BioBridge

This is the official implementation of the paper BioBridge: Bridging Biomedical Foundation Models via Knowledge Graphs (ICLR 2024).

Main components

checkpoints/

The trained BioBridge model checkpoints.

data/

• data/PrimeKG: The raw PrimeKG data in .zip format.

• data/Processed: The node features obtained from different databases, e.g., protein's sequence.

• data/embeddings/: The KG node embeddings extracted from unimodal FMs, such as PubMedBERT, UniMol, and ESM-2.

• data/BindData/: The preprocessed BioBridge related data and its configurations.

• data/mouse_protein/: The preprocessed mouse protein data and its configurations regarding the mouse protein prediction task in the paper.

dataset/

The guidelines for data preprocessing.

src/

The source code of BioBridge.

unimodal/

The source code of unimodal FMs, including PubMedBERT and ESM-2, for encoding node features.

train.sh

The example script for training BioBridge.

eval.sh

The example script for using BioBridge for cross-modality prediction.

experiments/mouse_protein

The example script for training BioBridge on the mouse protein prediction task.

experiments/molecule_generation

The example script for testing BioBridge on prompting LLMs for molecule generation and Q&A tasks.

References

If you find this repository useful in your research, please cite the following paper:

@inproceedings{wang2023biobridge,
  title={BioBridge: Bridging Biomedical Foundation Models via Knowledge Graphs},
  author={Wang, Zifeng and Wang, Zichen and Srinivasan, Balasubramaniam and Ioannidis, Vassilis N and Rangwala, Huzefa and Anubhai, Rishita},
  booktitle={International Conference on Learning Representations},
  year={2024}
}