Fig. 7

Diagnostic potential of gut microbiotas and fecal metabolites during multistep CKD progression. Gut metagenomic and metabolomic markers for noninvasively discriminating patients with A mild CKD, B moderate CKD, and C ESRD from the healthy controls (HC). These makers identified by random-forest (RF) classifiers based on species (red), or metabolites (green) alone, and the combination (blue) of the two features. Performance of RF classifiers was examined by receiver operating characteristic (ROC) curves and evaluated by 100-fold cross-validation. The box plots represent the average accuracy of models. D ROC analysis for the common model, taking into account fecal hydroquinone, L-cystine, 12-keto-tetrahydro-leukotriene B4, and Ruminococcus bromii, could correctly detect each of CKD groups from the HC. E Predicted score comparisons of the common model between the each of CKD groups and HC group using Mann–Whitney U-test (***P < 0.001). F ROC analysis for distinguishing each of CKD groups from HC group based on serum creatinine alone. mod-CKD, moderate CKD; ESRD, end-stage renal disease; LTB4, leukotriene B4; TXB2, thromboxane B2; 12-KETE, 12-keto-5,8,10,14-eicosatetraenoic acid