Implementation of the KNIME retropath2.0 workflow. Takes for input the minimal (dmin) and maximal (dmax) diameter for the reaction rules and the maximal path length (maxSteps). The tool expects the following files: rules.csv, sink.csv and source.csv and produces results in an output folder.

• Python 3
• KNIME (code was tested on 4.5.0 version)

The conda package manager is required. Fresh instructions on how to install conda are available online.

On Linux system, the tool tries to install the KNIME Anlytical Platform as well as the RetroPath2.0 node dependancies.

On MacOS, the user needs to first install by himself KNIME, and be sure that the RetroPath2.0 node dependancies are installed. To install node dependancies, we recommand to follow the RetroPath2.0 instructions from https://www.myexperiment.org/workflows/4987.html.

Install in the <my_env> conda environment:

conda install -c conda-forge -n <my_env> retropath2_wrapper 

Disclaimer: we recommand to provide absolute path to files, problems can arise with relative paths.

python -m retropath2_wrapper <sink-file> <rules-file> <out-dir> --source_file <source-file>

python -m retropath2_wrapper <sink-file> <rules-file> <out-dir> --source_file <source-file> --kexec <path-to-knime-exec>

The minimal required arguments are sink_file, source_file, rules_file and outdir.

from retropath2_wrapper import retropath2

r_code = retropath2(
    sink_file='/path/to/sink/file',
    source_file='/path/to/source/file',
    rules_file='/path/to/rules/file',
    outdir='/path/to/outdir'
    )

Exploration settings have default values and can be tuned:

from retropath2_wrapper import retropath2

r_code = retropath2(
    sink_file='/path/to/sink/file',
    source_file='/path/to/source/file',
    rules_file='/path/to/rules/file',
    outdir='/path/to/outdir',
    max_steps=3,
    topx=100,
    dmin=0,
    dmax=100,
    mwmax_source=1000,
    mwmax_cof=1000,
    )

For convenience, the retropath2_wrapper make an attempt to install the KNIME application, and the needed KNIME dependancies. KNIME installation can be skipped by setting kexec to the KNIME executable path, and KNIME dependancies can be skipped using kpkg_install=False.

Already installed KNIME app can hence be used that way, eg:

from retropath2_wrapper import retropath2

r_code = retropath2(
    sink_file='/path/to/sink/file',
    source_file='/path/to/source/file',
    rules_file='/path/to/rules/file',
    outdir='/path/to/outdir'
    kexec='/Applications/KNIME 4.3.0.app/Contents/MacOS/knime',
    kpkg_install=False
    )

Executions can be timed out using the timeout arguments (in minutes).

retropath2() function returns one of the following codes:

• 0: No error
• 1: Source has been found in the sink
• 2: Cannot find source-in-sink file
• 3: Running the RetroPath2.0 Knime program produced an OSError
• 4: RetroPath2.0 has found no solution
• 5: Time limit reached

Test can be run with the following commands:

conda install -c conda-forge pytest
python -m pytest tests

For further tests and development tools, a CI toolkit is provided in cicd-toolkit folder (see cicd-toolkit/README.md).

Please cite:

Delépine B, Duigou T, Carbonell P, Faulon JL. RetroPath2.0: A retrosynthesis workflow for metabolic engineers. Metabolic Engineering, 45: 158-170, 2018. DOI: https://doi.org/10.1016/j.ymben.2017.12.002