Fig. 2

Vaginal microbiota of pregnant women differs by SARS-CoV-2 infection. A Alpha diversity is shown using the Chao1 estimator comparing pregnant women with SARS-CoV-2 infection during pregnancy and pregnant healthy controls (HC). B Alpha diversity is shown using the Chao1 estimator comparing pregnant women with SARS-CoV-2 infection early, late, or active vs. HC. Significance was determined using Linear regression and pairwise comparison with estimated marginal means. C, D Beta diversity analyses by groups: C all pregnant women with SARS-CoV-2 infection during pregnancy compared to HC; D pregnant women with SARS-CoV-2 infection early, late, or active compared to HC. Beta diversity comparisons were performed using PERMANOVA analysis with pairwise comparisons and BETADISPER for dispersion analysis. For PERMANOVA and BETADISPER analyses we used Sørensen dissimilarities. P values were all adjusted by false discovery rate. E, F, G Bacterial taxa (at the sequence variant or SV) were selected by the random forest classification (RFC) and ranked according to their importance in the classification. RFC comparisons are shown in E pregnant women with SARS-CoV-2 infection during pregnancy vs. HC, F pregnant women with early SARS-CoV-2 infection vs. HC, G pregnant women with active SARS-CoV-2 infection vs. HC. Bars’ colors indicate the comparison group (i.e., SARS-CoV-2 or HC), and each bar indicates the importance by which the increase on an SV predicts a particular comparison group. The selection of the variables for RFC was performed with Boruta algorithm. We also used the Local Interpretable Model-agnostic Explanation (LIME) to estimate a threshold of the abundance of the SV selected with Boruta that predicts a particular comparison group. *Padj < 0.050, **Padj < 0.010