MicNet toolbox

MicNet toolbox integrates dimensionality reduction, density-based clustering, and compositional co-occurrence network inference to analyze microbial community structure and interactions.

Key Features:

• UMAP (Uniform Manifold Approximation and Projection): Applies UMAP for dimensionality reduction to capture local patterns within microbial community data.
• HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise): Uses HDBSCAN for density-based clustering to identify community structure and noise.
• Enhanced SparCC (Sparse Correlations for Compositional data): Employs an enhanced version of SparCC optimized for larger datasets to estimate correlations in compositional microbiome data.
• Co-occurrence network construction: Constructs co-occurrence networks from inferred correlations to represent interactions among microbial taxa.
• Network theory analyses and structural balance metrics: Integrates network theory analyses, including structural balance metrics, to characterize network topology and interaction patterns.

Scientific Applications:

• Kombucha ASV network: Applied to a well-characterized kombucha biological network comprising 48 Amplicon Sequence Variants (ASVs).
• Simulated community networks: Evaluated on simulated community networks with predefined topologies to assess topology differentiation.
• Archean Domes large-scale dataset: Applied to an Archean Domes dataset featuring over 2,000 ASVs to demonstrate scalability and robustness.

Methodology:

Computational methods explicitly include UMAP for dimensionality reduction, HDBSCAN for clustering, an enhanced SparCC for estimating correlations in compositional data, and network theory analyses (including structural balance metrics) to construct and interpret co-occurrence networks.