manta: a Clustering Algorithm for Weighted Ecological Networks.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Microbial network inference and analysis are key for generating biological hypotheses from sequencing data.
• Network clustering is an essential step in microbial network analysis.

Purpose of the Study:

• To introduce manta, a novel heuristic network clustering algorithm for weighted networks.
• To address limitations of existing algorithms by exploiting negative edges and differentiating cluster assignment strength.

Main Methods:

• Manta is a heuristic algorithm that clusters nodes in weighted networks.
• It differentiates between weak and strong cluster assignments and exploits negative edges.
• The algorithm assesses the robustness of cluster assignments, enhancing resilience to noisy data.

Main Results:

• Manta equals or outperforms existing algorithms on noise-free synthetic data.
• It successfully identifies biologically relevant subcompositions in real-world datasets, such as cheese rind and ocean microbial communities.
• Manta identified taxa groups correlating with intermediate moisture content in cheese rinds.

Conclusions:

• Manta is a robust and versatile tool for identifying biologically informative groups within microbial networks.
• Its unique strengths include handling intermediate nodes, utilizing negative edges, and assessing robustness.
• The algorithm requires no parameter tuning, is easy to install and use, and integrates with existing tools.