It is absolutely essential to plan and execute the experiments with utmost care. It is important to begin the planning of the microarray experiment with a proper question. The experimental model/system should be well-characterized or well-defined with an independent experimental verification. For example, if a growth factor was added which induces differentiation in 24-48 hours, but you are collecting RNA at three hours post-treatment, you should still check a parallel culture for verification that differentiation occurred at the 24-48 hours period. If possible, a quick check for a gene that is known to be affected by the treatment should be performed.
It is recommended that all experimental treatments be carried out in triplicates to compensate for biological and experimental variation. In vitro experiments using cultured cells should be conducted three different times (not three replicates performed on the same day) strictly following the same experimental procedures. Tumor specimens should be devoid of adjacent tissues and if possible, microdissected to obtain as pure a tumor sample as possible. Cell populations may also be further purified using cell-sorting techniques such as FACS. Dead cells also should be removed by density centrifugation. For comparative gene expression analysis, it is essential that all the experimental conditions such as temperature, CO2, media, reagents, and sample processing be kept identical for all samples.