Figure 2
Maturation of the microbiota. A 21-feature random forest regression model was constructed with the appropriate growth samples (R2 = 58%, p = 0.001). The number of features was selected by cross-validation (a). The predicted microbiota maturity age increased with postmenstrual age (b). The model was applied to a separate cohort of infants with appropriate growth (c) and to infants with growth failure (d). The spline derived from the appropriate growth infants in the primary cohort is shown in each panel (b–d). The 21 features in the model and their abundance (rarefied counts) over time (i.e., postmenstrual age) are shown (e). Relative microbiota maturity and microbiota-for-age Z scores were similar in the two appropriate growth cohorts, but significantly lower in growth failure (f,g). Microbiota maturity was also significantly lower when the analysis was restricted to infants without sepsis, necrotizing enterocolitis, or intestinal perforation (f). Relative microbiota maturity age was lower in infants with growth failure than infants with appropriate growth when infants were stratified by gestational age at birth (h). **p < 0.05 by pairwise Wilcoxon rank sum test with Benjamini-Hochberg adjustment. AG, appropriate growth. GF, growth failure. LOS, late-onset sepsis. NEC, necrotizing enterocolitis. SIP, spontaneous intestinal perforation. Val-AG, validation cohort appropriate growth.
