Microbial interactions: from networks to models

• Published in: Nature reviews. Microbiology 2012-08, Vol.10 (8), p.538-550
• Main Authors: Faust, Karoline, Raes, Jeroen
• Format: Article
• Language: English
• Subjects: 41; 631/158/2463; 631/326/2565; 631/553/2393; Bacteria; Biological models; Biomedical and Life Sciences; Biota; Biotic communities; Competition; Cooperation; Ecosystem structure; Ecosystem studies; Ecosystems; Humans; Infectious Diseases; Life Sciences; Medical Microbiology; Metabolism; Metagenomics - methods; Microbial Interactions; Microbiology; Microorganisms; Mutualism; Oceans; Parasitology; review-article; Virology
• ISSN: 1740-1526; 1740-1534
• DOI: 10.1038/nrmicro2832

Key Points Microorganisms form various ecological relationships, ranging from mutualism to competition, that in addition to other factors (such as niche preferences and random processes) shape microbial abundances. Recently, network inference techniques have frequently been applied to microbial presence–absence or abundance data to detect significant patterns of co-presence and mutual exclusion between taxa and to represent them as a network. In addition to predicting links between taxa and between environmental traits and taxa, the analysis of microbial association networks reveals niches, points out keystone species and indicates alternative community configurations. However, several pitfalls in the construction and interpretation of these networks exist, ranging from data normalization to multiple test correction. Thorough evaluation is needed to determine the best-performing network inference technique. Recent advances in the cultivation of unknown microorganisms, combinatorial labelling and parallel cultivation may soon allow systematic co-culturing and perturbation (that is, species removal) experiments. Interaction strengths that have been obtained from static networks or that have been measured experimentally can serve as inputs for dynamic models of microbial communities, which in turn can simulate the behaviour of the system in various conditions. In the long run, dynamic models could help to engineer microbial communities. The theory of dynamic systems can contribute to our understanding of microbial communities. For instance, alternative community states can arise as a consequence of system dynamics without being driven by environmental differences. Correlation and co-occurrence patterns found in metagenomic and phylogenetic data sets are increasingly being used to predict species interactions in the environment. Here, Faust and Raes describe the range of approaches for predicting microbial network models, the pitfalls that are associated with each approach and the future for developing ecosystem-wide models. Metagenomics and 16S pyrosequencing have enabled the study of ecosystem structure and dynamics to great depth and accuracy. Co-occurrence and correlation patterns found in these data sets are increasingly used for the prediction of species interactions in environments ranging from the oceans to the human microbiome. In addition, parallelized co-culture assays and combinatorial labelling experiments allow high-throughput discovery of cooper