National Toxicology Program, National Institute of Environmental Health Sciences, NIH, RTP, NC 27709, USA.
Toxicol Appl Pharmacol. 2010 Mar 15;243(3):300-14. doi: 10.1016/j.taap.2009.11.021. Epub 2009 Dec 11.
Identification of carcinogenic activity is the primary goal of the 2-year bioassay. The expense of these studies limits the number of chemicals that can be studied and therefore chemicals need to be prioritized based on a variety of parameters. We have developed an ensemble of support vector machine classification models based on male F344 rat liver gene expression following 2, 14 or 90 days of exposure to a collection of hepatocarcinogens (aflatoxin B1, 1-amino-2,4-dibromoanthraquinone, N-nitrosodimethylamine, methyleugenol) and non-hepatocarcinogens (acetaminophen, ascorbic acid, tryptophan). Seven models were generated based on individual exposure durations (2, 14 or 90 days) or a combination of exposures (2+14, 2+90, 14+90 and 2+14+90 days). All sets of data, with the exception of one yielded models with 0% cross-validation error. Independent validation of the models was performed using expression data from the liver of rats exposed at 2 dose levels to a collection of alkenylbenzene flavoring agents. Depending on the model used and the exposure duration of the test data, independent validation error rates ranged from 47% to 10%. The variable with the most notable effect on independent validation accuracy was exposure duration of the alkenylbenzene test data. All models generally exhibited improved performance as the exposure duration of the alkenylbenzene data increased. The models differentiated between hepatocarcinogenic (estragole and safrole) and non-hepatocarcinogenic (anethole, eugenol and isoeugenol) alkenylbenzenes previously studied in a carcinogenicity bioassay. In the case of safrole the models correctly differentiated between carcinogenic and non-carcinogenic dose levels. The models predict that two alkenylbenzenes not previously assessed in a carcinogenicity bioassay, myristicin and isosafrole, would be weakly hepatocarcinogenic if studied at a dose level of 2 mmol/kg bw/day for 2 years in male F344 rats; therefore suggesting that these chemicals should be a higher priority relative to other untested alkenylbenzenes for evaluation in the carcinogenicity bioassay. The results of the study indicate that gene expression-based predictive models are an effective tool for identifying hepatocarcinogens. Furthermore, we find that exposure duration is a critical variable in the success or failure of such an approach, particularly when evaluating chemicals with unknown carcinogenic potency.
鉴定致癌活性是两年生物测定的主要目标。这些研究的费用限制了可以研究的化学物质的数量，因此需要根据各种参数对化学物质进行优先级排序。我们已经基于雄性 F344 大鼠肝脏在接触一组肝致癌物（黄曲霉毒素 B1、1-氨基-2,4-二溴蒽醌、N-亚硝二甲胺、甲基丁香酚）和非肝致癌物（对乙酰氨基酚、抗坏血酸、色氨酸）后的 2、14 或 90 天的基因表达，开发了一组支持向量机分类模型。基于个体暴露时间（2、14 或 90 天）或暴露组合（2+14、2+90、14+90 和 2+14+90 天）生成了 7 个模型。除了一个模型产生的交叉验证错误率为 0%外，所有数据集都生成了模型。使用接触一组烯基苯调味剂的大鼠肝脏的表达数据对模型进行了独立验证。根据使用的模型和测试数据的暴露时间，独立验证错误率范围为 47%至 10%。对独立验证准确性影响最大的变量是烯基苯测试数据的暴露时间。随着烯基苯数据暴露时间的增加，所有模型的性能都普遍提高。模型区分了先前在致癌性生物测定中研究过的肝致癌物（茴香脑和黄樟素）和非肝致癌物（大茴香脑、丁香酚和异丁香酚）。在黄樟素的情况下，模型正确地区分了致癌和非致癌剂量水平。模型预测，如果在雄性 F344 大鼠中以 2mmol/kg bw/天的剂量水平研究两年，两种以前未在致癌性生物测定中评估的烯基苯，马郁兰和异黄樟素，将具有弱肝致癌性；因此，相对于其他未经测试的烯基苯，这些化学物质应在致癌性生物测定中具有更高的优先级。该研究的结果表明，基于基因表达的预测模型是鉴定肝致癌物的有效工具。此外，我们发现暴露时间是此类方法成功或失败的关键变量，特别是在评估具有未知致癌潜力的化学物质时。
