HUGMi: Human Uro-Genital Microbiome database and hybrid classifier for improved species level annotation of 16S rRNA amplicon sequences

Abstract

The human urogenital microbiota encompasses commensal microorganisms harbouring the urinary and reproductive tracts, and has been associated with the pathophysiology of multiple urogenital disorders. Here, the Human Urogenital Microbiome (HUGMi) workflow is introduced, wherein a manually curated, niche-specific, 16S rRNA database is integrated with a novel hybrid classification algorithm to enhance amplicon-based species-level taxonomic resolution in urogenital microbiome investigations. The HUGMi database is curated through rigorous quality assessment of sequences, nomenclature correction of taxonomies and meticulous selection of bacterial species biologically significant to the human urogenital region. Quantitative evaluation of taxonomic assignment with mock communities demonstrates that HUGMi consistently achieved superior accuracy compared to conventional databases, with statistically significant improvement in F1 scores. Comparative analyses across three sets of biological data also reveal a robust and versatile performance, with both QIIME2 BLAST and sklearn-based taxonomic classifiers and with varied 16S regions. Following these outcomes the HUGMi hybrid classifier was developed that combines the specificity of BLAST-based methods with the probabilistic framework of learning-based classifiers. This approach enabled superior species-level identification through varying confidence thresholds. In contrast to the limited number of species identified in real-time studies, the integrated outcome of the HUGMi database and hybrid classifier, not just replicated the original results but simultaneously revealed a spectrum of previously unidentified species. Despite its crucial role in human health, the urogenital microbiome is comparatively understudied with respect to other body sites. The HUGMi pipeline addresses this empirical lacuna by identifying a substantially expanded range of species-level taxa. This enhanced resolution will not only augment downstream functional studies, but also aid in the investigation of specific bacterial influence on urogenital homeostasis and disease pathogenesis.