Inferring microbial interaction networks based on consensus similarity network fusion.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• High-throughput metagenomic sequencing generates vast amounts of microbial community data.
• Existing methods for inferring microbial interaction networks often analyze single datasets, limiting their scope.
• Data integration offers a powerful approach to enhance the accuracy and comprehensiveness of microbial network analysis.

Purpose of the Study:

• To develop and validate a novel data integration method for inferring microbial species relations.
• To improve the identification of microbial interaction networks by leveraging multiple datasets.
• To introduce a modified similarity network fusion (SNF) approach, termed consensus SNF (CSNF), incorporating consensus k-nearest neighbors (Ck-NN).

Main Methods:

• The study proposes a similarity network fusion (SNF) method for microbial relation inference.
• SNF integrates species similarities from different datasets via a cross-network diffusion process.
• A consensus k-nearest neighbors (Ck-NN) approach is integrated into SNF, creating the CSNF method.

Main Results:

• The CSNF method successfully integrates microbial data from multiple sources.
• Application of CSNF on genus profiles from three microbiome datasets revealed a modular structure.
• This modular structure, indicative of microbial interactions, was not identifiable using single-dataset analyses.

Conclusions:

• The CSNF method effectively augments species relationships by aggregating evidence from diverse datasets.
• CSNF leverages data complementarity to provide a more robust understanding of microbial community interactions.
• The findings highlight the utility of data integration for discovering complex microbial interaction network structures.