WISP enables the quantitative evaluation of explainability methods applied to machine learning models. Details of the workflow are described in the following papers:

• When Machine Learning Models Learn Chemistry I: Quantifying Explainability with Matched Molecular Pairs
• When Machine Learning Models Learn Chemistry II: Applying WISP to Real-World Examples

Try out WISP in your browser — no installation required:

Launch the WISP Web App

The web application provides an interactive interface to explore explainability methods and quantitatively evaluate their performance — no prior coding knowledge required.

• python=3.11.11
• scikit-learn==1.6.1
• chemprop==2.2.0
• torch==2.7.1
• mmpdb==2.1
• rdkit==2024.9.6
• shap==0.47.2

WISP is directly installable from GitHub:

pip install git+https://github.com/kerjans/ml-XAI.git

Or, if you want to clone the repository for development:

git clone https://github.com/kerjans/ml-XAI.git
cd ml-XAI
pip install -e .
git checkout code-paper-version-1

We have provided different notebooks to:

• Reproduce the publications
• See how WISP is being used and
• Explore how the atom attributor is used as a standalone tool.

In general, you can execute WISP by simply plugging in your dataset of interest:

wisp = WISP('working-dir-path',
     'dataset-path',
     'ID-column-name',
     'smiles-column-name', 
     'property-of-interest-column-name',
     model_available=None)

Please acknowledge this work by citing the following if you use the package:

When Machine Learning Models Learn Chemistry I: Quantifying Explainability with Matched Molecular Pairs:

@misc{jans2025wisp1,
      title={When Machine Learning Models Learn Chemistry I: Quantifying Explainability with Matched Molecular Pairs}, 
      author={Kerrin Janssen, Jan M. Wollschläger, Jonny Proppe and Andreas H. Göller},
      year={2025},
      archivePrefix={ChemRxiv},
      url={https://doi.org/10.26434/chemrxiv-2025-d57kp}, 
}

When Machine Learning Models Learn Chemistry II: Applying WISP to Real-World Examples:

@misc{jans2025wisp2,
      title={When Machine Learning Models Learn Chemistry II: Applying WISP to Real-World Examples}, 
      author={Kerrin Janssen, Jan M. Wollschläger, Jonny Proppe and Andreas H. Göller},
      year={2025},
      archivePrefix={ChemRxiv},
      url={https://doi.org/10.26434/chemrxiv-2025-zdqgr}, 
}