MiCoDe

MiCoDe clusters microbial taxa into communities from taxonomic abundance data using a Bayesian weighted stochastic block model.

Key Features:

• Automatic Data Filtering: The software filters data automatically to address challenges of microbiome high-throughput sequencing data.
• High-Dimensionality Handling: The model accounts for high-dimensionality in microbiome sequencing data.
• Compositionality Correction: The method addresses the compositional nature of microbiome data.
• Zero Inflation Accommodation: The model handles zero inflation common in microbiome datasets.
• Nonlinearity Modeling: The approach accounts for nonlinearity in microbial community structure.
• Unsupervised Learning: MiCoDe employs a rigorous statistical unsupervised learning model for community detection.
• Data Transformation Options: Users can select from options regarding data transformation prior to analysis.
• Network Estimation: The tool provides selections for network estimation as part of the analysis workflow.
• Community Number Selection: Users specify the number of communities for the clustering analysis.
• Default Recommended Settings: The tool provides default settings based on the authors' recommendations.

Scientific Applications:

• Microbiome Community Detection: MiCoDe identifies communities of microbial taxa from taxonomic abundance networks in microbiome studies.

Methodology:

MiCoDe uses a Bayesian weighted stochastic block model. Users upload a csv file containing taxonomic abundance data with rows as samples and columns as taxa, then select options for data transformation, network estimation, and number of communities.