Control generation of --output-stats when running umi-tools dedup

[lars-work-sund]

Description of feature

umitools dedup uses large amounts of memory and runs slowly. To speed it up it is recommended to only run it on a single chromosome, see the FAQ point number 4.

Currently the option is hard-coded, see github

The suggested solution to only run on a single chromosome would cause the pipeline to only analyze a single chromosome.
Since --chrom seems to be mostly for debugging purposes, see here

I suggest either making the --output-stats optional, or running a second round of deduplication on a single chromosome to generate the output stats.

[lars-work-sund]

Running the analysis on a single chromosome is more tricky than I expected. In the NFCORE_RNASEQ:RNASEQ:DEDUP_UMI_UMITOOLS_TRANSCRIPTOME:UMITOOLS_DEDUP step each transcript is considered a contig.

[drpatelh]

Ok. In that case, we definitely need to fix the module in the next release.

@lars-work-sund tried this config below which failed as reported in the comment above:

process {
    withName: 'NFCORE_RNASEQ:RNASEQ:DEDUP_UMI_UMITOOLS_GENOME:UMITOOLS_DEDUP' {
        cpus = 12
        memory = '72.GB'
        time = '16.h'
        ext.args = '--chrom=chr22'
    }

    withName: 'NFCORE_RNASEQ:RNASEQ:DEDUP_UMI_UMITOOLS_TRANSCRIPTOME:UMITOOLS_DEDUP' {
        cpus = 12
        memory = '72.GB'
        time = '16.h'
        ext.args = '--chrom=chr22'
    }
}

We will need to update the nf-core/module to make --output-stats optional and possibly add a parameter to the pipeline to control this behaviour e.g. --umitools_dedup_output_stats which is off by default.

[drpatelh]

nf-core/module updated in nf-core/modules#1689
Fixed in pipeline in drpatelh@8102aa3

Added a new boolean parameter to the pipeline called --umitools_dedup_stats that when provided will generate the output stats. By default, the stats won't be generated.