Reference Database Selection Influences Both Taxonomic Annotation and Ecological Inference in Arbuscular Mycorrhizal Fungal Communities

Abstract

Arbuscular mycorrhizal (AM) fungi are keystone microorganisms in soil ecosystems, and their community composition plays a crucial role in regulating ecosystem functions. Although high-throughput sequencing is widely used to characterize their communities, the selection of reference databases for sequence annotation remains inconsistent. To assess database performance at the species level, rhizosphere soil and root samples from four plant species with contrasting communities were subjected to amplicon sequencing. Taxonomic annotations generated using the MaarjAM, SILVA, and NCBI databases were compared. NCBI provided the highest taxonomic resolution, identifying 18 genera and 108 species, including Rhizophagus intraradices and Funneliformis mosseae, with assignments supported by spore isolation and nomenclatural verification. In contrast, MaarjAM recovered the most ASVs but contained numerous virtual taxa, whereas SILVA showed limited annotation depth and fewest assignments. Comparative analyses showed that NCBI was more sensitive in distinguishing samples based on alpha diversity and community structure and produced species distributions more consistent with ecological expectations. Although core metabolic profiles were conserved across databases, suggesting functional redundancy that may contribute to ecosystem stability, database selection influenced relationship interpretation between environmental variables and communities. Overall, NCBI is recommended for species-level annotation and community analysis, whereas inappropriate database selection may bias biodiversity assessment and ecological inference.