Maximum entropy networks show that plant-arbuscular mycorrhizal fungi associations are anti-nested and modular.

There is uncertainty in whether there is a common pattern of nestedness and modularity in plant-arbuscular mycorrhizal (AM) fungi associations, partly because of limitations arising from the use of null models that randomly rewire the observed connections to test for non-random patterns in the network. Here, we overcome these limitations by generating null association matrices using maximum entropy network modelling, and specifically the bipartite binary configuration model (BiCM) with degree distributions as soft constraints. This was used to test the hypothesis that nestedness and modularity are prevalent in plant-AM fungi associations. In contrast to past findings, we found most plant-AM fungi associations were anti-nested and modular. This pattern was almost universal, being consistent across habitat types, multiple spatial scales, and multiple levels of plant node aggregation, from communities and species to populations. Anti-nestedness can easily emerge from modularity when network patterns are determined by the identity of the plant and AM fungal nodes. Our findings emphasize the need for experiments that test the factors that cause the observed network structure and how that structure determines the function and stability of plant-AM fungi association networks.