Inferring microbial co-occurrence networks from amplicon data: a systematic evaluation.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Microbial communities exhibit complex interactions, often studied using 16S ribosomal RNA (16S rRNA) amplicon sequencing.
• Inferring microbial co-occurrence networks from 16S data involves multiple analytical steps with varying tools and parameters.
• The impact of these choices on the resulting network's robustness and uniqueness is not well understood.

Purpose of the Study:

• To systematically analyze the influence of different tools and parameters on microbial co-occurrence network construction from 16S rRNA data.
• To identify critical steps in the analysis pipeline that contribute significantly to network variance.
• To develop guidelines and consensus algorithms for generating robust microbial association networks.

Main Methods:

• Performed a meticulous step-by-step analysis of a 16S data to network inference pipeline.
• Evaluated the effect of various algorithms and parameter choices on co-occurrence network structure.
• Developed and benchmarked consensus network algorithms using mock and synthetic datasets.

Main Results:

• Different tool and parameter choices substantially affect the inferred microbial co-occurrence networks.
• Identified specific analytical steps that contribute most to network variance.
• Developed and validated consensus network algorithms for improved robustness.

Conclusions:

• Systematic analysis provides crucial insights into the factors influencing microbial network inference from 16S data.
• Guidelines are provided for selecting appropriate tools and parameters for robust network analysis.
• The MiCoNE tool and consensus algorithms facilitate comparative analyses and understanding of microbial community assembly.