Random forest modeling including protein concentrations generated from the last swab before 28 weeks of gestation was used to generate models using the combination of all cytokines, chemokines, immune-related growth factors, and antimicrobial molecules in this study (black line), maternal characteristics (blue line), or the combination of all cytokines, chemokines, immune-related growth factors, and antimicrobial molecules in this study together with maternal characteristics (red line) for (A) all sPTB, (B) all spontaneous preterm labor with intact membranes (sPTL), (C) sPTL who delivered before 34 weeks of gestation, (D) all preterm prelabor rupture of membranes (PPROM), and (E) PPROM who delivered before 34 weeks of gestation. Area under the curve (AUC) values and 95% confidence intervals are given for each curve. P-values correspond to the comparisons between the curves of maternal characteristics alone and the curves of maternal characteristics together with all cytokines, chemokines, immune-related growth factors, and antimicrobial molecules within each study group.
Figure 7—source data 1. Comparison of AUC vs. AUPR statistics for vaginal immunoproteome predictive models of sPTB and subtypes.AUC and AUPR calculations used preterm birth and related subgroups as the positive group while the reference (i.e., negative) group always constituted normal pregnancy group controls.