Abstract
The vaginal microbiota can be classified into five major community state types (CSTs) based on the bacterial content. However, the link between different CST subtypes and vaginal infection remains unclear. Here, we analyzed 2017 vaginal microbiota samples from women of a reproductive age with vaginal infections that were published in the last decade. We found that L. iners was the most dominant in 34.8% of the vaginal samples, followed by L. crispatus (21.2%). CST I was common in healthy individuals, whereas CST III and IV were associated with dysbiosis and infection. CST III-B, IV-A, IV-B, and IV-C0 were prevalent in patients with bacterial vaginosis (BV). Based on the relative abundance of bacteria at the (sub)genus level, a random forest classifier was developed to predict vaginal infections with an area under the curve of 0.83. We further identified four modules of co-occurring bacterial taxa: L. crispatus, Gardnerella, Prevotella, and Bacteroides. The functional prediction revealed that nucleotide biosynthesis pathways were upregulated in patients with human papilloma virus, and carbohydrate degradation pathways were downregulated in patients with BV. Overall, our study identified the bacterial signatures of healthy and infected vaginal microbiota, providing unique insights into the clinical diagnosis and health status prediction of women of a reproductive age.