A framework for comparing microbial networks reveals core associations

Microbial network construction and analysis is an important tool in microbial ecology. As microbial interactions are challenging to infer experimentally, such networks are often constructed from statistically inferred associations and may not represent ecological interactions. Hence, microbial association networks contain a large number of errors and their derived properties do not necessarily reflect true community structure. Such errors can be identified with the use of appropriate null models. We have developed anuran, a toolbox for investigation of noisy networks with null models, for identification of non-random patterns in groups of association networks. This toolbox compares multiple networks to identify conserved subsets (core association networks, CANs) and other network properties that are shared across all networks. Such groups of networks can be generated from a collection of time series data or from cross-sectional sample sets. We use data from the Global Sponge Project to demonstrate that different orders of sponges have a larger CAN than expected at random.