https://github.com/d-f/binding-affinity
Protein-Ligand binding affinity prediction with Protein LLMs and Graph Attention Networks.
https://github.com/d-f/binding-affinity
binding-affinity binding-affinity-prediction graph-attention-networks pdb-dataset protein-llm pytorch
Last synced: about 1 year ago
JSON representation
Protein-Ligand binding affinity prediction with Protein LLMs and Graph Attention Networks.
- Host: GitHub
- URL: https://github.com/d-f/binding-affinity
- Owner: d-f
- Created: 2025-03-20T17:33:38.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-03-20T17:43:54.000Z (about 1 year ago)
- Last Synced: 2025-03-20T18:34:48.602Z (about 1 year ago)
- Topics: binding-affinity, binding-affinity-prediction, graph-attention-networks, pdb-dataset, protein-llm, pytorch
- Language: Python
- Homepage:
- Size: 9.77 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
```process_dataset.py``` processes the PDBBind dataset through the deepchem library. It will embed the sequence of amino acid abbreviations with a Protein LLM, convert the rdkit Molecule objects into torch_geometric graphs and saves the binding affinity each in separate directories.
```determine_bond_types.py``` determines which bond types are present within the dataset for nomalization purposes
```train_models.py``` allows for training a pure GAT or a combination of a GAT and Transformer depending on what is set for the model_type parameter. If a pure GAT is used, protein LLM embeddings are concatenated to atomic features when creating a ligand molecule graph and the graph is used for whole graph regression. If a combination of transformer and GAT is used, the GAT will be used to embed the graph, and a transformer will predict the binding affinity between the protein embedding and embedded ligand graph.