HyMM

HyMM predicts disease-associated genes by leveraging multiscale module structures in biological networks to improve identification of genes implicated in complex human diseases.

Key Features:

• Multiscale Module Extraction: Extracts multiscale modules from biological networks using three multiscale-module-decomposition algorithms: Modularity Optimization (MO), Agglomerative Shelling (AS), and Hierarchical Clustering (HC).
• Module Resolution Analysis: Performs statistical analysis showing that functional consistency improves with increased resolution across multiscale modules, indicating hierarchical functional relationships.
• Gene Ranking: Ranks genes based on their associations with the extracted multiscale modules.
• Ranking Integration: Integrates multiple gene rankings via a hybrid approach that combines algorithms including Random Walk with Restart (RWR) and its hierarchical variant (RWRH).
• Performance Validation: Validates disease-gene predictions using 5-fold cross-validation and independent testing, reporting superior performance for MO combined with RWR or RWRH.

Scientific Applications:

• Disease-Gene Prediction: Predicts genes associated with complex human diseases by integrating multiscale module information from biological networks.
• Disease-Module Analysis: Analyzes the modular organization of disease-associated network regions to inform understanding of disease mechanisms and potential therapeutic targets.

Methodology:

Multiscale-module decomposition using MO, AS, and HC; statistical analysis of module functional consistency across resolutions; gene ranking from module associations; integration of rankings using Random Walk with Restart (RWR) and RWRH; performance assessment via 5-fold cross-validation and independent testing.