Remove requirement to run `seabreeze` within the repository directory

[dr-joe-wirth]

The problem

Currently, users are required to place all of their input files into a directory named data within the repository itself. This means that users would need to replace (or move) this directory if a second run was being performed on different inputs. This introduces opportunities for user error and feels kind of clunky. It would significantly improve the user experience if users could run the program in the directory of their choice.

Proposed solution

Although by default, snakemake looks for Snakefile in ./, this is not required. It is possible to run snakemake elsewhere by using following syntax:

snakemake -d ./my/working/directory/ --snakefile path/to/seabreeze/Snakefile --cores 42

Within the Snakefile, you can reference the directory containing the snakefile with variable workflow.basedir and the working directory with the variable workflow.overwrite_workdir. This will allow you to do something like:

query_path = os.path.join(workflow.overwrite_workdir, '02_genomes', '{sample}.fasta')
script = os.path.join(workflow.basedir, 'bin', 'scripts', 'find_reindex_bases.py')

instead of hardcoding it like this

query_path = 'data/02_genomes/{sample}.fasta')
script = 'bin/scripts/find_reindex_bases.py'

This is part of the review at openjournals/joss-reviews#8065

[ira-zibbu]

Hi @dr-joe-wirth,

I agree with the issue you have raised and it would improve usability. Your suggested solution would be a good way to implement this. Will work on it.

[ira-zibbu]

Hi,

I have updated seabreeze so that it can be run from any directory. It was implemented in the following way:

• seabreeze now uses pip to build.
• seabreeze uses a CLI-wrapper src/seabreeze/seabreeze.py around the Snakefile to execute the desired workflows. The path to the Snakefile is determined from where the package is built

def get_snakefile():

    """
    Fetch path to the seabreeze directory where the snakefile
    args: None
    returns: Path to the snakefile
    """

    seabreeze_package_path = importlib.resources.files(seabreeze)
    print(f"seabreeze package path is {seabreeze_package_path}")

    snakefile_path = os.path.join(seabreeze_package_path, "Snakefile")
    return snakefile_path

The path to the Snakefile is passed as an argument to the snakemake command.

cmd = (
        "snakemake --snakefile {snakefile} --use-conda --cores {cores} {workflow} --config {config}"
        " {args} "
    ).format(
        snakefile=get_snakefile(),
        cores=threads,
        workflow=workflow,
        args=" ".join(snakemake_args),
        config=config_options
    )

• Additionally, the user can also specify the directory for output, if they desire it to be different from the current working directory. It is supplied as an option --dir to the seabreeze CLI.

$ seabreeze batch --help

Usage: seabreeze batch [OPTIONS] {analyse_genome_sizes|predict_IS_elements|pre
                       dict_structural_variants|predict_replichore_balance|pre
                       dict_SV_mechanism|annotate_SV_regions|run_all}
                       [SNAKEMAKE_ARGS]...

  Execute given workflow for batch processing

Options:
  -d, --dir PATH     Set working directory where output will be created.
                     Default is current working directory
  -b, --data PATH    Path to data.csv file. Default is data.csv in current
                     working dir
  --oridif PATH      Path to csv file with sequences for the origin and
                     terminus. Default is ori_dif_coords.csv in the current
                     working dir. Required for predict_replichore_balance or
                     run_all workflows
  --threads INTEGER  Resources to be used.  [default: 4]
  --masked           Mask insertion sequences
  -h, --help         Show this message and exit.