[Help required] setting cutoff

[David]

Hello there
It's not really a feature request, more a "help required"

I'm using derfinder to reannotate the Rat genome using RNA-seq data. I'm running fullcoverage() and regionmatrix() to id all the genomic region corresponding to transcript. However, fixing a unique cutoff value is problematic:
For weakly expressed genes, I lose the regions
For highly expressed genes, pre-transcript reads are not filtered enough and are identified as new regions.

I would like to know if there is a way to identify only region with a burst or a drop in reads according to surrounding background ?

Best
David