Objectives: Membranous nephropathy (MN) and minimal change disease (MCD) are two common types of nephrotic syndrome that have similar clinical presentations but require different treatment strategies. Currently, the definitive diagnosis for these conditions relies on invasive renal biopsy, which can be limited in clinical practice.Methods: In this study, we aimed to differentiate idiopathic MN (IMN) from MCD using clinical data and gut microbiota. We collected clinical data and stool samples from 115 healthy individuals, 115 IMN, and 45 MCD at the onset of disease and performed 16S rRNA sequencing. Through machine learning methods including random forest, logistic regression, and support vector machine, a classifier to differentiate IMN from MCD was constructed.Results: Baseline clinical data comparing the IMN and MCD groups showed that the MCD had higher levels of hemoglobin, uric acid, cystatin C, β2-microglobulin, α1-microglobulin, total cholesterol, and low-density lipoprotein and lower levels of albumin and CD4+ T-cell counts. The gut microbiota of the two groups differed at all levels of the phylum and genus. Differential gut microbiota may disturb the integrity of the intestinal wall and lead to the passage of inflammatory mediators through the intestinal barrier, causing kidney injury. We constructed a noninvasive classifier with a discrimination efficacy of 0.939 that combined the clinical data and gut microbiota information to identify IMN and MCD.Conclusions: The classifier of the gut microbiota combined with clinical indicators has achieved good performance in identifying IMN and MCD, which provides a new approach for the noninvasive discrimination of different pathological types of kidney disease.
Keywords: 16S rRNA; Idiopathic membranous nephropathy; gut microbiota; machine learning; minimal change disease.