Some remarks on design and analysis of microarray experiments

Microarray experiments have become the centerpiece of genetic and genomic studies. The great advantage of such experiments is that they allow simultaneous examination of expression levels of thousands of genes. Several articles have been written on design of microarray experiments considering different aspects such as efficiency and power. Many of these articles consider comparing a discrete set of designs and search for optimal designs within a specific experimental framework. Besides the design issues, there are many statistical problems that occur in the analysis of data from microarray experiments. Some of these problems include data cleansing, normalization, testing for differential expression and clustering of genes. Here, we provide a brief overview of statistical problems that occur in the design and analysis of microarray experiments. We specifically focus on methods that have been suggested for identification of genes that are differentially expressed under the two experimental conditions. A new method is proposed and is shown to be more flexible including some of the previous methods as special cases.