`CELLRANGER_MKVDJREF` and `CELLRANGER_MULTI` failing in the tests

[delfiterradas]

Description of the bug

Hi,

I am encountering two errors when running the cellrangermulti test profile.

1. CELLRANGER_MKVDJREF error
   The first one is in the CELLRANGER_MKVDJREF module:

[error] An empty constant regions file was generated/detected for your custom species. Please check if there are hits to the IMGT database or run cellranger vdj in denovo mode without reference.

The issue however seems to come from the CELLRANGER_MKGTF module. It can be avoided by adding the following configuration:

process {
    withName: 'NFCORE_SCRNASEQ:SCRNASEQ:CELLRANGER_MULTI_ALIGN:CELLRANGER_MKGTF' {
        ext.args = "--attribute=gene_biotype:protein_coding --attribute=gene_biotype:IG_V_gene --attribute=gene_biotype:IG_J_gene --attribute=gene_biotype:IG_D_gene --attribute=gene_biotype:IG_C_gene --attribute=gene_biotype:TR_V_gene --attribute=gene_biotype:TR_J_gene --attribute=gene_biotype:TR_D_gene --attribute=gene_biotype:TR_C_gene"
    }
}

2. CELLRANGER_MULTI error
   After applying the above modification, the pipeline proceeds further but then fails in the CELLRANGER_MULTI module with:

These library types are not compatible with Flex: Multiplexing Capture

This appears to be related to how the FRNA probeset is added in main.nf. The error can be avoided by modifying the logic introduced in this commit:

nf-core/modules@2501026

Specifically, changing:

frna_probeset = gex_frna_probeset_name != '' ? "probe-set,./$gex_frna_probeset_name" : ''

to

frna_probeset = include_frna && gex_frna_probeset_name != '' ? "probe-set,./$gex_frna_probeset_name" : ''

I am not entirely sure what side effects this change might have on other use cases, or whether this behavior is expected for the test_cellrangermulti profile.
Any guidance on this would be very helpful.

Command used and terminal output

nextflow run main.nf -profile singularity,test_cellrangermulti,human --outdir results

Relevant files

No response

System information

No response

[fmalmeida]

Hi @delfiterradas ,
Reopening this one following the theead we had in the modules repository (nf-core/modules#10922).

We need to check if the test is correct, in the cellranger outputs from before if the frna probeset in the test configured is really used or not.

For example is it there by mistake and it was never actually used before?

If so, then the solution would actually be removing it from the test config.

If on the other hand, it was indeed used and not having it changes the output, then we must come back to ensure the logic allows both cases.

[mahesh-panchal]

To get the test to run through

nextflow run main.nf -profile test_cellrangermulti,human,docker -dump-hashes -c testing/custom.config --outdir results

I had to add:

#
# Copy fb_reference to avoid symlink corruption
#
if fb_reference_path:
    fb_ref_parts = fb_reference_path.split(",")
    if len(fb_ref_parts) == 2:
        original_fb_path = Path(fb_ref_parts[1])
        if original_fb_path.exists():
            copy_fb_path = original_fb_path.parent / f"fb_reference_copy.csv"
            copy_fb_path.write_text(original_fb_path.read_text())
            fb_reference_path = f"{fb_ref_parts[0]},{copy_fb_path}"

to the cellranger_multi.py

and

        //
        // Create per-sample probeset channel
        // Use the FRNA CSV presence to determine if sample is Flex
        //
        def probeset_file = ch_gex_frna_probeset  // Capture the file value
        ch_frna_sample_csv
            .map { frna_csv ->
                // If sample has FRNA CSV (is Flex), pass probeset; otherwise empty
                frna_csv ? probeset_file : []
            }
            .set { ch_per_sample_probeset }
            
        // Create per-sample fb_reference channel
        // Only pass to samples that have AB or CRISPR data (not Flex)
        // Use FIFO order by checking AB and CRISPR channels directly
        //
        def fb_reference_file = ch_fb_reference
        ch_grouped_fastq.ab
            .map { meta, fastqs -> 
                // Check if AB data exists (not EMPTY file)
                fastqs.toString().contains('EMPTY') ? false : true
            }
            .concat(
                ch_grouped_fastq.crispr
                    .map { meta, fastqs ->
                        // Check if CRISPR data exists (not EMPTY file)  
                        fastqs.toString().contains('EMPTY') ? false : true
                    }
            )
            .buffer(size: 2)  // Collect [has_ab, has_crispr] for each sample
            .map { has_fb_list ->
                // If sample has AB OR CRISPR, pass fb_reference
                has_fb_list.any { it == true } ? fb_reference_file : []
            }
            .set { ch_per_sample_fb_reference }

        //
        // MODULE: cellranger multi
        //
        CELLRANGER_MULTI(
            ch_grouped_fastq.gex.map{ it[0] },
            ch_grouped_fastq.gex,
            ch_grouped_fastq.vdj,
            ch_grouped_fastq.ab,
            ch_grouped_fastq.beam,
            ch_grouped_fastq.cmo,
            ch_grouped_fastq.crispr,
            ch_cellranger_gex_index,
            ch_per_sample_probeset,
            ch_gex_target_panel,
            ch_cellranger_vdj_index,
            ch_vdj_primer_index,
            ch_per_sample_fb_reference,
            ch_beam_antigen_panel_csv,
            ch_beam_control_panel_csv,
            ch_gex_cmo_set,
            ch_cmo_barcode_csv,
            [],
            ch_frna_sample_csv,
            ch_ocm_barcode_csv,
            params.skip_cellranger_renaming
        )
        ch_versions = ch_versions.mix(CELLRANGER_MULTI.out.versions)

to CELLRANGER_MULTI_ALIGN.

[fmalmeida]

Hi @mahesh-panchal ,
Cool!

Thanks for the help on it. Makes super sense in having the checkers in the workflow.

Looks neat.

@delfiterradas ,
We can proof test a little bit more with other cellranger multi use cases so we can then have the solution implemented :)