All publications

LightMHC: A Light Model for pMHC Structure Prediction with Graph Neural Networks

Antoine P. Delaunay, Yunguan Fu, Nikolai Gorbushin, Robert McHardy, Bachir A. Djermani, Liviu Copoiu, Michael Rooney, Maren Lang, Andrey Tovchigrechko, Uğur Şahin, Karim Beguir, Nicolas Lopez Carranza

In NeurIPS MLSB 2023 · 2023

DOI

Abstract

The peptide-major histocompatibility complex (pMHC) is a crucial protein in cell-mediated immune recognition and response. Accurate structure prediction is potentially beneficial for protein interaction prediction and therefore helps immunotherapy design. However, predicting these structures is challenging due to the sequential and structural variability. In addition, existing pre-trained models such as AlphaFold 2 require expensive computation thus inhibiting high throughput in silico peptide screening. In this study, we propose LightMHC: a lightweight model (2.2M parameters) equipped with attention mechanisms, graph neural networks, and convolutional neural networks. LightMHC predicts full-atom pMHC structures from amino-acid sequences alone, without template structures. The model achieved comparable or superior performance to AlphaFold 2 and ESMFold (93M and 15B parameters respectively), with five-fold acceleration (6.65 seconds/sample for LightMHC versus 36.82 seconds/sample for AlphaFold 2), potentially offering a valuable tool for immune protein structure prediction and immunotherapy design.

Cite

@inproceedings{delaunay-etal-2023-lightmhc,
    title = "LightMHC: A Light Model for pMHC Structure Prediction with Graph Neural Networks",
    author = "Delaunay, Antoine P.  and
      Fu, Yunguan  and
      Gorbushin, Nikolai and
      McHardy, Robert and
      Djermani, Bachir A. and
      Copoiu, Liviu and
      Rooney, Michael and
      Lang, Maren and
      Tovchigrechko, Andrey and
      Şahin, Uğur and
      Beguir, Karim and
      Lopez Carranza, Nicolas
      ",
    booktitle = "Machine Learning in Structural Biology Workshop at the 37th Conference on Neural Information Processing Systems (NeurIPS)",
    month = dec,
    year = "2023",
    address = "New Orleans, Louisiana",
    publisher = "",
    url = "https://www.mlsb.io/",
    doi = "10.1101/2023.11.21.568015",
    pages = "",
}