Experimental Design

A thorough experimental design will help to identify sources of variation in your microarray experiment, minimize experimental costs, yield high-quality data, and provide a sound basis for statistical analysis. The most common sources of variation are biological variation, technical variation (related to the experimenter, to the extraction, labeling, and hybridization of samples, and to the slide scanning), and measurement error.

The design of a two-color microarray experiment usually has the following layers: (1) the treatment and the time points to be measured (a treatment can be any attribute, including sex, strain, or tissue type); (2) the number of independent biological replicates to allow a determination of biological variation; (3) the number of technical replicates to allow a determination of experimental variability; and (4) the design of the microarray itself, which includes the source of the nucleic acids to print, the number of duplicate spots, and internal controls.

General design principles:

Always keep the goals of the experiment in mind. Experiments that are designed to address a particular question are more likely to be interpretable than large-scale random fishing expeditions.
Perform adequate biological replication. Keep in mind that biological replicates are truly independent only if the experimental materials on which the measurements were obtained were handled separately at all stages of the experiment that are likely to introduce variation.
In two-color experiments, it is important to perform dye swaps or introduce a looping design, to balance dye efficiencies and samples (even if indirect dye incorporation methods are used).
Perform quality control for every step in the process (“garbage in, garbage out”). Don't try to rescue a poor hybridization with statistical methods.