Count reads around breakpoint (rna-seq)

[spoonman]

Hej there,

I'm trying to detect gene fusion with rna-seq. The draft below is showing
an example of how reads spanning read and split reads can be used to detect
fusions. Counts of spanning reads and split reads are given by the tool that I
use, that works fine. Additionally, I would like to count the reads/fragments that
contradict the fusion.
That would be:
sum(R2/1 (only if R2/2 maps to GENEA but on the other side of the breakpoint) + R3/1(this reads contradict the fusion, no need that R3/2 maps to GENEA)

same could be calculated for GENEB, note that I'm dealing with rna-seq data and the breakpoint I'm showing is on rna level.

GENEA R1/1 -----``````````````````GENEB ------R1/2
```````R2/1 ----- ``R2/2-----
++++++++++++++|----------------````` -----------------|+++++++++++++++
``````````R3/1 ------ ```R3/2-----
``````````R4/1---`````````````````````````` ---R4/1 -------R4/2

legend: ++++ Exon, | breakpoint, - intron, ----R1/1 read1 pair1, R1/2 read 1 pair 2; R4/1 is a split read, half the read maps on GENEA the other half on GENEB, R1/1 and R1/2 are spanning reads whereas 1 pair maps to GENEA the other to GENEB,```this are just placeholder to show correct "alignment" of the reads

I was thinking about using bedtools or samtools view but I can't rely only on intervals, I would really need to check wether eg R2/1 maps "left" to the breakpoint of GENEA and R2/2 has to map "right" to the breakpoint which would be the next exon of GENEA.

Is there any tools out there that does that for me, or do I need to write something custom. Hope the draft is not too confusing, but I thought I give it a try. Thx for any help.

Best,

###############################
###############################
Edit:
I figured I explained this far too complicated. In fact all I need is a way to count read across splice-junctions. Let's say I have exon1 exon2 of GENE A. I want to count all reads that map across exon1 and exon2 + all reads were pair 1 maps on exon1 and pair2 maps on exon 2. That's it.