Gene expression profiling – Clusters of possibilities
详细信息 查看全文
- 作者：Anders Bergkvist ; Vendula Rusnakova ; Radek Sindelka ; Jose Manuel Andrade Garda ; Bjö ; rn Sjö ; green ; Daniel Lindh ; Amin Forootan ; Mikael Kubista
- 关键词：qPCR ; Statistical analysis ; Xenopus laevis ; Hierarchical clustering ; Heatmap ; PCA ; SOM ; Hypotheses
- 刊名：Methods
- 出版年：2010
- 出版时间：April 2010
- 年：2010
- 卷：50
- 期：4
- 页码：323-335
- 全文大小：1478 K
文摘
Advances in qPCR technology allow studies of increasingly large systems comprising many genes and samples. The increasing data sizes allow expression profiling both in the gene and the samples dimension while also putting higher demands on sound statistical analysis and expertise to handle and interpret its results. We distinguish between exploratory and confirmatory statistical studies. In this paper we demonstrate several techniques available for exploratory studies on a system of Xenopus laevis development from egg to tadpole. Techniques include hierarchical clustering, heatmap, principal component analysis and self-organizing maps. We stress that even though exploratory studies are excellent for generating hypotheses, results have not been proven statistically significant until an independent confirmatory study has been performed. An exploratory study may certainly be valuable in its own right, and there are often not enough resources to report both an exploratory and a confirmatory study at the same time. However, exploratory and confirmatory studies are intimately connected and we would like to raise that awareness among qPCR practitioners. We suggest that scientific reports should always have a hypothesis focus. Reports are either hypothesis generating, from an exploratory study, or hypothesis validating, from a confirmatory study, or both. In either case, we suggest the generated or validated hypotheses be specifically stated.