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

Microbes tend to organize into communities consisting of hundreds of species involved in complex interactions with each other. 16S ribosomal RNA (16S rRNA) amplicon profiling provides snapshots that reveal the phylogenies and abundance profiles of these microbial communities. These snapshots, when collected from multiple samples, have the potential to reveal which microbes co-occur, providing a glimpse into the network of associations in these communities. The inference of networks from 16S data is prone to statistical artifacts. There are many tools for performing each step of the 16S analysis workflow, but the extent to which these steps affect the final network is still unclear. In this study, we perform a meticulous analysis of each step of a pipeline that can convert 16S sequencing data into a network of microbial associations. Through this process, we map how different choices of algorithms and parameters affect the co-occurrence network and estimate steps that contribute most significantly to the variance. We further determine the tools and parameters that generate the most accurate and robust co-occurrence networks based on comparison with mock and synthetic datasets. Ultimately, we develop a standardized pipeline (available at https://github.com/segrelab/MiCoNE) that follows these default tools and parameters, but that can also help explore the outcome of any other combination of choices. We envisage that this pipeline could be used for integrating multiple data-sets, and for generating comparative analyses and consensus networks that can help understand and control microbial community assembly in different biomes.ImportanceTo understand and control the mechanisms that determine the structure and function of microbial communities, it is important to map the interrelationships between its constituent microbial species. The surge in the high-throughput sequencing of microbial communities has led to the creation of thousands of datasets containing information about microbial abundances. These abundances can be transformed into networks of co-occurrences across multiple samples, providing a glimpse into the structure of microbiomes. However, processing these datasets to obtain co-occurrence information relies on several complex steps, each of which involves multiple choices of tools and corresponding parameters. These multiple options pose questions about the accuracy and uniqueness of the inferred networks. In this study, we address this workflow and provide a systematic analysis of how these choices of tools and parameters affect the final network, and on how to select those that are most appropriate for a particular dataset.