HetMatPy

HetMatPy provides Python-based storage, manipulation, and analysis of heterogeneous networks (hetnets) to identify and score significant paths connecting biomedical entities.

Key Features:

• Hetnet data model: Provides data structures and operations for storage, manipulation, and analysis of heterogeneous networks (hetnets).
• Hetionet implementation: Implements Hetionet, a hetnet interconnecting 11 node categories — including genes, diseases, drugs, pathways, and anatomical structures — via over 2 million edges spanning 24 relationship types.
• Hetnet connectivity search: Implements the hetnet connectivity search procedure to identify significant path types between any two nodes without requiring a supervised training set.
• Degree-aware enrichment: Identifies path types that occur more frequently than expected based solely on nodes' degrees.
• Path scoring: Scores and ranks individual paths, prioritizing those most specific to their path type.
• Precomputation and scalability: Includes optimizations and precomputation of significant instances of node connectivity to handle large-scale knowledge graphs.

Scientific Applications:

• Drug repurposing: Supports identification of candidate compound-disease relationships, exemplified by analyses such as metformin and breast cancer.
• Discovery of biological interactions: Enables exploration of complex biomedical connections, exemplified by investigations such as the GJA1 gene and insomnia.

Methodology:

Uses the hetnet connectivity search procedure that detects path types enriched relative to node-degree expectations, applies a scoring mechanism that evaluates and prioritizes individual paths by type-specificity, and employs precomputation and scalability optimizations for large knowledge graphs.