Microbiome Networks: A Systems Framework for Identifying Candidate Microbial Assemblages for Disease Management

Network models of soil and plant microbiomes provide new opportunities for enhancing disease management, but also challenges for interpretation. We present a framework for interpreting microbiome networks, illustrating how observed network structures can be used to generate testable hypotheses about candidate microbes affecting plant health. The framework includes four types of network analyses. "General network analysis" identifies candidate taxa for maintaining an existing microbial community. "Host-focused analysis" includes a node representing a plant response such as yield, identifying taxa with direct or indirect associations with that node. "Pathogen-focused analysis" identifies taxa with direct or indirect associations with taxa known a priori as pathogens. "Disease-focused analysis" identifies taxa associated with disease. Positive direct or indirect associations with desirable outcomes, or negative associations with undesirable outcomes, indicate candidate taxa. Network analysis provides characterization not only of taxa with direct associations with important outcomes such as disease suppression, biofertilization, or expression of plant host resistance, but also taxa with indirect associations via their association with other key taxa. We illustrate the interpretation of network structure with analyses of microbiomes in the oak phyllosphere, and in wheat rhizosphere and bulk soil associated with the presence or absence of infection by Rhizoctonia solani.