The identification of novel biomarkers of renal toxicity using automatic data reduction techniques and PCA of proton NMR spectra of urine

Early detection of drug-induced toxic lesions is of considerable importance in the pharmaceutical industry. Many drugs and toxins produce characteristic patterns of biochemical perturbations in the urinary profile related to the site or mechanism of the lesion. H-1 nuclear magnetic resonance (NMR) spectroscopy of biofluids has been shown to be a useful technique for characterising such lesions. We present here an efficient approach to the analysis and classification of complex urine NMR spectra obtained from rats treated with various nephrotoxins (glomerular, papillary and proximal tubular) based on the automatic generation of descriptors for the spectra with subsequent PCA. Urinalysis was performed using 600 MHz H-1 NMR spectroscopy and the site of renal lesion was confirmed by renal histology. A plot of the first three PCs showed distinct clustering of urine samples reflecting the site of toxicity within the kidney. Interrogation of the eigenvectors showed which NMR spectral regions contributed most to the separation of classes. These regions were examined visually for perturbations in metabolite profile and sets of 'marker' metabolites that characterised tissue-specific lesions were defined. These studies have shown that automatic data reduction of the spectra followed by multivariate techniques such as principal components analysis (PCA) is a reliable method for screening for biomarkers of organ or tissue-specific chemically-induced lesions. (C) 1998 Elsevier Science B.V. All rights reserved.