Résumé
Objective To evaluate the applicability of small molecular markers of nephrotoxicity that in prediction of drug toxicity.Method Extracts of five kinds of traditional Chinese medicines (Tripterygium wilfordii,Strychni semen,Aristolochiafangchi,Rhei Radix et Rhizoma,and Xanthium sibiricum) that had known as nephrotoxicity were ig given to rats to establish renal injury models,and the blood samples were collected after administration for 1 and 7 d.Then blood samples were analyzed by UPLC/Q-TOF-MS for five kinds of small molecule biomarkers-thymidine,lyso-phosphatidylcholine (LPC 16:1),LPC (18:4),LPC (20:5),and LPC (22:5).The support vector machine (SVM) prediction model was established to determine the toxicity.The levels of Cr and BUN in serum were determined by automatic biochemical analyzer.The rats in each group were sacrificed after blood collection,and the kidneys were taken for HE staining.Result No toxicity was observed in the control group,and the biochemical test results showed no renal injury after mentioned five kinds Chinese herbs were given for 1 d,while SVM model of nephrotoxicity had been found abnormal.After administration for 7 d,the results of SVM model show renal toxicity,which were consistent with biochemical and pathological examination.Conclusion Metabonomics combined with the earlier established SVM model enabled prediction of drug nephrotoxicity more sensitively,quickly and \ccurately,and it is of great significance for the discovery of drug toxicity as well as the prevention and treatment of drug-induced renal injuries in clinic.