stjude / CICERO

CICERO: a versatile method for detecting complex and diverse driver fusions using cancer RNA sequencing data.
https://stjude.github.io/CICERO
Other
35 stars 19 forks source link

Performance issues #115

Closed YussAb closed 12 months ago

YussAb commented 1 year ago

Dear St.Jude Researchers,

lately I was trying you tool for rnaseq fusion detection. I made a run with a pair-end bulk rnaseq sample (150bp) with 16 milions reads. I used 16 threads and 64 GB of memory with the following command:

Cicero.sh -n 16  \
                 -b $PWD/sample.Aligned.sortedByCoord.out.bam \
                 -g "GRCh38_no_alt" \
                 -r $PWD/reference/  \
                 -o sample \
                 -j sample.Aligned.sortedByCoord.out.bam.junctions.tab.shifted.tab

I takes a lot of time to finish 10 h 55 m 48 s.

Furthermore it gives me a huge rate of ambiguous calls. (I attached the output file)

Am I doing something wrong? Or it is normal?

Thank you in advance! Best regards, Youssef

sample.Aligned.sortedByCoord.out.final_fusions.txt

adthrasher commented 1 year ago

The run time is fairly normal. From the output, I see 87 calls, of which, 12 were labeled as HQ. You can also use the score column to filter to higher confidence calls.