QUAST QC step fails when running de novo assembly without reference genome

[wbau]

Description of the bug

Bug Summary

When running genomeassembler with --use_ref false for de novo assembly, the QUAST QC process is silently skipped due to a channel joining issue in the workflow logic. This also causes the final REPORT step to fail.

Current Behavior

The QUAST module itself (modules/local/quast/main.nf) correctly handles no-reference mode through conditional parameters:

def features = use_gff ? "--features ${gff}" : ''
def reference = use_fasta ? "-r ${fasta}" : ''
def reference_bam = params.use_ref ? "--ref-bam ${ref_bam}" : ''

However, the workflow in subworkflows/local/qc/quast/main.nf always attempts to join the inputs_references (containing ref_fasta and ref_gff) and aln_to_ref channels, even when they are empty/null:

inputs
    .map { row -> [row.meta, row.ref_fasta, row.ref_gff] }
    .set { inputs_references }

assembly
    .join(inputs_references)     // <- This fails when ref_fasta/ref_gff are null
    .join(aln_to_ref)            // <- This is also empty when use_ref=false
    .join(aln_to_assembly)
    .set { quast_in }

This causes the channel join to fail, and QUAST never executes.

Expected Behavior

QUAST should generate assembly statistics even without a reference genome, as the module already supports this functionality.

Workaround / Example Solution

Add conditional logic to handle de novo mode in subworkflows/local/qc/quast/main.nf:

if (params.quast) {
    if (params.use_ref) {
        // Original behavior with reference
        inputs
            .map { row -> [row.meta, row.ref_fasta, row.ref_gff] }
            .set { inputs_references }

        assembly
            .join(inputs_references)
            .join(aln_to_ref)
            .join(aln_to_assembly)
            .set { quast_in }
    }
    else {
        // De novo mode: skip reference channels
        assembly
            .join(aln_to_assembly)
            .map { meta, consensus, bam ->
                [meta, consensus, [], [], [], bam]
            }
            .set { quast_in }
    }
    
    QUAST(quast_in, params.use_ref, false)
    QUAST.out.results.set { quast_results }
    QUAST.out.tsv.set { quast_tsv }
    QUAST.out.versions.set { versions }
}

Testing

I've tested this fix with: ONT + Illumina hybrid assembly, with '--use_ref false' and '--lift_annotations false'. QUAST now successfully runs and generates assembly statistics.

Command used and terminal output

nextflow run nf-core/genomeassembler \
  --input samplesheet.csv \
  --ont \
  --short_reads \
  --assembler flye \
  --use_ref false \
  --lift_annotations false \ 
  --genome_size 20000000 \
  --quast \
  ...

Relevant files

No response

System information

No response

[nschan]

Hi,

that is interesting, thanks for reporting. I am currently in the process of a larger rewrite of genomeassembler (#171) and I will see if the same behaviour occurs there. I realize that there are other issues with the quast module (e.g. it references a param that is not input) and will see if I can switch to the nf-core module, and handle this logic correctly.

[nschan]

I have checked, and the upcoming pipeline version handles this correctly https://github.com/nf-core/genomeassembler/pull/171/files#diff-e70f72ad408ddbfa665ad2ab9c6b6ab5fe98c4240faba7dc234b91bea579598aR16-R31
I have also addressed the issue that the quast module referenced a param. I will leave this issue open as this bug currently exists in the release verison, and I doubt that there will be another release of this pipeline before v2.