Although emerging evidence revealed that the gut microbiome served as a tool and as biomarkers for predicting and detecting specific cancer or illness, it is yet unknown if vaginal microbiome-derived bacterial markers can be used as a predictive model to predict the severity of CIN. In this study, we sequenced V3 region of 16S rRNA gene on vaginal swab samples from 66 participants (24 CIN 1-, 42 CIN 2+ patients) and investigated the taxonomic composition. The vaginal microbial diversity was not significantly different between the CIN 1- and CIN 2+ groups. However, we observed Lactobacillus amylovorus dominant type (16.7%), which does not belong to conventional community state type (CST). Moreover, a minimal set of 33 bacterial species was identified to maximally differentiate CIN 2+ from CIN 1- in a random forest model, which can distinguish CIN 2+ from CIN 1- (area under the curve (AUC) = 0.952). Among the 33 bacterial species, Lactobacillus iners was selected as the most impactful predictor in our model. This finding suggests that the random forest model is able to predict the severity of CIN and vaginal microbiome may play a role as biomarker.
Keywords: cervical intraepithelial neoplasia; machine learning; vaginal microbiome.