Background: The microbiome, a community of microbes that co-reside in biotic or abiotic environments, underlies biogeochemical cycling, plant and animal development, and human health. Increasing evidence shows that much of the role the microbiome plays is executed through complex interactions among microbes. Thus, network reconstruction has been increasingly used as a tool to disentangle internal workings within microbial communities.
Results: We developed a general framework for recovering microbial interaction networks from any design of microbial experiment. This framework represents a quasi-dynamic game model (qdGM) derived from the seamless integration of evolutionary game theory and allometric scaling laws. The qdGM can not only characterize how individual microbes act singly, but also reveal how different microbes interact with each other to govern microbial community assembly. Beyond existing approaches that can only identify a single overall microbial network from a number of samples, our framework can track and visualize how interaction networks vary from sample to sample and covert sample-specific (personalized) networks into context-specific networks. More importantly, this framework can reconstruct such mobile microbial networks from steady-state data, facilitating the widespread use of network tools to understand the impact of the microbiome on natural processes.
Conclusions: As proof of concept, we used the new framework to analyze human gut microbiota data and interspecific animal-associated microbiota data. Mobile networks reconstructed from each dataset can characterize previously unknown mechanisms that drive the change of microbial interaction architecture and organization along spatiotemporal gradients. This framework provides a tool to generate process-specific microbiome networks that can be readily translated into various biotechnological applications and evolutionary studies.