Skip to main content

(DRExM³L) Drug REpurposing using and eXplainable Machine Learning and Mechanistic Models of signal transduction"

Project description

DOI DOI PyPI version pdm-managed

Drug REpurposing using eXplainable Machine Learning and Mechanistic Models of signal transduction

Repository for the drexml python package: (DRExM³L) Drug REpurposing using eXplainable Machine Learning and Mechanistic Models of signal transduction

Citation

Find the associated publication here:

Esteban-Medina M, de la Oliva Roque VM, Herráiz-Gil S, Peña-Chilet M, Dopazo J, Loucera C. drexml: A command line tool and Python package for drug repurposing. Computational and Structural Biotechnology Journal 2024;23:1129–43. https://doi.org/10.1016/j.csbj.2024.02.027.

Part of the Intelligent Biology and Medicine special issue:

https://www.sciencedirect.com/journal/computational-and-structural-biotechnology-journal/special-issue/10XRHM1G1LS

And the BIB file:

@article{EstebanMedina2024,
  title = {drexml: A command line tool and Python package for drug repurposing},
  volume = {23},
  ISSN = {2001-0370},
  url = {http://dx.doi.org/10.1016/j.csbj.2024.02.027},
  DOI = {10.1016/j.csbj.2024.02.027},
  journal = {Computational and Structural Biotechnology Journal},
  publisher = {Elsevier BV},
  author = {Esteban-Medina,  Marina and de la Oliva Roque,  Víctor Manuel and Herráiz-Gil,  Sara and Peña-Chilet,  María and Dopazo,  Joaquín and Loucera,  Carlos},
  year = {2024},
  month = dec,
  pages = {1129–1143}
}

The article was written using drexml version v1.1.0. Install it using:

pip install drexml==1.1.0

Version v1.1.1 improves the documentation and README by including a reference to the published article for easier access.

Setup

To install the drexml package use the following:

conda create -n drexml python=3.10
conda activate drexml
pip install drexml

If a CUDA~10.2/11.x (< 12) compatible device is available use:

conda create -n drexml --override-channels -c "nvidia/label/cuda-11.8.0" -c conda-forge cuda cuda-nvcc cuda-toolkit gxx=11.2 python=3.10
conda activate drexml
pip install --no-cache-dir --no-binary=shap drexml

To install drexml in an existing environment, activate it and use:

pip install drexml

Note that by default the setup will try to compile the CUDA modules, if not possible it will use the CPU modules.

Run

To run the program for a disease map that uses circuits from the preprocessed KEGG pathways and the KDT standard list, construct an environment file (e.g. disease.env):

  • using the following template if you have a set of seed genes (comma-separated):
seed_genes=2175,2176,2189
  • using the following template if you want to use the DisGeNET [1] curated gene-disease associations as seeds.
disease_id="C0015625"
  • using the following template if you know which circuits to include (the disease map):
circuits=circuits.tsv.gz

The TSV file circuits.tsv has the following format (tab delimited):

index	in_disease
P-hsa03320-37	0
P-hsa03320-61	0
P-hsa03320-46	0
P-hsa03320-57	0
P-hsa03320-64	0
P-hsa03320-47	0
P-hsa03320-65	0
P-hsa03320-55	0
P-hsa03320-56	0
P-hsa03320-33	0
P-hsa03320-58	0
P-hsa03320-59	0
P-hsa03320-63	0
P-hsa03320-44	0
P-hsa03320-36	0
P-hsa03320-30	0
P-hsa03320-28	1

where:

  • index: Hipathia circuit id
  • in_disease: (boolean) True/1 if a given circuit is part of the disease

Note that in all cases you can restrict the circuits to the physiological list by setting use_physio=true in the env file.

To run the experiment using 10 CPU cores and 0 GPUs, run the following command within an activated environment:

drexml run --n-gpus 0 --n-cpus 10 $DISEASE_PATH

where:

  • --n-gpus indicates the number of gpu devices to use in parallel (-1 -> all) (0 -> None)
  • --n-cpus indicates the number of cpu devices to use in parallel (-1 -> all) 8
  • DISEASE_PATH indicates the path to the disease env file (e.g. /path/to/disease/folder/disease.env)

Use the --debug option for testing that everything works using a few iterations.

Note that the first time that the full program is run, it will take longer as it downloads the latest versions of each background dataset from Zenodo:

https://doi.org/10.5281/zenodo.6020480

Contribute to development

The recommended setup is:

  • setup pipx
  • setup miniforge
  • use pipx to install pdm
  • ensure that pdm is version >=2.1, otherwise update with pipx
  • use pipx to inject pdm-bump into pdm
  • use pipx to install nox
  • run pdm config venv.backend conda
  • run make, if you want to use a CUDA enabled GPU, use make gpu=1
  • (Recommended): For GPU development, clear the cache using pdm cache clear first

Documentation

The documentation can be found here:

https://loucerac.github.io/drexml/

References

[1] Janet Piñero, Juan Manuel Ramírez-Anguita, Josep Saüch-Pitarch, Francesco Ronzano, Emilio Centeno, Ferran Sanz, Laura I Furlong. The DisGeNET knowledge platform for disease genomics: 2019 update. Nucl. Acids Res. (2019) doi:10.1093/nar/gkz1021

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

drexml-1.1.3.tar.gz (2.1 MB view details)

Uploaded Source

Built Distribution

drexml-1.1.3-py3-none-any.whl (2.0 MB view details)

Uploaded Python 3

File details

Details for the file drexml-1.1.3.tar.gz.

File metadata

  • Download URL: drexml-1.1.3.tar.gz
  • Upload date:
  • Size: 2.1 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/5.1.0 CPython/3.12.5

File hashes

Hashes for drexml-1.1.3.tar.gz
Algorithm Hash digest
SHA256 c99bdf6f0679b9f984a177aaf38cc306b072994d733335269b06c954c499734c
MD5 a5687ae3183b9834f93d6ed31e703dcb
BLAKE2b-256 03a2cb25efa198a2d746c548d5bc7eaffc49be8e15344bab72d49e36b633edd0

See more details on using hashes here.

File details

Details for the file drexml-1.1.3-py3-none-any.whl.

File metadata

  • Download URL: drexml-1.1.3-py3-none-any.whl
  • Upload date:
  • Size: 2.0 MB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/5.1.0 CPython/3.12.5

File hashes

Hashes for drexml-1.1.3-py3-none-any.whl
Algorithm Hash digest
SHA256 7bb7d6eed946de9c7726740ce51e93cac01e0800006309a8e9899955a73bacd5
MD5 6302b398fcb3776e5df0f46ad83e0439
BLAKE2b-256 f5da315fb0cd3221d5212bdc7dd7ba7b8a1aab8e8252b8fc6a7ff6b5250535c9

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page