netANOVA

netANOVA performs distance-based hierarchical clustering of graphs and assesses the statistical significance of detected clusters to identify disease subtypes from individual-level biological networks for applications in precision medicine.

Key Features:

• Unsupervised Hierarchical Clustering: Employs an unsupervised hierarchical algorithm to compute similarities between networks using distance measures and group similar networks into classes.
• Optimal Cluster Determination: Determines the optimal number of clusters by recursively testing inter-group distances inspired by distance-wise ANOVA algorithms.
• Significance Assessment via Permutation Testing: Evaluates the statistical significance of observed heterogeneity by permuting between-object distance matrices.
• Flexibility and Adaptability: Provides multiple selectable options tailored to specific contexts and network types for analysis.

Scientific Applications:

• Simulation Performance: Demonstrates high performance in simulation scenarios while controlling type I error rates.
• Benchmarking: Ranks among top performers compared to state-of-the-art methods on synthetic and real datasets.
• Disease Subtype Identification: Identifies disease subtypes from individual-level biological networks to inform precision medicine, diagnosis, monitoring, and prevention programs.

Methodology:

Similarity computation using distance measures, unsupervised hierarchical clustering, recursive inter-group distance testing for optimal cluster determination, and permutation-based significance assessment of between-object distance matrices.