rsem/star index fails with large genome

[ChristopherBarrington]

Error when using --aligner star_rsem in the RSEM_PREPAREREFERENCE process:

  EXITING because of FATAL PARAMETER ERROR: limitGenomeGenerateRAM=31000000000is too small for your genome
  SOLUTION: please specify --limitGenomeGenerateRAM not less than 33524373088 and make that much RAM available

This was trying to index the GRCm38 genome, pipeline version 2.0 (-r 2.0)

[drpatelh]

Thanks @ChristopherBarrington. So as we have discussed on Crick Slack this would be a relatively easy fix if RSEM add a separate parameter to provide additional options when building the indices with STAR. A couple of other issues have been reported along the same lines but it appears that the authors are recommending users to change the source code or haven't responded to issues e.g. here and here. I have managed to build the index successfully by installing RSEM and STAR via Conda and changing the source code to increase limitGenomeGenerateRAM. This however, isn't ideal (and should really be fixed upstream) but a more flexible solution would be to change the nf-core/module to detect whether the user has provided --star* specific flags, to build the STAR index with STAR in order to be able to provide additional options and then just use RSEM to create the remainder of the files it needs. This means the approach will work for others using the same module file for example with Bowtie, Bowtie2 or HISAT2.

[drpatelh]

Updated the nf-core/module for rsem-prepare-reference nf-core/modules#87 so once that is merged I will re-install it in this PR with the fix for the rnaseq pipeline #512