Department of Animal Ecology, Institute of Ecological Science, VU University Amsterdam, Amsterdam, The Netherlands.
Toxicol Sci. 2010 May;115(1):34-40. doi: 10.1093/toxsci/kfq043. Epub 2010 Feb 4.
Environmental pollution is a worldwide problem, and metals are the largest group of contaminants in soil. Microarray toxicogenomic studies with ecologically relevant organisms, such as springtails, supplement traditional ecotoxicological research but are presently rather descriptive. Classifier analysis, a more analytical application of the microarray technique, is able to predict biological classes of unknown samples. We used the uncorrelated shrunken centroid method to classify gene expression profiles of the springtail Folsomia candida exposed to soil spiked with six different metals (barium, cadmium, cobalt, chromium, lead, and zinc). We identified a gene set (classifier) of 188 genes that can discriminate between six different metals present in soil, which allowed us to predict the correct classes for samples of an independent test set with an accuracy of 83% (error rate = 0.17). This study shows further that in order to apply classifier analysis to actual contaminated field soil samples, more insight and information is needed on the transcriptional responses of soil organisms to different soil types (properties) and mixtures of contaminants.
环境污染是一个全球性的问题，而金属是土壤中最大的污染物群体之一。利用生态相关生物（如跳虫）进行的微阵列毒理学基因组学研究补充了传统的生态毒理学研究，但目前还相当具有描述性。分类器分析是微阵列技术的一种更具分析性的应用，能够预测未知样本的生物学类别。我们使用不相关收缩质心方法对暴露于六种不同金属（钡、镉、钴、铬、铅和锌）污染土壤的跳虫 Folsomia candida 的基因表达谱进行分类。我们确定了一个由 188 个基因组成的基因集（分类器），可以区分土壤中存在的六种不同金属，这使得我们能够以 83%的准确率（错误率=0.17）预测独立测试集样本的正确类别。这项研究进一步表明，为了将分类器分析应用于实际污染的田间土壤样本，需要更多地了解土壤生物对不同土壤类型（特性）和污染物混合物的转录反应信息。
