Validation of oligonucleotide microarray data using microfluidic low-density arrays: a new statistical method to normalize real-time RT-PCR data
Validation of oligonucleotide microarray data using microfluidic low-density arrays: a new statistical method to normalize real-time RT-PCR data
复制标题
DOI：
10.2144/05385mt01
复制
发表时间：
2005-05-01
期刊：
影响因子：
2.7
通讯作者：
Coombes, KR
中科院分区：
文献类型：
--
作者：
Abruzzo, LV;Lee, KY;Coombes, KR
利用微阵列进行基因表达谱研究以测量信使RNA（mRNA）的表达，常常会鉴定出一长串差异表达的基因。差异表达通常使用实时逆转录聚合酶链反应（RT - PCR）分析来验证。在传统的实时RT - PCR分析中，表达量是参照一个对照基因，即管家基因进行标准化的。然而，没有一个单一的管家基因可用于所有研究。我们使用了TaqMan®低密度阵列，这是一种利用微流体技术进行实时RT - PCR的中通量方法，可同时检测9例慢性淋巴细胞白血病（CLL）样本中96个基因的表达。我们开发了一种基于线性混合效应模型的新型统计方法来分析数据。这种方法能自动识别在样本中表达无显著差异的基因，从而可将它们用于对其余基因进行标准化。我们将标准化后的实时RT - PCR值与从Affymetrix Hu133A基因芯片®寡核苷酸微阵列获得的结果进行了比较。我们发现使用TaqMan低密度阵列的实时RT - PCR在7个数量级上都能产生可重复的测量结果。我们的模型识别出了许多表达水平几乎恒定的基因，包括管家基因PGK1、GAPD、GUSB、TFRC和18S rRNA。在参照无差异基因的几何平均数进行标准化后，对于在样本中适度表达且存在差异的基因，实时RT - PCR和微阵列之间的相关性很高。
Profiling studies using microarrays to measure messenger RNA (mRNA) expression frequently identify long lists of differentially expressed genes. Differential expression is often validated using real-time reverse transcription PCR (RT-PCR) assays. In conventional real-time RT-PCR assays, expression is normalized to a control, or housekeeping gene. However no single housekeeping gene can be used for all studies. We used TaqMan (R) Low-Density Arrays, a medium-throughput method for real-time RT-PCR using microfluidics to simultaneously assay the expression of 96 genes in nine samples of chronic lymphocytic leukemia (CLL). We developed a novel statistical method, based on linear mixed-effects models, to analyze the data. This method automatically identifies the genes whose expression does not vary significantly over the samples, allowing them to be used to normalize the remaining genes. We compared the normalized real-time RT-PCR values with results obtained from Affymetrix Hu133A GeneChip (R) oligonucleotide microarrays. We found that real-time RT-PCR using TaqMan Low-Density Arrays yielded reproducible measurements over seven orders of magnitude. Our model identified numerous genes that were expressed at nearly constant levels, including the housekeeping genes PGK1, GAPD, GUSB, TFRC, and 18S rRNA. After normalizing to the geometric mean of the unvarying genes, the correlation between real-time RT-PCR and microarrays was high for genes that were moderately expressed and varied across samples.