Scott likes two approaches:
1. GoMiner: http://discover.nci.nih.gov/gominer/htgm.servlet
But this requires gene names in a text file (MUST have .txt ending) to work; useful hack to get gene names from, e.g., blast -m 8 results (presuming fish.t has only the blast result identifier, e.g. "gi|5729849|ref|NM_006496.1"):
...
Then send fish.t.genes.genenames off to gominer, allowing it to select the background (you could send the annot file of course...)
2. DAVID: http://david.abcc.ncifcrf.gov/tools.jsp
In contrast to GoMiner, DAVID is fine having NCBI identifiers (NM_..., NP_..., etc.) that can easily be parsed out from blast results.
Why would you chose one or the other?
...