Make BigWig files and use RSeQC geneBody_coverage2.py about rnaseq HOT 6 CLOSED

Hi Justin,

Sorry for this problem. We’ve hit it a bunch of times as well, it can be annoying. We’ve already split this task out by itself for this reason and subsample the BAM file down to one million reads first to limit the memory usage. I suspect here that it’s actually this subsampling method which is eating all of the memory. We tried to be clever and randomise the subsampling but it would perhaps be better to just take the first one million reads instead. I’ll take a look into this.

Phil

from rnaseq.

Hi @justinjeyakani,

I just pushed an update in #37 which should hopefully help the memory problem. It would be great if you could pull the latest version of the code and give it a try.

I've tested with a minimal dataset and it seems to work ok. I'd like to see how many reads the subsampled BAM files are ending up with using the current default subsampling target of ~10GB - we may be able to tweak this either up (to get better / more representative gene body coverage profiles) or down (for faster execution if the profiles aren't affected).

I'm trying to test on some larger data now myself, but our HPC is having problems again 😣

Phil

from rnaseq.

Dear Phil,

Sure, will pull the latest and test with large bam files.
Any concern to include generate bigwig and run geneBody_coverage2.py on it. This serves to make use of the entire bam with less memory and an additional feature of the pipeline for visualization of the alignment.

from rnaseq.

Hi Justin,

Yes, perhaps you're right. I skipped over this suggestion initially because we'd already half-implemented the subsampling solution. But as you say, perhaps doing a bigwig conversion is the better approach (and would be simpler pipeline code too).

Phil

from rnaseq.

https://deeptools.readthedocs.io/en/develop/content/tools/bamCoverage.html

Should be sufficient to get this going?

from rnaseq.

This is now in the dev branch of the pipeline :-)

from rnaseq.