Fig. 2: Characterization of vaginal metabolome dynamics.

A Partial least squares discriminant analysis (PLS-DA) shows a partial but notable separation between baseline and follow-up samples for the vaginal metabolome. Significance was determined using permutational multivariate analysis of variance. B The resultant metabolite abundance matrix was then subjected to co-abundance clustering using WGCNA. Box plots indicate the normalized module eigenvalue of MetaG10 (p = 0.0037) and MetaG1 (p = 0.023). The central line indicates the median. The lower and upper hinges indicate the first and third quartiles. Significance was determined using the Wilcoxon signed rank test. C Leave-one-out analysis was conducted to determine the contribution of 389 identified metabolites in differentiating baseline and follow-up groups. The top ten essential metabolites in MetaG10, 1, 3, 5, and 7 are shown for display purposes. For 50 displayed metabolites, the normalized abundances (z-score normalization of the log2 transformed intensity) are shown in the heatmap. D The distribution of bray-curtis distance for VM and vaginal metabolome at baseline and 6 M follow-up. E For the 65 patients included, the intra-individual distance was calculated based on the VM and the vaginal metabolome, respectively. The central line indicates the median. The lower and upper hinges indicate the first and third quartiles.