The intraclass correlation coefficient applied for evaluation of data correction, labeling methods, and rectal biopsy sampling in DNA microarray experiments
The intraclass correlation coefficient applied for evaluation of data correction, labeling methods, and rectal biopsy sampling in DNA microarray experiments
复制标题
DOI：
10.1152/physiolgenomics.00111.2003
复制
发表时间：
2003-12-16
影响因子：
4.6
通讯作者：
Keijer, J
中科院分区：
文献类型：
--
作者：
Pellis, L;Franssen-van Hal, NLW;Keijer, J
我们表明，类内相关系数（ICC）可以作为一种相对简单的统计措施来评估DNA微阵列分析的方法学和生物学差异。ICC是一种确定变量可重复性的度量，可以很容易地从方差分析表中计算出来。它基于对系统偏差和随机变异的评估，便于同时对多个样本进行比较。我们首先使用ICC来优化我们的微阵列数据归一化方法，发现使用中位数而不是平均值可以提高数据的正确率。然后对不同标记方法的可重复性进行了评价，间接荧光染料掺入标记的可重复性优于直接标记。最后，我们通过分析基因表达的总体变异来确定最佳活检取样。当取两次而不是一次活检时，人体内直肠活检的基因表达变异减少，但当取两次以上活检时，基因表达变异没有明显改善，这表明在我们的实验条件下，每个人取两次活检就足够进行DNA微阵列分析了。为了优化微阵列数据的准确性，每组应使用至少6个不同人的活检。
We show that the intraclass correlation coefficient ( ICC) can be used as a relatively simple statistical measure to assess methodological and biological variation in DNA microarray analysis. The ICC is a measure that determines the reproducibility of a variable, which can easily be calculated from an ANOVA table. It is based on the assessment of both systematic deviation and random variation, and it facilitates comparison of multiple samples at once. We used the ICC first to optimize our microarray data normalization method and found that the use of median values instead of mean values improves data correction. Then the reproducibility of different labeling methods was evaluated, and labeling by indirect fluorescent dye incorporation appeared to be more reproducible than direct labeling. Finally, we determined optimal biopsy sampling by analyzing overall variation in gene expression. The variation in gene expression of rectal biopsies within persons decreased when two biopsies were taken instead of one, but it did not considerably improve when more than two biopsies were taken from one person, indicating that it is sufficient to use two biopsies per person for DNA microarray analysis under our experimental conditions. To optimize the accuracy of the microarray data, biopsies from at least six different persons should be used per group.