junction reads missing

[pabloacera]

Hi,

Thanks for maintaining this pipeline!
I have an issue regarding junction reads. When I run the recount-pump and I look for junction reads spanning from acceptor 31853031 in chr17 in the file 'X!X!hg38!local.sjout.zst' I see this:

This is strange because it is missing reads going from the canonical donor 31856798. Then I use STAR outside the pipeline with exact same filters used in monorail and I indeed see the junction reads in the SJ.out generated by STAR.

This is causing the unify pipeline not counting these junction reads and the sqlite generated is missing them. This does not happens for all the samples, which is strange, and it is causing a bunch of downstream issues. I tried to look on the recount-pump logs to see any errors or warnings but I cannot see anything weird. Please do you have any inputs about this? I will apreciate much!! thanks a lot in advance
std.txt
This is the log of the pump. please let me know if anything went wrong. Thanks!

[ChristopherWilks]

Hi @pabloacera,

My quick read of the issue is that there is something that's different between the STAR run within Monorail pump and the manual/external one you're running. This is down to the *local.sjout.zst file not containing the expected junction. If you had only not seen it in the later files I'd suspect a filtering issue in Monorail Unify rather than a STAR command difference.

Could you post/double check your external STAR params just so we can make sure we're on the same page? Also, as part of that, could you post/check the STAR reference index you're using?

I ask, because the only root cause I can think of at the moment is that there's some difference in either the params or the STAR index, or both, around annotated junctions, e.g. the external run is either doing a 2nd pass alignment and/or using an annotation whereas the Monorail pump one is not doing either of those things.

Thanks,
Chris