cellranger multi => MTX_TO_H5AD: input file name collision

[nick-youngblut]

Description of the bug

Running CellRanger with all GEX samples, in which there are multiple barcodes per sample, but the all go to the same sample (see the samples & barcodes tables below).
This results in a file name collision at the MTX_TO_H5AD step.
I haven't been able to determine why, based on the pipeline code.

Command used and terminal output

The command:

nextflow run main.nf \
  -ansi-log false \
  -profile singularity \
  -process.executor slurm \
  -process.queue cpu_batch \
  -work-dir /scratch/$(id -gn)/$(whoami)/nextflow-work/scrnaseq \
  --aligner cellrangermulti \
  --skip_cellrangermulti_vdjref \
  --skip_emptydrops \
  --gex_frna_probe_set ${PROBE_REF_DIR}/Chromium_Human_Transcriptome_Probe_Set_v1.0.1_GRCh38-2020-A.csv \
  --cellranger_index ${GENOME_REF_DIR}/refdata-gex-GRCh38-2020-A/ \
  --cellranger_multi_barcodes tmp/sample_barcodes.csv \
  --input tmp/samples.csv \
  --outdir tmp/scrnaseq_output

ERROR ~ Error executing process > 'NFCORE_SCRNASEQ:SCRNASEQ:MTX_CONVERSION:MTX_TO_H5AD (2)'

Caused by:
  Process `NFCORE_SCRNASEQ:SCRNASEQ:MTX_CONVERSION:MTX_TO_H5AD` input file name collision -- There are multiple input files for each of the following file names: barcodes.tsv.gz, features.tsv.gz, matrix.mtx.gz

Relevant files

The samples table (full paths removed for clarity):

sample,fastq_1,fastq_2,feature_type
20240905_ADI_batch3_flex_1,20240905_ADI_batch3_flex_1_S1_L001_R1_001.fastq.gz,20240905_ADI_batch3_flex_1_S1_L001_R2_001.fastq.gz,gex
20240905_ADI_batch3_flex_2,20240905_ADI_batch3_flex_2_S1_L001_R1_001.fastq.gz,20240905_ADI_batch3_flex_2_S1_L001_R2_001.fastq.gz,gex
20240925_ADI_batch5_flex_1,/20240925_ADI_batch5_flex_1_R1_001.fastq.gz,/20240925_ADI_batch5_flex_1_R2_001.fastq.gz,gex
20240925_ADI_batch5_flex_2,/20240925_ADI_batch5_flex_2_R1_001.fastq.gz,/20240925_ADI_batch5_flex_2_R2_001.fastq.gz,gex
20240925_ADI_batch5_flex_3,/20240925_ADI_batch5_flex_3_R1_001.fastq.gz,/20240925_ADI_batch5_flex_3_R2_001.fastq.gz,gex
20240925_ADI_batch5_flex_4,/20240925_ADI_batch5_flex_4_R1_001.fastq.gz,/20240925_ADI_batch5_flex_4_R2_001.fastq.gz,gex

The sample barcodes table:

sample,multiplexed_sample_id,probe_barcode_ids,cmo_ids,description
20240905_ADI_batch3_flex_1,20240905_ADI_batch3_flex_1,BC001|BC002|BC003|BC004,,
20240905_ADI_batch3_flex_2,20240905_ADI_batch3_flex_2,BC001|BC002|BC003|BC004,,
20240925_ADI_batch5_flex_1,20240925_ADI_batch5_flex_1,BC001|BC002|BC003|BC004,,
20240925_ADI_batch5_flex_2,20240925_ADI_batch5_flex_2,BC001|BC002|BC003|BC004,,
20240925_ADI_batch5_flex_3,20240925_ADI_batch5_flex_3,BC001|BC002|BC003|BC004,,
20240925_ADI_batch5_flex_4,20240925_ADI_batch5_flex_4,BC001|BC002|BC003|BC004,,

System information

Nextflow: 24.04.4.5917
Hardward: HPC
Executor: SLURM
Engine: Apptainer
OS: Ubuntu
Pipeline: 2.7.1

[nick-youngblut]

Adding mtx_matrices.view() to MTX_CONVERSION shows that all of the samples have the same file names, which is causing the name collision:

[
  [id:20240925_ADI_batch5_flex_3, ...],
  [
    /path/to/sample1/barcodes.tsv.gz,
    /path/to/sample1/features.tsv.gz,
    /path/to/sample1/matrix.mtx.gz,
    ...
  ]
]

[nick-youngblut]

It might help to include info on how to handle multiple barcodes per sample in the sample barcode table: https://nf-co.re/scrnaseq/2.7.1/docs/usage ("Additional samplesheet for multiplexed samples").

For example:

sample,multiplexed_sample_id,probe_barcode_ids,cmo_ids,description
20240905_ADI_batch3_flex_1,20240905_ADI_batch3_flex_1,BC001|BC002|BC003|BC004,,

From the 10X docs:

If multiple Probe Barcodes were used for a sample, separate IDs with a pipe (e.g., BC001|BC002).