Common cervicovaginal sequencing methods result in discordant molecular diagnoses of bacterial vaginosis and reveal strain level effects of Gardnerella vaginalis

Bacterial vaginosis (BV) is associated with HIV transmission and pre-term birth, yet the etiology of BV remains unknown. Our analysis addressed that knowledge gap by comparing diagnostic techniques and using Bayesian inference to find species-specific associations with clinical indicators. We also assessed the effect of sequencing methodology on the results of molecular BV profiling. We observed significant differences in microbial diversity within BV-associated CSTs based on clinical diagnosis. CST assignments were substantially influenced by sequencing methodology, with concordance between methods as low as 59% for metatranscriptomic and metataxonomic-based CST assignment. We also found that <i>Gardnerella</i> has a strain-dependent association with individual Amsel's criteria, and that <i>Dialister micraerophilus</i> and <i>Parvimonas micra</i> are positively associated with Amsel's criteria while <i>Lactobacillus</i> is negatively associated. These results highlight the challenge of characterizing a condition without a single etiological agent, reinforcing the need for more granular diagnoses and treatments that are sensitive to BV variability.