Fig. 6

Differential microbial features between STAT and control. a LEfSe cladograms showing discriminant taxa between control and STAT at weeks 4, 11, 16, and 30, respectively, with corresponding diet. All identified taxa were significantly altered by Kruskal–Wallis test (p <0.05) and had at least twofold increase by LDA. b Inter-week comparisons in control (upper) or STAT (lower). The week 4 to 11 comparison shows changes across weaning, the week 11 to 16 comparison shows changes from the introduction of HFD, and the week 16 to 30 comparison shows changes with increasing age. c A Random Forest classification model was built to predict disease outcome (class) based on bacterial OTU relative abundance (features) for each week of life. Heat map indicates the importance of each OTU (as mean increase error %) to the disease prediction models at each stage of life. The mean increase error for each OTU indicates the incremental decrease in prediction accuracy if that OTU is removed from the model. Highlighted time points show HFD. The table lists the predictive accuracy of the model by week. d Average microbiota-by-age z-score (MAZ) over time; z-score = 0 indicates appropriate maturation over time; higher or lower z-scores indicate accelerated or delayed microbiota development, respectively. ***p <0.001 relative to Control, one-way ANOVA with Fisher’s LSD adjusted for false-discovery rate