The kidney is the most important organ for the excretion of pharmaceuticals and their metabolites. Among the complex structures of the kidney, the proximal tubule and renal interstitium are major targets of nephrotoxins. Despite its importance, there are only a few in silico models for predicting human nephrotoxicity for drug candidates. Here, we present quantitative structure鈥揳ctivity relationship (QSAR) models for three common patterns of drug-induced kidney injury, i.e., tubular necrosis, interstitial nephritis, and tubulo-interstitial nephritis. A support vector machine (SVM) was used to build the binary classification models of nephrotoxin versus non-nephrotoxin with eight fingerprint descriptors. To build the models, we constructed two types of data sets, i.e., parent compounds of pharmaceuticals (251 nephrotoxins and 387 non-nephrotoxins) and their major urinary metabolites (307 nephrotoxins and 233 non-nephrotoxins). Information on the nephrotoxicity of the pharmaceuticals was taken from clinical trial and postmarketing safety data. Though the mechanisms of nephrotoxicity are very complex, by using the metabolite information, the predictive accuracies of the best models for each type of kidney injury were better than 83% for external validation sets. Software to predict nephrotoxicity is freely available from our Web site at
http://bmdrc.org/DemoDownload.