Fig. 4

Independent machine learning approach identifies similar differential taxa between infants with FPIAP and healthy controls. A–C Random forest prediction accuracy across all samples (A) and in those over 6 months comparing controls with resolved FPIAP cases (C). Most important features identified in differentiating disease states across all samples (B) and those over 6 months (D). Colored taxa in B and D represent those also independently identified in our multivariate analyses shown in Fig. 3