Repository for pydebiaseddta, Python implementation of the DebiasedDTA training framework for robust generalization in drug-target affinity prediction models [1]. The documentation of the library is available here.
The package can be installed using the \texttt{pip} package management system. Before installing the package, we highly recommend creation of a virtual environment with the appropriate Python version, using the Anaconda software framework:
conda create --name pydebiaseddta python=3.9.7
conda activate pydebiaseddta
python3 -m pip install pydebiaseddtaWe highly recommend the user to start their exploration with the Jupyter notebook file examples/quickstart.ipynb, which provides a practical yet comprehensive introduction to pydebiaseddta. You can view the contents of this notebook directly through this repository, or download it and run it on your machine.
To run this file, open it with an IDE that can read notebook files, such as VS Code or Jupyter Lab. In this program, you will have to select the interpreter as pydebiaseddta, corresponding to the name of the Anaconda environment you just installed. The cells of the notebook will onboard you to the main functionalities of the repository.
To see examples of a broader range of functionalities, check the notebook file examples/additional_use_cases.ipynb. This notebook illustrates uses of various additional experiment settings. This repository also includes the script scripts/pydebiaseddta_experiment.py that allows running experiments through command line with an appropriate config file.
The repository includes a small sample from the KIBA dataset [3], which can be loaded using the function pydebiaseddta.utils.load_sample_dta_data. This allows the user to experiment with pydebiaseddta's features without needing external data.
For users interested in experimenting with real datasets we make the full BDB [2] and KIBA [3] datasets available to download from the following link. The datasets have been separated into splits called train, test_warm, test_cold_prot, test_cold_lig, and test_cold_both. The various test splits serve to test algorithm performance in novel proteins, novel ligands, and both novel proteins and ligands. Validation counterparts of the test splits have also been provided for model selection. Five different random splits have been provided to allow the user to assess the effect of randomization on results.
If you use pydebiaseddta in your research, please consider citing:
@article{ozccelik2023debiaseddta,
title={DebiasedDTA: DebiasedDTA: Improving the Generalizability of Drug-Target Affinity
Prediction Models},
author={{\"O}z{\c{c}}elik, R{\i}za and Ba{\u{g}}, Alperen and At{\i}l, Berk and Barsbey, Melih and {\"O}zg{\"u}r, Arzucan and {\"O}zk{\i}r{\i}ml{\i}, Elif},
journal={arXiv preprint arXiv:2107.05556},
year={2023}
}[1] Özçelik, Rıza, Alperen Bağ, Berk Atıl, Melih Barsbey, Arzucan Özgür, and Elif Özkırımlı. “DebiasedDTA: A Framework for Improving the Generalizability of Drug-Target Affinity Prediction Models.” arXiv, January 8, 2023. http://arxiv.org/abs/2107.05556.
[2] Özçelik, Rıza, Hakime Öztürk, Arzucan Özgür, and Elif Ozkirimli. “ChemBoost: A Chemical Language Based Approach for Protein - Ligand Binding Affinity Prediction.” Molecular Informatics 40, no. 5 (May 2021): e2000212. https://doi.org/10.1002/minf.202000212.
[3] Tang, Jing, Agnieszka Szwajda, Sushil Shakyawar, Tao Xu, Petteri Hintsanen, Krister Wennerberg, and Tero Aittokallio. “Making Sense of Large-Scale Kinase Inhibitor Bioactivity Data Sets: A Comparative and Integrative Analysis.” Journal of Chemical Information and Modeling 54, no. 3 (March 24, 2014): 735–43. https://doi.org/10.1021/ci400709d.