Li Jiang, Spletter Maria L, Johnson Jeffrey A
School of Pharmacy, University of Wisconsin-Madison, Madison, Wisconsin 53705-2222, USA.
Physiol Genomics. 2005 Mar 21;21(1):43-58. doi: 10.1152/physiolgenomics.00214.2004. Epub 2004 Dec 21.
This paper compares the gene expression profiles identified by short (Affymetrix U95AV2) or long (Agilent Hu1A) oligonucleotide arrays on a model for upregulation of a cluster of antioxidant responsive element-driven genes by treatment with tert-butylhydroquinone. MAS 5.0, dCHIP, and RMA were applied to normalize the Affymetrix data, and Lowess regression was considered for Agilent data. SAM was used to identify the differential gene expression. A set of biological markers and housekeeping genes were chosen to evaluate the performance of multiple normalization approaches. Both arrays illustrated a definite set of overlapping genes between the data sets regardless of data mining tools used. However, unique gene expression profiles based on the platform used were also revealed and confirmed by quantitative RT-PCR. Further analysis of the data revealed by alternative approaches suggested that alternative splicing, multiple vs. single probe(s) measurement, and use or nonuse of mismatch probes may account for the discrepant data. Therefore, these two microarray technologies offer relatively reliable data. Integration of the gene expression profiles from different array platforms may not only help for cross-validation but also provide a more complete view of the transcriptional scenario.
本文比较了短寡核苷酸芯片（Affymetrix U95AV2）和长寡核苷酸芯片（安捷伦Hu1A）在叔丁基对苯二酚处理诱导抗氧化反应元件驱动的基因簇上调模型中所鉴定的基因表达谱。应用MAS 5.0、dCHIP和RMA对Affymetrix数据进行标准化处理，对安捷伦数据则采用局部加权回归法。使用SAM来鉴定差异基因表达。选择一组生物标志物和管家基因来评估多种标准化方法的性能。无论使用何种数据挖掘工具，两种芯片在数据集之间都显示出一组明确的重叠基因。然而，基于所用平台的独特基因表达谱也通过定量逆转录-聚合酶链反应得到了揭示和证实。对替代方法所揭示的数据进行的进一步分析表明，可变剪接、多探针与单探针测量以及错配探针的使用与否可能是造成数据差异的原因。因此，这两种微阵列技术可提供相对可靠的数据。整合来自不同阵列平台的基因表达谱不仅有助于交叉验证，还能提供更完整的转录情况视图。
