Fig. 5

CiRNAseq: the CVM changes upon hrHPV infection. A Unsupervised clustering analysis of randomly selected cervical smears from hrHPV-negative and hrHPV-positive women (CIN2+) profiled at the RNA level shows three distinct clusters from left to right: the first cluster (1) includes a higher proportion of hrHPV-negative women, who have a microbiome characterized of Lactobacillus species, and particularly L. crispatus (CST I). The second cluster (2) contains a higher proportion of hrHPV-positive women with CIN2+ lesions, who possess a diverse microbiome (CST IV) containing distinctive bacteria such as Atopobium vaginae, Dialister micraerophilus, Gardnerella vaginalis, Lactobacillus iners, Megasphaera genomosp type 1, Sneathia amnii, and Sneathia sanguinegens. The third cluster (3) includes both hrHPV-negative and hrHPV-positive women with predominantly hrHPV-negative women, who have a unique microbiome characterized by Lactobacillus species such as L. gasseri (CST II), L. iners (CST III), L. jensenii (CST V), and L. acidophilus. Clustering distance for columns: Canberra; clustering method: Ward (unsquared distances); Row scaling: Pareto scaling. The CVMP was shortened (CVMPs) to only include species with URC. URC from bacterial isolates in our CVMP were considered for analysis. B Principal component analysis (PCA) shows that hrHPV-negative and hrHPV-positive samples are correlated with PC1. The loading score of PC1 (data not shown) indicates that anaerobic bacteria have the stronger association with PC1 (Additional file 19). Original values are ln(x + 1)-transformed. No scaling is applied to rows; SVD with imputation is used to calculate principal components. C Histogram of the LDA scores computed for features differentially abundant between hrHPV-negative (negative) and hrHPV-positive women (positive)