Comparison of commonly used software pipelines for analyzing fungal metabarcoding data

Metabarcoding targeting the internal transcribed spacer (ITS) region is commonly used to characterize fungal communities of various environments. Given their size and complexity, raw ITS sequences are necessarily processed and quality-filtered with bioinformatic pipelines. However, such pipelines are not yet standardized, especially for fungal communities, and those available may produce contrasting results. While some pipelines cluster sequences based on a specified percentage of base pair similarity into operational taxonomic units (OTUs), others utilize denoising techniques to infer amplicon sequencing variants (ASVs). While ASVs are now considered a more accurate representation of taxonomic diversity for prokaryote communities based on 16S rRNA amplicon sequencing, the applicability of this method for fungal ITS sequences is still debated.

Here we compared the performance of two commonly used pipelines DADA2 (inferring ASVs) and mothur (clustering OTUs) on fungal metabarcoding sequences originating from two different environmental sample types (fresh bovine feces and pasture soil). At a 99% OTU similarity threshold, mothur consistently identified a higher fungal richness compared to DADA2. In addition, mothur generated homogenous relative abundances across multiple technical replicates (n = 18), while DADA2 results for the same replicates were highly heterogeneous.

Our study highlights a potential pipeline-associated bias in fungal metabarcoding data analysis of environmental samples. Based on the homogeneity of relative abundances across replicates and the capacity to detect OTUs/ASVs, we suggest using OTU clustering with a similarity of 97% as the most appropriate option for processing fungal metabarcoding data.