EMNGly: predicting N-linked glycosylation sites using the language models for feature extraction

Xiaoyang Hou, Yu Wang, Dongbo Bu, Yaojun Wang^∗ and Shiwei Sun^∗
Paper link on Bioinformatics

Environments

git clone git@github.com:StellaHxy/EMNgly.git
conda create -n EMNgly python=3.7
conda activate EMNgly
pip install -r requirements.txt

Datasets

• Download the N-GlycositeAltas dataset and N-GlyDE dataset under folder data/:

  ├── data
  │   ├── N-GlycositeAltas
  │   │     ├── train.csv
  │   │     ├── test.csv
  │   │     ├── test_features.csv   
  │   │     └── pdb_file
  │   ├── N-GlyDE
  │   │     ├── train.csv
  │   │     ├── test.csv
  │   │     └── pdb_file

Usage

Quick predict N-linked glycosylation sites of N-GlycositeAltas dataset:

Download the checkpoint of N-GlyAltas_classifier under folder checkpoints/:

├── checkpoints
│       └── N-GlyAltas_classifier.pkl
├── data
├── log
├── model
├── scripts
├── main.py
├── predict.py

python predict.py --mode=test_features --data_path=./data/N-GlycositeAltas --ckpt_path=./checkpoints/N-GlyAltas_classifier.pkl 

Train the classifier of EMNgly on N-GlycositeAltas dataset:

bash scritps/get_N-GlycositeAltas_train_features.sh
python main.py --mode=train --data_path=./data/N-GlycositeAltas --output_path=./checkpoints/N-GlyAltas_classifier.pkl

Predict N-linked glycosylation sites of N-GlycositeAltas dataset:

bash scritps/get_N-GlycositeAltas_test_features.sh
python predict.py --mode=test --data_path=./data/N-GlycositeAltas --ckpt_path=./checkpoints/N-GlyAltas_classifier.pkl