Extended Data Fig. 7. Metabolic models provide accurate predictions of putrescine, histamine, and tyramine.

a–c, Putrescine (a), histamine (b), and tyramine (c) predictions derived from microbiome metabolic models (NMPC; Methods; y-axis) plotted against measured metabolite levels (x-axis), showing good accuracy for all (Spearman ρ = 0.64; ρ = 0.54; and ρ = 0.62, respectively, P < 10⁻¹⁰ for all). d, Model coverage (y-axis; line, median; box, IQR; whiskers, 1.5*IQR), described as the fraction of total sample abundance represented by metabolic models, for each subgroup separately. Samples from White women had higher model coverage compared to samples from Black women, despite the lower accuracy for tyramine prediction in the former group. N = 173 for Black women; N = 21 for White women with sPTB; N = 30 for White women with TB. e, Spearman ρ between metabolic model predictions (NMPCs) and metabolite measurements (y-axis) for models that only contain a maximum of N most abundant species (x-axis). As our metabolic models account for the abundance of each microbe, and as the vaginal microbiome has a skewed distribution, our models are robust to lack of representation of low-abundance microbes.