1
$\begingroup$

I run the cellranger count with this code:

cellranger count \
  --id=run_count_SRR11537950 \
  --transcriptome=refdata-gex-GRCh38-2024-A \
  --create-bam=true \
  --fastqs=SRR11537950 \
  --sample=SRR11537950 \
  --chemistry=fiveprime \
  --localcores=1 \
  --localmem=5

and error showed:

[error] Pipestance failed. Error log at: run_count_SRR11537950/SC_RNA_COUNTER_CS/SC_MULTI_CORE/MULTI_GEM_WELL_PROCESSOR/COUNT_GEM_WELL_PROCESSOR/_BASIC_SC_RNA_COUNTER/_MATRIX_COMPUTER/ALIGN_AND_COUNT/fork0/chnk0-u0de8b9d2e0/_errors
Log message: Job failed in stage code
signal: killed

What should I do? Thanks in advance!

$\endgroup$

1 Answer 1

2
$\begingroup$

I am almost certain that the localmem parameter is the issue here. CellRanger uses STAR aligner under the hood which is resource-hungry. We usually allocate 18 cores and at least 40GB of RAM to CellRanger while disabling all of its secondary analysis, such as clustering and tSNE. Five GB of RAM for sure are not sufficient.

$\endgroup$
3
  • $\begingroup$ I have to use upwards of 100GB to ensure things definitely run to completion. $\endgroup$
    – Ram RS
    Commented Aug 12 at 15:01
  • $\begingroup$ @RamRS I just checked some older logs and we routinely use 18 cores and 40GB of RAM, but disabling all these secondary analysis CR can do (like tSNE etc). Still takes several hours to complete, but it usually works just fine. $\endgroup$
    – ATpoint
    Commented Aug 12 at 15:16
  • 1
    $\begingroup$ We also use 40 GB RAM when running STAR with the human or mouse genome for regular RNA-Seq. It might be able to do with less memory if the parameters are tweaked, but that may not be possible when it's evoked through CellRanger. $\endgroup$
    – Cloudberry
    Commented Aug 12 at 19:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.