detection and node classification in complex networks.
• Link Prediction: Predicts potential connections between nodes to identify novel interactions or associations.
• Network Reconstruction: Infers missing links to reconstruct incomplete networks from existing data.
• Rare Disease–Gene Association Discovery: Applied to predict links and cluster genes associated with rare diseases in biomedical network analyses.

Methodology:

MultiVERSE leverages the VERSE framework and applies Random Walks with Restart on Multiplex (RWR-M) and Multiplex-Heterogeneous (RWR-MH) networks by simulating random walks that restart at their origin nodes to capture both local and global network structure.