This is the code implementation for SCOP.

Dataset

Pre-training dataset

• AlphaFold Swiss-Prot: A protein structure database predicted by the AlphaFold model, containing approximately 542K protein 3D structure information. You can download this dataset clicking the website.

Benchmark datasets

1. EC: A public dataset used to predict the protein's enzyme commission (EC) numbers, which describe their catalysis of biochemical reactions. You can download this dataset clicking the website.
2. GO-BP: A dataset related to the biological process (BP) terms of a protein, representing a specific objective that the organism is genetically programmed to achieve. You can download this dataset clicking the website.
3. GO-MF: A dataset about the protein's molecular functions (MF), which correspond to activities that can be performed by individual gene products. You can download this dataset clicking the website
4. GO-CC: A dataset with the cellular component (CC) terms of a protein, referring to the locations about cellular structures where a gene product performs a function. You can download this dataset clicking the website
5. Glycoprotein Dataset (self-built dataset used in Case Study): You can find this dataset in the folder datasets

Requirements

In order to reproduce our code, you need to install the following pip dependencies:

numpy==1.22.4
pandas==1.4.3
rdkit-pypi==2022.3.5
torch==1.12.0
torch-cluster==1.6.0
torch-geometric==2.0.4
torch-scatter==2.0.9
torch-sparse==0.6.14
torch-spline-conv==1.2.1
torchdrug==0.2.0

Usage

1. Preprocess the pre-training dataset
   • After downloading the full AlphaFold Swiss-Prot dataset, you are supposed to leverage the src\alphafold.py to transform the original protein PDB file into a NetworkX file.
2. Pretrain the protein structure model and sequence model.
   • Run the file src\pretrain_model.py to pretrain two models simultaneously.
3. Evaluate the pretrained structure model on benchmark datasets.
   • Transform the downloaded dataset to the NetworkX format.
   • Run the file src\pipeline.py to evaluate the model on downstream tasks.

Cite

If you feel this work helpful for your research, please cite the following papers:

@inproceedings{ma2024scop,
  author={Ma, Runze and He, Chengxin and Zheng, Huiru and Wang, Xinye and Wang, Haiying and Zhang, Yidan and Duan, Lei},
  booktitle={2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)}, 
  title={SCOP: A Sequence-Structure Contrast-Aware Framework for Protein Function Prediction}, 
  year={2024},
  pages={79-84},
  doi={10.1109/BIBM62325.2024.10822541}
}