Benchmark on Drug Target Interaction Modeling from a Structure Perspective

This is the official codebase of the paper Benchmark on Drug Target Interaction Modeling from a Structure Perspective.

GTB-DTI is a comprehensive benchmark customized for GNN and Transformer-based methodologies for DTI prediction.

Reproduction

Before you begin, you can install the required libraries using:

First, clone the repository to your local machine:

git clone https://github.com/justinwjl/GTB-DTI.git
cd your-repo

Traning the Model

python main.py -c config/model.yaml

The config.yaml file contains all the configurable parameters for training the model. You can edit this file to adjust parameters such as learning rate, batch size, and number of epochs.

Memory Evaluation

Set the 'train' to 'memory_test' in the config.yaml file

task:
  class: regression
  model:
    class: model
    param: 
  train: memory_test

Benchmark Result

Benchmark of Regression task

Benchmark of Classification task

License

This codebase is released under the MIT License as in the LICENSE file.

Citation

If you find this paper or code helpful in your research, please cite the following paper.

@misc{zhang2024benchmarkdrugtargetinteraction,
      title={Benchmark on Drug Target Interaction Modeling from a Structure Perspective}, 
      author={Xinnan Zhang and Jialin Wu and Junyi Xie and Tianlong Chen and Kaixiong Zhou},
      year={2024},
      eprint={2407.04055},
      archivePrefix={arXiv},
      primaryClass={q-bio.QM},
      url={https://arxiv.org/abs/2407.04055}, 
}