Food web construction
Trophic groups and classes: We assigned the retrieved taxonomically annotated MOTUs to 10 trophic classes including autotrophs, decomposers, detritivores, phytophages or phytoparasites, plant mutualists, bacterivores, fungivores, omnivores, predators and zooparasites. We then refined those trophic classes by subdividing them into 55 trophic groups in total. These were defined by separating phylogenetic distant groups that could have a different set of prey/predators (e.g., bacterivorous mites vs. bacterivorous nematodes) and groups differing in their resources acquisition strategy (e.g., different types of mycorrhiza and saprotrophs). The taxonomic rank used to delimitate phylogenetic distant groups comprised Bacteria, Fungi, Protista, the different phyla within Metazoa, and the different classes or orders within Arthropoda and Annelida. (Appendix 2).
To facilitate the assignment of the taxonomically annotated MOTUs to trophic classes, we first built a knowledge graph integrating information about the trophic interactions and feeding habits of soil-associated consumers from multiple data sources using the ontology-based data integration pipeline described in Le Guillarme & Thuiller (2023). This trophic knowledge graph uses the NCBITaxon ontology and the Soil Food Web Ontology (SFWO)(Le Guillarme et al. 2023) as global schemas for reconciling taxonomic and semantic heterogeneities between the data sources. It provides a unified access to multisource trophic information across taxonomic groups and trophic levels. The list of data sources and details on the assignment criteria can be found in Appendix 2.
Building the metaweb: A metaweb is a potential network containing all trophic groups and their potential interactions (Dunne 2006). We added three basal resources to construct our metaweb: light (or chemical energy), plants, and organic matter. Trophic links were then added between groups based on their main feeding preferences. Therefore, plant mutualists and phytoparasites were linked with the plant resource, detritivores and decomposers were linked with the organic matter resource and autotrophs were associated with the light resource. Bacterivores were linked with all trophic groups containing bacteria, and fungivores with all trophic groups containing fungi. The trophic interactions of omnivores, predators and zooparasites, were determined through a literature review of dietary preferences of taxa representing over 90% of the read abundance of the group (Appendix 2).
Local food webs: From the metaweb, the composition and structure of the local soil food webs were deduced based on locally present trophic groups, i.e., local webs are subgraphs of the metaweb. This assumes that co-occurring groups interact locally as defined in the metaweb. We approximated local relative abundance of each trophic group using a double-transformation, where first, total read counts were transformed into proportions within the sample, and second, the resulting proportions were standardised by the largest proportion observed across samples for each trophic group (‘eDNA index’ in Kellyet al. 2019). The relative abundance of each trophic group varied thus between 0 (absent) to 1 (largest observed proportion), allowing to have a comparable measure across trophic groups. Abundance values of the basal resources (light, plant, organic matter) were set to non-null (0.001) to not affect diversity measurements. For trophic class abundance estimates, we summed group relative abundances.