Baty Florent, Facompré Michaël, Wiegand Jan, Schwager Joseph, Brutsche Martin H
Pulmonary Gene Research, Universitätsspital Basel, Petersgraben 4, 4031 Basel, Switzerland.
BMC Bioinformatics. 2006 Sep 29;7:422. doi: 10.1186/1471-2105-7-422.
Evaluating the importance of the different sources of variations is essential in microarray data experiments. Complex experimental designs generally include various factors structuring the data which should be taken into account. The objective of these experiments is the exploration of some given factors while controlling other factors.
We present here a family of methods, the analyses with respect to instrumental variables, which can be easily applied to the particular case of microarray data. An illustrative example of analysis with instrumental variables is given in the case of microarray data investigating the effect of beverage intake on peripheral blood gene expression. This approach is compared to an ANOVA-based gene-by-gene statistical method.
Instrumental variables analyses provide a simple way to control several sources of variation in a multivariate analysis of microarray data. Due to their flexibility, these methods can be associated with a large range of ordination techniques combined with one or several qualitative and/or quantitative descriptive variables.
在微阵列数据实验中，评估不同变异来源的重要性至关重要。复杂的实验设计通常包含构建数据的各种因素，这些因素应予以考虑。这些实验的目的是在控制其他因素的同时探索某些给定因素。
我们在此提出了一类方法，即关于工具变量的分析方法，该方法可轻松应用于微阵列数据的特定情况。在研究饮料摄入对外周血基因表达影响的微阵列数据案例中，给出了一个工具变量分析的示例。将这种方法与基于方差分析的逐个基因统计方法进行了比较。
工具变量分析为在微阵列数据的多变量分析中控制多种变异来源提供了一种简单方法。由于其灵活性，这些方法可与大量排序技术以及一个或多个定性和/或定量描述变量相结合。
