Hi, maybe a noob question but sarek doesn't run st...
# sarek
a
Hi, maybe a noob question but sarek doesn't run strelka as a variant caller. It just skips completely the variant calling step (unless there are other caller). I run in tumor only mode. Updating Nextflow and sarek to 3.7.1 did not help. Any way to troubleshoot this ? Thanks
m
#C043FMKUNLB here on nf-core, so nothing to worry about
strelka does not run in tumor-only mode
a
Thanks Maxime. I saw an issue about that that was closed after a hackathon though
m
could you link the issue?
That might remind me of the context
a
m
Oh yeah, so long story short, we used to run strelka in germline mode on the tumor sample for tumor only variant calling. but that was only because that's what my old lab was doing, and def not best practices
that being said, you can always trick sarek into believing your tumor samples are actually normal by modifying the status in the recalibrated csv files and run strelka if that's what you really want
a
I see ! It's not really tumor sample, more like germline with very low vaf (1%). Would strelka in germline mode work ?
m
I would say yes
the figure says so
but the docs says no
@Rike Hanssen screams
the true test would be checking the code
so yeah, code confirm strelka runs in germline mode for each germline sample
a
Thanks, I'll try germline then :)
m
You shouldn't have to rerun the whole mapping step and you could restart from the variant calling step
that would be way faster
a
Yes,
-resume
is my go-to tool
m
yeah, but in that case, changing the status would kill the resumability of the pipeline
a
Right, good point !
As a side note, do you know if there is a way to increase the sensibility of strelka ? Otherwise, I was thinking of using a custom model, like https://github.com/Illumina/strelka/blob/v2.9.x/docs/userGuide/trainingGermlineEmpiricalScore.md
m
also, we do use a tiny mismatch on the alignment with
-B 3
on the tumor samples
a
I've overidden that (not sure why but it seems to work better without it)
🙏 1
r
https://nf-co.re/docs/usage/getting_started/configuration#understanding-and-modifying-tool-arguments see here for accessikng and modifying tool arguments that are not exposed by the pipeline
i
I remembered correctly that there was a funny slide about this
😀 1
m
🤣
good memory @Igor Trujnara