Vaginal microbiota is closely associated with women's health, however, the correlation between HPV-related cervical disease (HRCD) and vaginal microbiota is still obscure. In this study, patients with HRCD (n = 98) and healthy controls (n = 58) in Hangzhou were recruited, and their vaginal microbiota were collected and analyzed. The composition of the vaginal microbial community was explored, and a disease classification model was developed by random forest algorithm. The results suggested that the diversity of vaginal microbiota was significantly higher in HRCDs than that in healthy controls (p < 0.05). Firmicutes is the dominant phylum in vaginal microbiota, and Lactobacillus was identified as the most altered genus between two groups (p < 0.01). Kyoto Encyclopedia of Genes and Genomes analysis suggested the difference in vaginal microbial community functions between two groups. Furthermore, we identified 10 biomarkers as the optimal marker sets for the random forest model and found a higher probability of disease values in HRCD group in discovery cohort (p < 0.0001), with an area under the receiver operating characteristic curve reaching 89.7% (95% confidence interval: 78.3%-100%). We further validated the model in both validation and independent diagnosis cohorts, confirming its accuracy in the prediction of HRCD. In conclusion, this study revealed the community composition of vaginal microbiota in HRCDs and successfully constructed a diagnostic model for HRCD.
Keywords: HPV-related cervical disease; community composition; disease prediction; human papillomavirus; random forest; vaginal microbiota.
© 2022 Wiley Periodicals LLC.