cuffmerge on cufflinks denovo or with -g

[secda1]

Hey all,

I have RNAseq data from various rice tissues, and would like to use this data to improve the current (poorly) annotated rice genome, in order to discover new genes.

So my question is:

Is it better to run cufflinks with the denovo mode, and then use cuffmerge with -g (providing the gff file)

or is it better to run cufflinks with -g directly, and then also use cuffmerge with -g ?

In term of time the second option is by far much slower, but I dont know if it makes a difference at the end.

thanks for your help

Cheers

David