Figure Box2. Comparison of OTU-based (OTU-s ,
singletons removed) and ASV-based approaches for inferring taxonomic
richness using the ITS2 subregion, full ITS region, and ITS + 28S based
on the dataset of Furneaux et al. (2021), which was taxonomically
re-analysed following Tedersoo et al. (in press). (a) average number of
reads retained; (b) richness of OTUs/ASVs; (c) richness of fungal
orders; and (d) kingdom and phylum-level distribution of OTUs/ASVs
across the entire dataset. In (d), letters indicate statistically
significant differences among groups based on Scheffe PostHoc tests
following sample-wise testing of log-ratio transformed proportions using
two-way ANOVAs including the two factors marker length and matrix type -
OTU, OTUs without singletons (OTU-s) and ASV. We conclude that 1) the
ASV approach and OTU-based approach with singletons removed recover
lower proportions of non-Dikarya and non-fungal taxa compared with the
OTU-based approach including singletons; 2) analyses based on the ITS2
region alone revealed a higher proportion of OTUs/ASVs that could not be
identified to a fungal phylum or eukaryote kingdom compared with longer
marker fragments due to less taxonomic information; 3) longer markers
had fewer sequences passing the quality control and revealed relatively
more singletons, suggesting accumulation of artefacts across the entire
1500-base amplicon; and 4) the ASV approach performs suboptimally for
metabarcoding of fungi and particularly poorly for the full ITS region
and longer markers.