Optimising go term tests

Dear @tanghaibao, 

I'm implementing a go-term enrichment analysis not directly based on genes but on cis-regulatory regions acting on nearby genes.
At the moment I can test all go terms, but I'm like to optimise the number of tests in order to reduce the effect of the FDR, maybe similar to how GOATOOLS does.
How could I do that?
Thanks a lot
F
