Simultaneous fecal microbial and metabolite profiling enables accurate classification of pediatric irritable bowel syndrome

Abstract

Background: We previously showed that stool samples of pre-adolescent and adolescent US children diagnosed with diarrhea-predominant IBS (IBS-D) had different compositions of microbiota and metabolites compared to healthy age-matched controls. Here we explored whether observed fecal microbiota and metabolite differences between these two adolescent populations can be used to discriminate between IBS and health.

Findings: We constructed individual microbiota- and metabolite-based sample classification models based on the partial least squares multivariate analysis and then applied a Bayesian approach to integrate individual models into a single classifier. The resulting combined classification achieved 84 % accuracy of correct sample group assignment and 86 % prediction for IBS-D in cross-validation tests. The performance of the cumulative classification model was further validated by the de novo analysis of stool samples from a small independent IBS-D cohort.

Conclusion: High-throughput microbial and metabolite profiling of subject stool samples can be used to facilitate IBS diagnosis.

Figures

Fig. 1

Schematic overview of the classification model generation. Pink and green points represent individual kIBS and kHLT samples, respectively, distributed in the simulated T-vs-Torthogonal PLS ordination space. Blue point represents an unknown sample that is classified by the PLS-DA models. M and C denote overall microbiota- and metabolite-based classification models, respectively; G is the group identifier; b0…n are numerical parameters and m1…n and c1…p are values of specific microbes and metabolites, respectively. See statistical data analyses section for the definitions of Bayesian model terms and parameters

Fig. 2

Improvement of sample classification based on the integration of microbiota- and metabolite-based PLS-DA models. a Sample classifications are shown as provided by the microbial genus abundance-based PLS-DA model (top row), metabolite-based PLS-DA model (middle row), and combined Bayesian model (bottom row). Each column represents a unique sample from IBS and healthy sets as shown. Each square is colored according to the group assignment confidence based on the gradient as shown in the legend. Average assignment accuracy and confidence for each model are indicated at the right of the figure. b Application of the Bayesian integration model to a set of four new IBS-D samples. c Density distribution plots of PDI values for IBS-D and healthy adolescent samples. Top three discriminating genera and metabolites were used to compute PDI values. The X axis shows the range of PDI values; the Y axis represents the density (frequency) of PDI values at each position along the X axis. PDI values for individual kIBS and kHLT samples are shown on the plots as discrete points. Blue points represent new IBS-D samples. d Receiver operating characteristic analysis of PLS-DA models (left panel) and patient discrimination indices (right panel). AUC area under the curve (represents the discrimination ability of each model; higher value equals better discrimination), G genus, M metabolite