This repository is a part of the EU-funded RECeSS project (#101102016), and hosts the code for the open-source Python package stanscofi for the development of collaborative filtering-based drug repurposing algorithms.
As of 2022, current drug development pipelines last around 10 years, costing $2billion in average, while drug commercialization failure rates go up to 90%. These issues can be mitigated by drug repurposing, where chemical compounds are screened for new therapeutic indications in a systematic fashion. In prior works, this approach has been implemented through collaborative filtering. This semi-supervised learning framework leverages known drug-disease matchings in order to recommend new ones.
The stanscofi package comprises method-agnostic training, validation, preprocessing and visualization procedures on several published drug repurposing datasets. The proper implementation of these steps is crucial in order to avoid data leakage, i.e., the model is learnt over information that should be unavailable at prediction time. Indeed, data leakage is the source of a major reproducibility crisis in machine learning. This will be avoided by building training and validation sets which are weakly correlated with respect to the drug and disease feature vectors. The main performance metric will be the area under the curve (AUC), which estimates the diagnostic ability of a recommender system better than accuracy in imbalanced datasets.
Medium-term outcomes to this package will significantly alleviate the economic burden of drug discovery pipelines, and will help find treatments in a more sustainable manner, especially for rare or tropical neglected diseases.
For more information about the datasets accessible in stanscofi, please refer to the following repository.
Run one of the following commands:
## Using pip: install in Python env
pip install stanscofi
## Using Anaconda: install in Python env
conda install -c recess stanscofi
## Using the Docker image: will open a container
docker push recessproject/stanscofi:2.0.0Documentation about stanscofi can be found at this page. The complete list of dependencies for stanscofi can be found at requirements.txt (pip) or meta.yaml (conda).
This repository is under an OSI-approved MIT license.
If you use stanscofi in academic research, please cite it as follows
Réda et al., (2024). stanscofi and benchscofi: a new standard for drug repurposing by collaborative filtering.
Journal of Open Source Software, 9(93), 5973, https://doi.org/10.21105/joss.05973
Pull requests and issue flagging are welcome, and can be made through the GitHub interface. Support can be provided by reaching out to recess-project[at]proton.me. However, please note that contributors and users must abide by the Code of Conduct.