Use of short-term transcriptional profiles to assess the long-term cancer-related safety of environmental and industrial chemicals.

The Hamner Institutes for Health Sciences, Research Triangle Park, North Carolina 27709, USA.
Toxicological Sciences (Impact Factor: 4.33). 09/2009; 112(2):311-21. DOI: 10.1093/toxsci/kfp233
Source: PubMed

ABSTRACT The process for evaluating chemical safety is inefficient, costly, and animal intensive. There is growing consensus that the current process of safety testing needs to be significantly altered to improve efficiency and reduce the number of untested chemicals. In this study, the use of short-term gene expression profiles was evaluated for predicting the increased incidence of mouse lung tumors. Animals were exposed to a total of 26 diverse chemicals with matched vehicle controls over a period of 3 years. Upon completion, significant batch-related effects were observed. Adjustment for batch effects significantly improved the ability to predict increased lung tumor incidence. For the best statistical model, the estimated predictive accuracy under honest fivefold cross-validation was 79.3% with a sensitivity and specificity of 71.4 and 86.3%, respectively. A learning curve analysis demonstrated that gains in model performance reached a plateau at 25 chemicals, indicating that the size of current data set was sufficient to provide a robust classifier. The classification results showed that a small subset of chemicals contributed disproportionately to the misclassification rate. For these chemicals, the misclassification was more closely associated with genotoxicity status than with efficacy in the original bioassay. Statistical models were also used to predict dose-response increases in tumor incidence for methylene chloride and naphthalene. The average posterior probabilities for the top models matched the results from the bioassay for methylene chloride. For naphthalene, the average posterior probabilities for the top models overpredicted the tumor response, but the variability in predictions was significantly higher. The study provides both a set of gene expression biomarkers for predicting chemically induced mouse lung tumors and a broad assessment of important experimental and analysis criteria for developing microarray-based predictors of safety-related end points.

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The use of genes for distinguishing classes of toxicity has become well established. In this paper we combine the reconstruction of a gene dysregulation network (GDN) with a classifier to assign unseen compounds to their appropriate class. Gene pairs in the GDN are dysregulated in the sense that they are linked by a common expression pattern in one class and differ in this pattern in another class. The classifier gives a quantitative measure on this difference by its prediction accuracy. As an in-depth example, gene pairs were selected that were dysregulated between skin cells treated with either sensitizers or irritants. Pairs with known and novel markers were found such as HMOX1 and ZFAND2A, ATF3 and PPP1R15A, OXSR1 and HSPA1B, ZFP36 and MAFF. The resulting GDN proved biologically valid as it was well-connected and enriched in known interactions, processes and common regulatory motifs for pairs. Classification accuracy was improved when compared with conventional classifiers. As the dysregulated patterns for heat shock responding genes proved to be distinct from those of other stress genes, we were able to formulate the hypothesis that heat shock genes play a specific role in sensitization, apart from other stress genes. In conclusion, our combined approach creates added value for classification-based toxicogenomics by obtaining novel, well-distinguishing and biologically interesting measures, suitable for the formulation of hypotheses on functional relationships between genes and their relevance for toxicity class differences. Copyright © 2012 John Wiley & Sons, Ltd.
    Journal of Applied Toxicology 08/2012; · 2.60 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Around 40% of drug-induced liver injury (DILI) cases are not detected in preclinical studies using the conventional indicators. It has been hypothesized that genomic biomarkers will be more sensitive than conventional markers in detecting human hepatotoxicity signals in preclinical studies. For example, it has been hypothesized and demonstrated in some cases that (1) genomic biomarkers from the rat liver can discriminate drug candidates that have a greater or lesser potential to cause DILI in susceptible patients despite no conventional indicators of liver toxicity being observed in preclinical studies, and (2) more sensitive biomarkers for early detection of DILI can be derived from a "subtoxic dose" at which the injury in the liver occurs at the molecular but not the phenotypic level. With a public TGx data set derived from short-term in vivo studies using rats, we divided drugs exhibiting human hepatotoxicity into three groups according to whether elevated alanine aminotransferase (ALT) or total bilirubin (TBL) were observed in the treated rats: (A) The elevation was observed in the treated rats, (B) no elevation was observed for all of the treated rats, and (C) no elevation could be observed at a lower dose and shorter duration but occur when a higher or longer treatment was applied. A control group (D) was comprised of drugs known not to cause human hepatotoxicity and for which no rats exhibited elevated ALT or TBL. We developed classifiers for groups A, B, and C against group D and found that the gene signature from scenario A could achieve 83% accuracy for human hepatotoxicity potential of drugs in a leave-one-compound-out cross-validation process, much higher than scenarios B (average 45%) and C (61%). Furthermore, the signature derived from scenario A exhibited relevance to hepatotoxicity in a pathway-based analysis and performed well on two independent public TGx data sets using different chemical treatments and profiled with different microarray platforms. Our study implied that the human hepatotoxicity potential of a drug can be reasonably assessed using TGx analysis of short-term in vivo studies only if it produces significant elevation of ALT or TBL in the treated rats. The study further revealed that the value of "sensitive" biomarkers derived from scenario C was not promising as expected for DILI assessment using the reported TGx design. The study will facilitate further research to understand the role of genomic biomarkers from rats for assessing human hepatotoxicity.
    Chemical Research in Toxicology 11/2011; 25(1):122-9. · 3.67 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Several groups have employed genomic data from subchronic chemical toxicity studies in rodents (90 days) to derive gene-centric predictors of chronic toxicity and carcinogenicity. Genes are annotated to belong to biological processes or molecular pathways that are mechanistically well understood and are described in public databases. To develop a molecular pathway-based prediction model of long term hepatocarcinogenicity using 90-day gene expression data and to evaluate the performance of this model with respect to both intra-species, dose-dependent and cross-species predictions. Genome-wide hepatic mRNA expression was retrospectively measured in B6C3F1 mice following subchronic exposure to twenty-six (26) chemicals (10 were positive, 2 equivocal and 14 negative for liver tumors) previously studied by the US National Toxicology Program. Using these data, a pathway-based predictor model for long-term liver cancer risk was derived using random forests. The prediction model was independently validated on test sets associated with liver cancer risk obtained from mice, rats and humans. Using 5-fold cross validation, the developed prediction model had reasonable predictive performance with the area under receiver-operator curve (AUC) equal to 0.66. The developed prediction model was then used to extrapolate the results to data associated with rat and human liver cancer. The extrapolated model worked well for both extrapolated species (AUC value of 0.74 for rats and 0.91 for humans). The prediction models implied a balanced interplay between all pathway responses leading to carcinogenicity predictions. Pathway-based prediction models estimated from sub-chronic data hold promise for predicting long-term carcinogenicity and also for its ability to extrapolate results across multiple species.
    PLoS ONE 01/2013; 8(5):e63308. · 3.53 Impact Factor