GraphMHC

GraphMHC is a neoantigen prediction model that applyies graph neural networks to the molecular structure from amino acid sequences of MHC proteins and peptides. This model utilizes the MHC class I and MHC class II datasets from IEDB, and can also be used for other datasets on binding between amino acid sequences.

Requirements

• python=3.9
• pandas=1.4.4
• numpy=1.23.2
• rdkit=2022.03.2
• torch=1.11.0+cu113
• torch_geometric=2.1.0
• sklearn=1.1.2

Apply to IEDB class I data

• Download data:

  wget http://tools.iedb.org/static/main/binding_data_2013.zip
  unzip binding_data_2013.zip
  wget https://downloads.iedb.org/tools/mhci/3.1.2/IEDB_MHC_I-3.1.2.tar.gz
  tar xvzf IEDB_MHC_I-3.1.2.tar.gz

• Preprocessing:

  python preprocessing_i.py

• Apply GraphMHC model:

  python graphmhc.py --root iedb_i --train iedb_trainset.csv --test iedb_testset.csv --mhc mhc_sequence --peptide sequence --binding binding

Apply to IEDB class II data

• Download data:

  wget http://tools.iedb.org/static/download/classII_binding_data_Nov_16_2009.tar.gz
  tar xvzf classII_binding_data_Nov_16_2009.tar.gz
  wget https://downloads.iedb.org/tools/mhcii/3.1.6/IEDB_MHC_II-3.1.6.tar.gz
  tar xvzf IEDB_MHC_II-3.1.6.tar.gz

• Preprocessing:

  python preprocessing_ii.py

• Apply GraphMHC model:

  python graphmhc.py --root iedb_ii --train iedb_ii_train.csv --test iedb_ii_test.csv --mhc mhc --peptide peptide --binding binding

  Apply to the other dataset about binding between amino acid sequences

  python graphmhc.py --root <root directory> --train <trainset file> --test <testset file> --mhc <MHC field> --peptide <peptide field> --binding <binding affinity field>

Citation

https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0291223

@article{jeong2024graphmhc,
  title={GraphMHC: Neoantigen prediction model applying the graph neural network to molecular structure},
  author={Jeong, Hoyeon and Cho, Young-Rae and Gim, Jungsoo and Cha, Seung-Kuy and Kim, Maengsup and Kang, Dae Ryong},
  journal={Plos one},
  volume={19},
  number={3},
  pages={e0291223},
  year={2024},
  publisher={Public Library of Science San Francisco, CA USA}
}