Statistical issues in microarray data analysis.

Microarrays provide the ability to quantitatively measure the abundance of specific RNA transcripts through sample hybridization to a solid-state grid of oligonucleotides or amplicons. The prospect of measuring the entire transcriptome is extremely alluring, but as with any experiment, it should be met with caution and great consideration. The level of confidence we can assign to the results depends on the skill at which the experiment is conducted, the quality of the experimental design and subsequent analysis, and, most important, the power in the study. Any microarray experiment consists of several components: (1) carrying out an appropriately designed (replicated) plant experiment; (2) array processing, which includes several steps of data acquisition and normalization; and (3) analysis of expression data to identify differentially expressed genes and overall patterns of expression. Numerous software packages are available to assist in performing these steps and it is not our intent to provide a software users manual or a statistical review. It is our intent to provide a brief user's explanation of these various components and present the commonly used methods.