...
Statistical enrichments are always relative to some background set. If you do not provide a background set, then GREAT uses a whole genome background. Such a background is not always appropriate. You should provide a background set to account for any inherent biases in samplingsampling <a name="reference1">1</a>.
...