Cautionary notes on the use of co-occurrence networks in soil ecology

Soil ecology is witnessing exponential growth in the number of studies using co-occurrence network analysis. Researchers reconstruct networks based on the co-occurrence of taxa or genes across soil samples at a wide range of geographic scales from single aggregates to the whole planet and taxonomic scopes, some studies targeting specific taxa or guilds to others surveying the whole microbiome as well as micro-and mesofauna. Co-occurrence networks can be very useful to extract simple patterns from complex datasets. Applications include the detection of abiotic and biotic factors that determine community structure, the identification of keystone taxa and their relationship to specific soil functions, and the inference of mechanisms of community assembly. However, networks are more and more often misused and serve as mere graphic tools with no attempt at hypothesis testing. In this perspectives article, we first review the main usage of co-occurrence network analysis in soil ecology during the last decade. We then discuss the applications and caveats of network analysis in soil ecology, leaving apart strictly methodological aspects of network reconstruction, which is beyond the focus of this article. Finally, we include recommendation guidelines - such as the possibility of informing networks with geographic, environmental and/or phylogenetic information - with the hope that this will facilitate network analysis to become a useful tool that helps elucidate meaningful patterns in soil ecology.