NetCoMi

NetCoMi constructs, analyzes, and compares microbial association networks from high-throughput sequencing data to characterize interactions and differences among operational taxonomic units (OTUs) or taxa in microbiome studies.

Key Features:

• Network Construction: NetCoMi estimates associations between operational taxonomic units (OTUs) or taxa to construct microbial association networks from sequencing-derived count data.
• Data Normalization and Zero Handling: The package implements methods to handle zero values in compositional sequencing data and to apply normalization techniques required for accurate association estimation.
• Network Analysis: It provides tools to analyze single microbial networks for exploring taxa-level interactions, network structure, dynamics, and stability.
• Comparison of Networks: NetCoMi quantifies differences between microbial association networks across conditions (for example, healthy versus patients) to assess changes in network structure or specific taxa associations.
• Differential Network Construction: It constructs differential association networks to identify pairs of taxa with significant interaction differences between two groups.
• Dissimilarity Networks: NetCoMi creates and analyzes dissimilarity networks to summarize heterogeneity and overall community-structure differences across microbiome samples.

Scientific Applications:

• Microbial ecology: Characterizing taxa interactions and community structure within environmental and host-associated microbiomes.
• Human health and disease: Comparing microbial association networks between healthy individuals and patients to investigate disease-related community changes.
• Biomarker discovery: Identifying taxa pairs or network features with differential associations as potential biomarkers or mechanistic leads.
• Environmental microbiome studies: Applying network construction and comparison to datasets such as the GABRIELA study of settled dust samples from children’s rooms in Ulm and Munich to assess spatial or study-center differences.

Methodology:

Computational steps explicitly include handling zeros in compositional data, data normalization, estimation of associations, construction of differential association networks, creation of dissimilarity networks, and integration of statistical techniques within reproducible computational workflows.