Prediction and Classification of Drug Toxicity Using Probabilistic Modeling of Temporal Metabolic Data: The Consortium on Metabonomic Toxicology Screening Approach

摘要: Detection and classification of in vivo drug toxicity is an expensive time-consuming process. Metabolic profiling becoming a key enabling tool this area as it provides unique perspective on the characterization mechanisms response to toxic insult. As part Consortium Metabonomic Toxicology (COMET) project, substantial metabolic pathological database was constructed. We chose set 80 treatments build modeling system for prediction using NMR spectroscopy urine samples (n=12935) from laboratory rats (n=1652). The compound structures activities were diverse but there emphasis selection hepato nephrotoxins. developed two-stage strategy based assumptions that (a) adverse effects would produce profiles deviating those normal animals (b) such deviations be similar having physiological effects. To address first stage, we multivariate model urine, principal components analysis specially preprocessed 1H spectra. demonstrated high correspondence between occurrence abnormal profiles. In second extended density estimation method, "CLOUDS", compute multidimensional similarities treatments. Crucially, technique allowed distribution-free estimate similarity across multiple time points each treatment resulting matrix showed segregation liver toxins other Using matrix, able correctly identify target organ two "blind" treatments, even at sub-toxic levels. further validate approach, then applied leave-one-out approach predict main (liver or kidney) showing significant responses three most matches matrix. Where predictions could made, error rate 8%. sensitivities kidney 67 41%, respectively, whereas corresponding specificities 77 100%. some cases, not possible make because interference by drug-related metabolite signals (18%), inconsistent histopathological urinary (11%), genuine class overlap (8%), lack any (2%). This study constitutes largest validation date metabonomic preclinical toxicology assessment, confirming methodology offers practical utility rapid screening.