Bioinformatics matters: The accuracy of plant and soil fungal community data is highly dependent on the metabarcoding pipeline

Abstract

Fungal communities associated with plants and soil influence plant fitness and ecosystem functioning. They are frequently studied by metabarcoding approaches targeting the ribosomal internal transcribed spacer (ITS), but there is no consensus concerning the most appropriate bioinformatic approach for the analysis of these data. We sequenced an artificial fungal community composed of 189 strains covering a wide range of Ascomycota and Basidiomycota, to compare the performance of 360 software and parameter combinations. The most sensitive approaches, based on the USEARCH and VSEARCH clustering algorithms, detected almost all fungal strains but greatly overestimated the total number of strains. By contrast, approaches using DADA2 to detect amplicon sequence variants were the most effective for recovering the richness and composition of the fungal community. Our results suggest that analyzing single forward (R1) sequences with DADA2 and no filter other than the removal of low-quality and chimeric sequences is a good option for fungal community characterization. • A mock community of 189 fungal strains was used to compare 360 pipelines. • Observed fungal richness depended strongly on the bioinformatic approach. • Community richness and composition were well recovered by DADA2. • Single forward (R1) reads often gave better results than assembled reads. • Bioinformatic approach selection should be based on study goals.