DegradeMaster

Source Code of ISMB/ECCB'25 submitted paper "Accurate PROTAC targeted degradation prediction with DegradeMaster".

Dependencies

• python==3.10.13
• pyg=2.5.2
• pytorch-cuda=11.6
• torch==2.4.0
• torch-cluster==1.6.3+pt23cu118
• torch-scatter==2.1.2+pt23cu118
• torch-sparse==0.6.18+pt23cu118
• torch-spline-conv==1.2.2+pt23cu118
• torchaudio==2.3.0+cu118
• networkx==3.3
• numpy==1.23.5
• scipy==1.10.1

The full dependencies can be installed by executing the command below:

conda env create --name envname --file=protac.yml

Usage

Accurate PROTAC-targeted degradation prediction on datasets crafted from PROTAC-DB 3.0

To train and evaluate on PROTAC-8K:

1. Download the dataset from https://zenodo.org/records/14728925, and paste folders at ./data/PROTAC
2. Execute the command below:

python main.py --config config/config.yml

Case study

To conduct the case study #1 for VZ185 candidate degradation prediction:

python case_study.py

To conduct the case study #2 for ACBI3 on KRAS mutant degradation prediction:

1. Remove all the files in ./data/case_study/processed

cd ./data/case_study/processed
rm *.pt

2. Change the value of "dataset_type" in ./config/config_c.yml to "case_study_2"

3. Execute the command below:

python case_study.py

Reference

@article{liu2025accurate,
  title={Accurate PROTAC targeted degradation prediction with DegradeMaster},
  author={Liu, Jie and Roy, Michael and Isbel, Luke and Li, Fuyi},
  journal={bioRxiv},
  pages={2025--02},
  year={2025},
  publisher={Cold Spring Harbor Laboratory}
}