You are here:
Genomic Indicators in the blood predict drug-induced liver injury
Citation:
Huang, J., W. Shi, J. Zhang, J. W. Chou, R. S. Paules, K. Gerrish, J. LI, J. Lou, R. D. Wolfinger, W. Bao, C. Tzu-Ming, Y. Nikolsky, D. Dosymbekov, M. O. Tsyganova, L. Shi, X. Fan, C. CORTON, M. Chen, Y. Cheng, W. Tong, H. Fang, AND P. R. Bushel. Genomic Indicators in the blood predict drug-induced liver injury. PHARMACOGENOMICS. Ashley Publications LIMITED, London, Uk, 104(4):267-277, (2010).
Impact/Purpose:
This work was initiated to determine if liver toxicity could be predicted using blood gene expression. The findings indicate that an analysis of gene expression changes in the blood may be an appropriate surrogate for predicting toxicity in other tissues especially the liver. The fact that a small number of genes were consistently altered upon liver necrosis indicates that they may be suitable biomakers of effect that could be used in more high-throughput assays currently in use or being developed in the Tox Cast program for screening thousands of chemicals for potential toxicity.
Description:
Hepatotoxicity and other forms of liver injury stemming from exposure to toxicants and idiosyncratic drug reactions are major concerns during the drug discovery process. Animal model systems have been utilized in an attempt to extrapolate the risk of harmful agents to humans and to predict the manifestations of acute toxicity resulting from off target effects of pharmaceuticals. We used microarray-based gene expression data acquired from rats exposed to a compendium of hepatotoxicants in order to build several classifiers to determine if transcript data could be used to predict liver injury. Gene- and pathway- based predictors comprised of using expression data from whole blood were found to be highly predictive of liver necrosis (accuracy ranging from about 70% to 90%). The cross tissue predictability of several classifiers was confirmed using Agilent and Affymetrix array technologies. The best predictor was derived from a nearest centroid classifier containing genes obtained using a forward feature selection with a five-fold internal cross-validation strategy. Coherent co-expression Bicluster analysis of the compendium data yielded clusters of genes that over-represent biological mechanisms related to initiation of an immune (inflammatory) response, induction of apoptosis and targeting to the mitochondria. Pathway-based classifiers that were identified in blood and highly predictive of the necrotic response in the liver were associated with inflammation, apoptosis, mitochondrial damage, angiogenesis and Toll-like receptor (TLR) signaling. Finally, the best gene- and pathway-based classifier models from the blood predicted liver necrosis of independent test samples with high accuracy (> 92%) for samples exposed to acetaminophen or carbon tetrachloride (both cause centrilobular necrosis) and with moderate accuracy (> 75%) for samples exposed to allyl alcohol (causes periportal necrosis). These results support the hypothesis that genomic (gene- and pathway-based) indicators in the blood can serve as surrogate biomarkers of hepatotoxicity potentially for pharmaceutical toxicity testing, pre-clinical and clinical trial screening, and clinical diagnostics.
