About

DEER (Dense Enzyme Retrieval) provides a method for finding functionally related human-bacteria isozymes using learned dense vector representations (embeddings). This huggingface repository contains an example dataset for applying the DEER model for isozyme retrieval.

The working example dataset contains 5,849 enzyme structures in PDB format, including:

1,636 eukaryota enzymes as templates
4,213 bacteria enzymes as the database

The full codes and tutorial for using this dataset are available at GitHub: https://github.com/WangJiuming/deer/tree/main.

Citation

If you use the model in your research, please cite our paper with the following.

@misc{liu2025Exploring,
  author={Liu, H. and Shen, J. and others},
  title={Exploring Functional Insights into the Human Gut Microbiome via the Structural Proteome},
  year={2025},
  note={Manuscript under revision}
}