Benchmarking Denoising Strategies and Nucleic Acid Templates for ITS2 Amplicon Sequencing of Gut Mycobiome

The human gut mycobiome plays a critical role in host immunity, yet standardized workflows for fungal internal transcribed spacer (ITS) amplicon sequencing remain elusive. Unlike the bacterial 16S rRNA gene, where the impact of DNA versus RNA templates is well-documented, such comparative benchmarks are lacking for fungal studies. Furthermore, fungal ITS regions exhibit significant length heterogeneity, posing unique challenges for denoising algorithms that were originally designed for conserved bacterial amplicons. We performed fungal ITS2 sequencing on six fecal samples, generating 12 matched DNA and RNA-derived cDNA libraries for paired methodological benchmarking. We systematically compared three representative workflows, DADA2, Deblur, and VSEARCH, to evaluate read retention, feature recovery, and preservation of ITS2 length heterogeneity. To address dataset-size limitation, we further added an independent public fecal fungal ITS2 validation cohort from PRJNA419104, including 32 baseline CDI samples and 21 metadata-confirmed healthy controls, and reprocessed this dataset using the same comparative pipeline framework. AGI-versus-healthy comparisons in the internal cohort were treated as exploratory secondary analyses rather than disease-biomarker discovery. DADA2 best preserved fungal ITS2 length heterogeneity while avoiding the fixed-length truncation artifact observed with Deblur and the feature inflation typical of permissive OTU clustering. Paired DNA and RNA profiles showed concordant dominant community structure, with no ANCOM-detected differential abundance between templates across major taxonomic levels. Exploratory AGI-versus-healthy comparisons suggested candidate differential-abundance features, but these observations were considered hypothesis-generating and were not interpreted as validated disease biomarkers. In this small-sample paired methodological benchmark, DADA2 is the preferred workflow for fungal ITS2 amplicon analysis in this paired benchmark because it preserves variable-length sequence information. DNA sequencing provides a practical primary layer for routine dominant-community profiling, whereas RNA-based sequencing can serve as an optional complementary layer when low-abundance or activity-oriented features are of specific interest. The results presented in this article are limited by the small sample size, and larger clinical cohorts are required for disease-association or biomarker inference.