Statistical methods in experimentation recommendation models for discovering gene regulation pathways

In recent years, scientists and health professionals have become overwhelmed by the amount of genomic data available for epidemiological studies and patient care. In April 1996, the genome sequence of the yeast Saccharomyces cerevisiae was completed. By December 1999, the first plant's complete genome sequence from Arabidopsis thaliana was published. The Human Genome Project and Celera completed a working draft reference of the human genome DNA sequence in June 2000. Microarray and high throughput RNA sequence technology has opened a new era for gene expression studies. It enables us to measure the level of mRNA from different genes among different cells. Large microarray datasets provide opportunities to learn causal relationships among the genes, and compile in-depth knowledge of physiology of different diseases. We have to deal with many issues to achieve this, including: Validity of the microarray and high throughput RNA sequence measurements; Experimental design of gene expression studies; Analysis of gene expression experimental results. For meaningful interpretation of experimental results, we need to start with a sound experimental design. However, not most of the statistical methods concentrate on issues related to gene expression data analyses. They do not take into consideration experimental design of microarray studies. The main topic of this article is review of modeling methods of the expected value of experimentation for discovering gene regulation pathways. In this article, we use experimentation as both interventions (e.g., an experiment with gene knock-outs) and observations (e.g., passively observing the expression level of genes with no interventions). WIREs Comput Stat 2013, 5:121–134. doi: 10.1002/wics.1244