Fecal Metabolome: New Addition to the Toolbox for Dietary Assessment?

Frankenfeld discusses the study by Shinn et al. on the use of fecal metabolome to identify metabolite biomarkers for predicting intake of specific foods. Metabolomics profiles from fecal samples were assessed pre- and postintervention. The authors used random forest modeling to evaluate the change in relative concentration of 96 metabolites that remained after pre-processing. Based on results from the single-food models, the authors observed that there was poor prediction performance for avocado, broccoli, barley, and oats, whereas there was good performance for almond and walnut single-food models. The authors additionally observed, from the results of a mixed-food random forest model, good prediction accuracy in distinguishing almond and walnut intake.