Abstract: The pathogenesis of the nephrotic syndrome is complex and the pathological types are diverse, so the minor symptoms in its early phases are difficult to detect. Renal biopsy is the gold indicator for the diagnosis of renal pathology and progression, but poor patient compliance shows, and the optimal treatment time is often delayed. Therefore, the discovery of biomarkers for early diagnosis and disease progression monitoring is of great clinical significance. In this study, doxorubicin-injured podocyte models were used to simulate human kidney disease at different stages of progression. LC-MS-based metabolomic technology combined with statistical methods was used to screen and identify the potential biomarkers associated with early injury or progression of podocytes. The results of cell viability, apoptosis tests and podocyte structural protein analysis showed that the model was successfully constructed, and the degree of podocyte injury was significantly different between the two modeling methods. According to VIP > 1 and P < 0.05 based on the orthogonal partial least squares discriminant analysis (OPLS-DA) model, nine differential metabolites reflecting early podocyte injury and twelve differential metabolites reflecting the injury progression were screened, respectively. ROC analysis was adopted to focus on the potential biomarkers that can reflecting the early podocyte injury including L-tryptophan, guanosine triphosphate (GTP), 5′-thymidylic acid (dTMP) and thymidine, and the biomarkers reflecting the injury progression of podocytes composed of L-phenylalanine, L-tyrosine acid, uridine 5′-diphosphate (UDP) and guanosine 5′-diphosphate (GDP) AUC > 0.85. It indicated that these eight metabolites may have high sensitivity and diagnostic ability. This study provides a reference for the research on biomarkers of progressive diseases.