Add subworkflow for ensemble species detection with 3 taxonomic profilers

[ppreshant]

Goal:

Implementing Eddies' Bakeoff project with unified DBs for consensus species detection

The consensus will be determined using a custom script that combines their species table calls. Here's the idea (from Todd) :

• 2 out of 3 tools detecting presence of species
• Same criterea for parent species if looking at strain calls
• Ignore combining the abundance information for now and only focus on presence/absence.
• Should we include the relative abundance from all 3 tools in our final table or will that be more confusing?

What do we do with multiple samples? Should they be combined into a single output?

Plan

Prashant laying the ground-work with:

• Picking these 3 tools: Download the nf-core modules using nf-core modules install <module> on branch: add_taxprofilers

  1. Sylph : rebase part of the older branch: sylph_i73 at commit: f6c45ef
  2. Ganon2
  3. Kraken2

• Direct the pipeline to Eddy's bakeoff DB path:

unified_db_base_dir = "/home/Users/pacbio_bakeoff/data/ref_db/refseq03032025" // path to Eddy's unified databases (Ensemble analysis: species detection)

• Do a prelim test with the proper run command for sylph using test-modules/sylph-test.nf (same as subissue here: test sylph with the unified DB #108)

Continue from here on a new branch with something like ensemble_species_detection_i107

• (fce89de) Test the 3 modules using the respective unified DBs assigned in nextflow.config. Can start with this unfinished script: test-modules/ensemble-test.nf
• Using the test script above, create a new subworkflow that brings together all 3 tools and connects to the main workflow ; activated with analysis_type = species_detection
  • (think/elaborate more) Implement a switch with mode = comprehensive or something to turn on ensemble mode and activate the 3 tools in species_detection branch. Or use a fast mode to run sylph only here?

[ppreshant]

Notes

• Got an issue when updating nf-core: sylph/profile using nf-core modules update sylph/profile
  • Mentioned in nf-core/tools: #3848 and solved using micromamba update questionary to v2.1.1 now (6/Apr/26)
• Dry run of test-modules/ensemble-test.nf works with params.reads = './assets/examples/other_tools_files/sylph/o157_reads.fastq.gz'

(nf_base_env) pbk1@owlet03:~/Somatem$ nextflow run test-modules/ensemble-test.nf                                                      
                                                                                                                                      
 N E X T F L O W   ~  version 25.10.4                                                                                                 
                                                                                                                                      
Launching `test-modules/ensemble-test.nf` [chaotic_joliot] DSL2 - revision: ee7cb1e52f                                                

executor >  local (3)
[94/2980b8] SYLPH_PROFILE (o157_reads)   [100%] 1 of 1 ✔
[87/f1d5ac] GANON_CLASSIFY (o157_reads)  [100%] 1 of 1 ✔
[4a/21511c] KRAKEN2_KRAKEN2 (o157_reads) [100%] 1 of 1 ✔
Completed at: 17-Apr-2026 14:03:03
Duration    : 8m 12s
CPU hours   : 1.6
Succeeded   : 3

• moved into a subworkflow and got errors due to using new nf-core modules: Is caused by incompatibility of new topic: method of storing versions information as a value instead of a versions.yml file as previously. The old validation in utils_nfcore_pipeline module is still expecting a yml file and the tuple [process name, executable, version] throws it off with this error.

Apr-20 11:52:05.985 [Actor Thread 112] DEBUG nextflow.Session - Session aborted -- Cause: No signature of method: org.yaml.snakeyaml.Yaml.load() is applicable for argument types: (ArrayList) values: [[ORCHESTRATE_SOMATEM:SOMATEM:SPECIES_DETECTION:KRAKEN2_KRAKEN2, kraken2, 2.1.6]]
Possible solutions: load(java.io.InputStream), load(java.io.Reader), load(java.lang.String), load(java.io.File), load(java.nio.file.Path), load(java.io.File)

• fixed (3904f3e): by removing ch_versions collection from SPECIES_DETECTION subworkflow and collecting the channel.topic("versions") separately and mixing it in in somatem.nf
• (comprehensive template update; FOR LATER) Other suggestion from seqera AI was to update the template that is originally used for creating the pipeline but it requires lot of manual steps now to bring in sync including a TEMPLATE branch. read nf-core/template-syncs.
• Successful run of the subworkflow with the 3 tools:

[ef/44ed72] ORCHESTRATE_SOMATEM:SOMATEM:PREPROCESSING:RawNanoPlot (mock20)        [100%] 2 of 2 ✔
[56/65a3f3] ORCHESTRATE_SOMATEM:SOMATEM:PREPROCESSING:CHOPPER (mock9)             [100%] 2 of 2 ✔
[08/05f84f] ORCHESTRATE_SOMATEM:SOMATEM:PREPROCESSING:FinalNanoPlot (mock9)       [100%] 2 of 2 ✔
[80/06b239] ORCHESTRATE_SOMATEM:SOMATEM:SPECIES_DETECTION:SYLPH_PROFILE (mock9)   [100%] 2 of 2 ✔
[38/2e20cc] ORCHESTRATE_SOMATEM:SOMATEM:SPECIES_DETECTION:GANON_CLASSIFY (mock9)  [100%] 2 of 2 ✔
[0e/f5823f] ORCHESTRATE_SOMATEM:SOMATEM:SPECIES_DETECTION:KRAKEN2_KRAKEN2 (mock9) [100%] 2 of 2 ✔
-[nf-core/somatem] Pipeline completed successfully-
Completed at: 21-Apr-2026 15:26:45
Duration    : 13m 51s
CPU hours   : 2.7
Succeeded   : 12