Hi @jpjviro
You have two problems going on.
-
This prior topic described the first → Goseq NA NA NA NA values - #13 by jennaj
-
And I suspect that missing database is also influencing why all of the genes are equally “ranked” in the output.
So, you’ll either need to create that database or check to see if someone has created it then run in an R environment. Cow is a model organism, so I would definitely check, and maybe ask here https://support.bioconductor.org/ to see if anyone knows.
On the Galaxy side, we already have a ticket about this, but it isn’t an algorithm fix, instead that is just to remove the options not actually supported by GOSEQ itself from even being possible on our tool form.
Back you your graph → Try outputting all of the optional files and inspect the over/under p-values. Notice that each result line has exactly the same p-values, so all qualify for the “top ten”, and you end up with everything in that graph.
Hope this helps! 
Update: Hum, I can’t find that other ticket, so I just created a new one. The developers can consolidate as needed. But for reference, please see or follow here → Goseq with custom gene-go-category term inputs = bad scientific outputs · Issue #6500 · galaxyproject/tools-iuc · GitHub