Enable multiple assembler runs by default and compare results

[Daniel-VM]

Description of feature

Description

Update the pipeline to allow running multiple assemblers in parallel (e.g., Flye, MEGAHIT, Trycycler) followed by polishing, so users can compare results using tools like QUAST and BUSCO.

Goal

• Add a config setting (e.g., --run_multiple_assemblers)
• Ensure assemblies go through polishing stages
• Summarize results in QUAST and BUSCO reports for comparison
• Provide documentation

Note: This depends on having more than one assembler for each short, long, and hybrid assembly mode.

[ACCakut]

Ryan R. Wick's autocylcer does exactly this with the auto-script: Automated_Autocycler_Bash_script_by_Ryan_Wick

[d4straub]

Ah thanks for linking autocycler, this seems to be very interesting for bacass I think. It seems to be basically Trycycler just faster & easier to build into a pipeline, found here a nice overview of automated and optional hands-on tasks.
Having said that, I didnt try out autocycler yet.
Converting that bash script into a subworkflow for bacass seems a good and relatively easy solution indeed.

[Daniel-VM]

Thanks a lot for the proposal. I agree. I haven’t tested it yet either but it looks like a solid fit for a subworkflow! 🙌🏾

[d4straub]

Actually, reading the issue again, there might be several different proposals:
(1) produce multiple assemblies in parallel to produce several assemblies for each sample and choose manually "the best" assembly (sometimes not a trivial task, prone to error)
(2) run mutiple assemblers and combine all assemblies into one assembly -> trycycler or autocycler
I think the original thought was (1) and we are now discussing (2). But both ways seem valid ideas to me. In the past, I often run bacass with several setting (e.g. once with Unicycler, once with Dragonflye) and check what appears to be better result. But in my experience trycycler was always best, however it also needs most data and is laborious to run.

[ACCakut]

Yes, that's right. The word "exactly" was exagerated. Autocycler is focussing on automated consensum. But, the different assemblies are generated and are on hand for manual analysis and also allows manual curation.

Autocycler is considered the successor of Trycycler by Ryan (https://github.com/rrwick/Trycycler/wiki, https://github.com/rrwick/Autocycler) and gives great results for our Bacillus and Neisseria reads. Ryan also shows proof of Autocycler's superiority: see Figure 2

If that helps, I created a minimal Singularity/Apptainer container for Autocycler (alternative to autocycler biocontainer; I am new to this ;-):

Bootstrap: docker
From: mambaorg/micromamba

%post
    # install gnu parallel and git
    apt update
    apt install parallel git bash --yes
    apt upgrade --yes
    # mute citation nag screen
    echo 'will cite' | parallel --citation || true

    # clone from GitHub
    cd /opt
    git clone https://github.com/rrwick/Autocycler.git

    # install conda environment
    micromamba config append channels conda-forge
    micromamba config append channels defaults
    micromamba install -y -n base -f /opt/Autocycler/pipelines/Conda_environment_file_by_Ryan_Wick/environment.yml

    micromamba run plassembler download -d "$CONDA_PREFIX"/plassembler_db

    # clean
    micromamba clean -aqy

    micromamba shell init --shell bash --root-prefix ~/.local/share/mamba

    ## optional
    # save date of creation
    NOW=`date`
    echo "export NOW=\"${NOW}\"" >> $SINGULARITY_ENVIRONMENT

%environment
    # just change to the directory and wait for commands
    cd /opt/Autocycler/pipelines/Automated_Autocycler_Bash_script_by_Ryan_Wick/

%runscript
    #!/bin/bash
    echo "Container was created $NOW"
    echo "Arguments received: $*"
    echo "Running autocycler_full.sh"

    eval "$(micromamba shell hook --shell bash)"
    micromamba activate base
    bash /opt/Autocycler/pipelines/Automated_Autocycler_Bash_script_by_Ryan_Wick/autocycler_full.sh "$@"


%labels
    Author ACCakut
    Date 2025-07-18
    Version v0.1
    Name autocycler

%help
   A container running autocycler auto-script (https://github.com/rrwick/Autocycler/tree/main/pipelines/Automated_Autocycler_Bash_script_by_Ryan_Wick) installed on micromamba (miniconda alternative)

[d4straub]

Thats a handy container. However, nf-core pipelines are supposed to be rather flexible and modularized. Therefore I didnt want to suggest using that exact script but rather as a template for a subworkflow that contains several modules, each of which with its own software container. So I hope we can go with the default containers at bioconda (less headache with maintenance!).