Be aware of your experimental design

Before choosing a scoring method that ranks your genes, consider the underlying experimental design. Ask yourself the following questions: first, how many conditions do I examine? If you compare samples of a wild-type to a mutant group, you need a two-condition score. If you compare samples of more than two conditions, you are in a multi-condition setting. The next question is: Are the samples paired or unpaired? A 'wild-type versus mutant' comparison is unpaired: the samples are not related to each other. A 'before and after treatment' comparison on the same mouse is paired: the gene expression of the same animal was measured before and after treatment, that is each sample in one condition has its counterpart in the second condition. Like the two channels in a 'red-green' experiment, these coupled data must not be separated from each other. Figure 18.1 helps in deciding which design, and thus which scoring method, is appropriate.

0 0

Post a comment