feat(fastq_qc_trim_filter_setstrandedness): Add Bowtie2 as alternative rRNA removal tool

[pinin4fjords]

Summary

Add bowtie2 as a third option for rRNA removal alongside sortmerna and ribodetector, providing users with more flexibility in choosing their preferred alignment-based rRNA filtering approach.

Implementation Details

Paired-end Read Handling

For paired-end reads, proper rRNA removal requires filtering pairs where either mate aligned to rRNA. The implementation:

1. Aligns reads with BOWTIE2_ALIGN_PE
2. Filters BAM with samtools view -f 12 (flag 12 = both mates unmapped)
3. Converts filtered BAM back to FASTQ with samtools fastq

Single-end Read Handling

For single-end reads, bowtie2's --un-gz output is used directly via save_unaligned=true.

U→T Conversion

rRNA reference FASTAs may contain uracil (U) since they're RNA sequences, but sequencing reads contain thymine (T). The implementation converts U→T when building the bowtie2 index.

Changes

• Add BOWTIE2_ALIGN, BOWTIE2_ALIGN_PE, and BOWTIE2_BUILD module imports
• Add SAMTOOLS_VIEW and SAMTOOLS_FASTQ for paired-end filtering
• Add ch_bowtie2_index input channel for pre-built indexes
• Add make_bowtie2_index parameter for on-the-fly index building
• Implement single-end/paired-end branching with appropriate filtering strategy
• Update meta.yml with bowtie2 in ribo_removal_tool enum and new components
• Add test cases for both paired-end and single-end bowtie2 rRNA removal
• Update nextflow.config with bowtie2 and samtools process configurations

Test Results

Test	Input	After Trim	rRNA Detected	Output
PE fastp bowtie2	4169 pairs	1137 pairs	17 mates (10 pairs)	1127 pairs
SE fastp bowtie2	4169 reads	4162 reads	10 reads	4152 reads

The paired-end bowtie2 test achieves the same output (1127 pairs) as sortmerna, correctly removing all 10 synthetic rRNA pairs.

Test plan

• Run existing sortmerna and ribodetector tests to ensure no regression
• Run paired-end bowtie2 test
• Run single-end bowtie2 test
• Verify bowtie2 log output is captured for MultiQC

Proposed refactor before merge

Note: This subworkflow has grown substantially with three rRNA removal tools (sortmerna, ribodetector, bowtie2), each with their own index building and filtering logic. I propose we factor out the rRNA removal components into a dedicated fastq_remove_rrna subworkflow before merging this PR, once we confirm bowtie2 should be retained as a method.

🤖 Generated with Claude Code

[suhrig]

LGTM! Thanks for this great addition (at amazing speed)!

I agree that moving the rRNA removal into a separate space would be sensible, but if bandwidth is limited, then it should be okay the way it is.

[pinin4fjords]

@suhrig Factoring out complete!

The failure in CI is because the factoring-out triggered the issue described in hzi-bifo/RiboDetector#59 for ribodetector (I think). I think we'll need that fix to propagate to Conda channels before we can set a seed for testing purposes and this can be completed.

Edit: I also hadn't sorted the reads

[maxulysse]

We should be snapshoting the versions. There's a snippet for that

[pinin4fjords]

sorry, it was set to auto-merge! Will pr a fix

[pinin4fjords]

#9478