SAGEPhos: Sage Bio-Coupled and Augmented Fusion for Phosphorylation Site Detection

This repository contains the open-source implementation of the paper "SAGEPHOS: SAGE BIO-COUPLED AND AUGMENTED FUSION FOR PHOSPHORYLATION SITE DETECTION".

Table of Contents

• Introduction
• Installation
• Usage
  • Training
  • Testing
• Citation
• Contact

Introduction

SAGEPhos introduces a Bio-Coupled Modal Fusion method, distilling essential kinase sequence information to refine task-oriented local substrate feature space. It creates a shared semantic space that captures crucial kinase-substrate interaction patterns. Within the substrate’s intra-modality domain, it focuses on Bio-Augmented Fusion, emphasizing 2D local sequence information while selectively incorporating 3D spatial information from predicted structures to complement the sequence space.

Installation

1. Clone the repository:

   git clone https://github.com/ZhangJJ26/SAGEPhos.git
   cd SAGEPhos
   

2. Install dependencies:

   Option 1: Use the environment.yml file to set up the environment:

   conda env create -f environment.yml
   conda activate SAGEPhos
   

   Option 2: Manually install the environment:

   conda create -n SAGEPhos python=3.8
   conda activate SAGEPhos
   conda install torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia  # Use the appropriate CUDA version for your system.
   conda install torchdrug -c milagraph -c conda-forge -c pytorch -c pyg
   conda install easydict pyyaml -c conda-forge
   conda install scipy joblib wandb
   pip install git+https://github.com/facebookresearch/esm.git
   pip install ninja rdkit-pypi scikit-learn h5py atom3d
   

3. Download the pretrained model:

   • Download the pretrained model from the releases page.
   • Extract and place it in the checkpoint directory:

     mkdir checkpoint
     tar -xzvf checkpoint.tar.gz -C checkpoint/
     

Usage

Training

By default, we use 2 NVIDIA A40 GPUs for training. Adjust the batch size according to your GPU memory.

python script/downstream.py -c config/phos/esm_gearnet_parallel.yaml --ckpt null

• --ckpt: Specify the path to the model checkpoint. For training, set this to null.
• We have uploaded our SAGEPhos dataset on releases page.

Testing

To test the model, run the following command:

python script/downstream.py -c config/phos/esm_gearnet_parallel.yaml --ckpt checkpoint/checkpoint.pth

• --ckpt: Path to the model checkpoint (default is checkpoint/checkpoint.pth).

Citation

If you use SAGEPhos in your research, please cite our paper:

@inproceedings{zhang2025sagephos,
  title={SAGEPhos: Sage Bio-Coupled and Augmented Fusion for Phosphorylation Site Detection},
  author={Zhang, Jingjie and Hanqun, CAO and Gao, Zijun and Wang, Xiaorui and Gu, Chunbin},
  booktitle={The Thirteenth International Conference on Learning Representations}
}

Contact

If you have any questions, please feel free to contact the authors.

• Jingjie Zhang (1155224008@link.cuhk.edu.hk)