perbase in recount-unify docker image crashes if there are > 1032 chromosomes

[rmflight]

In the current version of perbase in the recount-unify docker image (v 1.1.1), it seems to crash if there are too many chromosomes.

I'm trying to get monorail working for zebrafish, and the current primary assembly has 993 chromosomes. Add in the 99 from ERCC and SIRV, and it pushes it out 1092.

When the merged junctions step runs, right around chromosome # 1032, there is sequence spit out into the .err file, and it crashes and halts the process.

I'm almost certain this is the issue, because after I trim down to the 25 primary chromosomes, and run through everything again, the junction step runs without errors. And the chromosome where the errors occur is one of the ERCC ones, not any of the zebrafish ones.

[ChristopherWilks]

Hi @rmflight,

Thanks for the info, cool to hear you're using Monorail for Zebrafish!

I'm not surprised perbase is crashing, it was a quick 'n dirty solution at the time, and I intended it mainly to work with human and mouse chrom's 1-22, X, Y, and MT, though it can support some number more as you found.

I'm traveling at the moment, but I will prioritize fixing and updating unifier docker image.
I'll let you know.

Chris

[rmflight]

Thanks Chris!

Just so we are clear, I'm pretty sure the current version of perbase on GH works fine, because I can compile it, and run it with the same commands that were failing in the snake make output, and it worked fine.

And most of the limits listed in the code seem like they should handle the zebrafish case just fine, if I'm reading them right.

I just can't tell what version is actually in the unify docker.

[ChristopherWilks]

Thanks for the clarification, I've forgotten about my later updates to that program.

So yes, I agree, it looks like the current version of the source for perbase is probably fine for what you're doing (as you said), as the # of chroms supported is quite high given the limits.

That expanded # of chroms change was made in 2024:
langmead-lab/recount-unify@f0bb51e

However, v1.1.1 of the unifier docker image was pushed in 2022, so almost for sure doesn't have the expanded chroms limit and very likely uses one of the older, much smaller limits, as you're seeing.

Of course I did these things myself but it's been long enough that I don't recall all the details.

I'll need to double check that the new version isn't doing anything particularly different (other than fixing bugs and expanding limits) before I push a new container image with updated version of perbase.