README.md

---
license: cc-by-4.0
pipeline_tag: graph-ml
tags:
- biology
- chemistry
---

## Data and Model Weights

All datasets and pre-trained model weights are available on Hugging Face:

🤗 **[https://huggingface.co/wenda8759/AbFlow](https://huggingface.co/wenda8759/AbFlow)**

### Download Options

#### Option 1: Using huggingface-cli (Recommended)

```bash
# Install huggingface_hub if not already installed
pip install huggingface_hub

# Download all files
huggingface-cli download wenda8759/AbFlow --local-dir ./

# Or download specific files
huggingface-cli download wenda8759/AbFlow checkpoints/multi_cdr_design.ckpt --local-dir ./
```

#### Option 2: Using Python

```python
from huggingface_hub import snapshot_download, hf_hub_download

# Download entire repository
snapshot_download(repo_id="wenda8759/AbFlow", local_dir="./")

# Or download specific file
hf_hub_download(
    repo_id="wenda8759/AbFlow",
    filename="checkpoints/multi_cdr_design.ckpt",
    local_dir="./"
)
```


### Model Checkpoints

| Model | Description | File |
|-------|-------------|------|
| Paratope-Centric Design | Design based on epitope | `checkpoints/paratope_centric_design.ckpt` |
| Multi-CDR Design | Design all 6 CDR regions | `checkpoints/multi_cdr_design.ckpt` |
| Structure Prediction | Predict antibody structure | `checkpoints/structure_prediction.ckpt` |
| Affinity Optimization | Optimize binding affinity | `checkpoints/affinity_optimization.ckpt` |
| ΔΔG Predictor | Predict binding energy changes | `checkpoints/ddg_predictor.ckpt` |

### Dataset Structure

After downloading, organize your data as follows:

```
AbFlow/
├── datasets/
│   ├── RAbD/
│   │   ├── train.json
│   │   ├── valid.json
│   │   ├── test.json
│   │   ├── train.pkl
│   │   ├── valid.pkl
│   │   ├── test.pkl
│   │   ├── train_surf.pkl
│   │   ├── valid_surf.pkl
│   │   └── test_surf.pkl
│   └── IgFold/
│       ├── train.json
│       ├── valid.json
│       ├── test.json
│       └── ...
└── checkpoints/
    ├── multi_cdr_design.ckpt
    ├── structure_prediction.ckpt
    └── ...
```