BaseMaskConfigException: You cannot assign UMIs ('M') if using bcl2fastq

[messersc]

Cases we might encounter with Samplesheets from digestiflow-server:

1. identical base mask without UMI -> bcl2fastq2
2. identical base mask with UMI -> Picard tools
3. more than one base mask without UMI -> bcl2fastq2
4. more than one base mask with (some) UMI -> see error below. Picard can only use one bask mask, so we chose bcl2fastq2, but that does not allow for UMIs.
5. multiple more configurations one could think of.

The traceback from the bug that triggered this discussion:

[Fri Nov 26 15:43:41 2021]
rule demux_bcl2fastq2_par:
    input: /fast/work/projects/raw-data-bih-genomics-charite/DEMUX_WORK_DIR/211124_A01077_0148_AHKLFLDRXY/illumina_basesmask/151T8B9M8B151T/SampleSheet.csv
    output: /fast/work/projects/raw-data-bih-genomics-charite/DEMUX_WORK_DIR/211124_A01077_0148_AHKLFLDRXY/illumina_basesmask/151T8B9M8B151T/bcl2fastq2.done
    jobid: 5
    wildcards: bases_mask=151T8B9M8B151T
    threads: 8

/fast/users/svc-bih-cubi-demux_c/work/miniconda3/bin/python /fast/scratch/users/svc-bih-cubi-demux_c/tmp/res-transfer-1.research.hpc.bihealth.org/2021-11-26/tmpmwxwsrx3-cubi-demux/.snakemake/scripts/tmpbcm5miv4.wrapper.py
Activating conda environment: /fast/scratch/users/svc-bih-cubi-demux_c/tmp/res-transfer-1.research.hpc.bihealth.org/2021-11-26/tmpmwxwsrx3-cubi-demux/.snakemake/conda/76bbdd51
Traceback (most recent call last):
  File "/fast/scratch/users/svc-bih-cubi-demux_c/tmp/res-transfer-1.research.hpc.bihealth.org/2021-11-26/tmpmwxwsrx3-cubi-demux/.snakemake/scripts/tmpbcm5miv4.wrapper.py", line 38, in <module>
    bases_mask_illumina = return_bases_mask(planned_reads, bases_mask)
  File "/fast/work/users/svc-bih-cubi-demux_c/digestiflow-demux/digestiflow_demux/bases_mask.py", line 92, in return_bases_mask
    raise BaseMaskConfigException("You cannot assign UMIs ('M') if using bcl2fastq")
digestiflow_demux.bases_mask.BaseMaskConfigException: You cannot assign UMIs ('M') if using bcl2fastq

In principle, the flowcells that fall under point 4. from the above list might need two digestiflow-demux runs.
However, for the specific flowcell, picard should be fine to use and I need to check why we end up with two sets of demux_reads if one should suffice.