Code, intermediate results and an interactive visualisation on prediction of putative novel enzymes and small molecule binding proteins presented in (Barrio-Hernandez et al. 2023).
- Predicted pockets with score > 60 and mean pLDDT > 90
- Pocket surfaces, one file per pocket
- Predicted GO/EC terms for all structures
- Residue-level saliency weights for MF/EC terms, one file per structure
A publicly accessible instance is currently accessible at this address on the Streamlit Community Cloud. Alternatively, install the web app locally by cloning the repository and setting up a conda/mamba environment as follows:
$ git clone git@github.com:jurgjn/af-protein-universe.git
$ cd af-protein-universe/
$ conda create -p streamlit-env python numpy matplotlib seaborn 'pandas<2.0.0'
$ conda run -p ./streamlit-env pip install -r requirements.txt
After that, run the web app locally with:
$ conda run -p ./streamlit-env streamlit run app.py