PreDDG

📦 Environment Setup

We recommend creating a dedicated conda environment:

  conda create -n PreDDG python=3.12
  conda activate PreDDG
  pip install numpy pandas scipy scikit-learn pathlib tqdm
  pip install torch==2.5.0 torchvision==0.20.0 torchaudio==2.5.0 --index-url https://download.pytorch.org/whl/cu118
  pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.5.0+cu118.html
  pip install tensorboard tensorboardX pytorch_lightning 
  pip install torch_geometric fair-esm
  pip install biopython

⚠️ Please ensure that the CUDA version matches your PyTorch and torch-geometric installation. For details, refer to the installation guide。

📂 Data Preparation

Datasets

Dataset	Download Link
cDNA	https://github.com/jozhang97/MutateEverything/
cDNA2	https://github.com/jozhang97/MutateEverything/
PTMul-NR	https://ddgemb.biocomp.unibo.it/datasets/
M28	https://github.com/GenScript-IBDPE/UniMutStab/tree/main/Dataset/Independent/multiple
M38	https://github.com/GenScript-IBDPE/UniMutStab/tree/main/Dataset/Independent/multiple

Datasets should be placed under ./data/dataset/ directory. The folder structure should be as follows:

data/
    dataset/
        M28/
            mutations/
                M28.csv

ISM Model Preparation

Download ISM-650M-UC30PDB，and place it in ./data/ism/ism_t33_650M_uc30pdb/ directory:

data/
    ism/
        ism_t33_650M_uc30pdb/
            config.json
            gitattributes
            ism_t33_650M_uc30pdb.pth
            model.safetensors
            special_tokens_map.json
            tokenizer_config.json
            vocab.txt

🚀 Running PreDDG for Prediction

Example: predicting on M28 dataset. Input files should be in .csv format with one of the following formats:

Format 1 (with both wild-type and mutant sequences):

pdb_id	wt_seq	mut_info	mut_seq

Format 2 (only mutation info provided):

pdb_id	wt_seq	mut_info

Note:

1. mut_info follows the format WT_POS_MUT, e.g., Y68R means the 68th position changes from Y to R.
2. Multiple mutations are separated by :, e.g., Y68R:A120V.
3. mut_seq is optional. If not provided, it will be computed based on wt_seq and mut_info.

cd PreDDG
python predict.py --test_name='M28' --device='cuda:0'

Predictions are saved under ./data/dataset/M28/predictions/. Example output:

pdb_id	wt_seq	mut_info	mut_seq	preddg

For more details, please refer to the paper and source code.

📖 Citation

If you find PreDDG useful, please cite our paper:

@article{
  title={},
  author={},
  journal={},
  year={}
}