Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2019 May 28;20(1):270.
doi: 10.1186/s12859-019-2892-4.

MHCSeqNet: a deep neural network model for universal MHC binding prediction

Poomarin Phloyphisut  1 , Natapol Pornputtapong  2   3 , Sira Sriswasdi  4   5 , Ekapol Chuangsuwanich  6   7   8
Affiliations

MHCSeqNet: a deep neural network model for universal MHC binding prediction

Poomarin Phloyphisut et al. BMC Bioinformatics. .

Abstract

Background: Immunotherapy is an emerging approach in cancer treatment that activates the host immune system to destroy cancer cells expressing unique peptide signatures (neoepitopes). Administrations of cancer-specific neoepitopes in the form of synthetic peptide vaccine have been proven effective in both mouse models and human patients. Because only a tiny fraction of cancer-specific neoepitopes actually elicits immune response, selection of potent, immunogenic neoepitopes remains a challenging step in cancer vaccine development. A basic approach for immunogenicity prediction is based on the premise that effective neoepitope should bind with the Major Histocompatibility Complex (MHC) with high affinity.

Results: In this study, we developed MHCSeqNet, an open-source deep learning model, which not only outperforms state-of-the-art predictors on both MHC binding affinity and MHC ligand peptidome datasets but also exhibits promising generalization to unseen MHC class I alleles. MHCSeqNet employed neural network architectures developed for natural language processing to model amino acid sequence representations of MHC allele and epitope peptide as sentences with amino acids as individual words. This consideration allows MHCSeqNet to accept new MHC alleles as well as peptides of any length.

Conclusions: The improved performance and the flexibility offered by MHCSeqNet should make it a valuable tool for screening effective neoepitopes in cancer vaccine development.

Keywords: Deep learning; MHC epitope prediction; Recurrent neural networks.

PubMed Disclaimer

Conflict of interest statement

The authors declare that they have no competing interests.

Figures

Fig. 1
Fig. 1
An overview of the MHCSeqNet’s architecture. The model is comprised of three main parts: the peptide sequence processing part (a & c), the MHC processing part (b & d), and the main processing part which accepts the processed information from the previous parts (e). The entire model is a single deep learning model which can be trained altogether. f Our models output binding probability for the given peptide and MHC allele on the scale of 0 to 1, with 1 indicating likely ligand
Fig. 2
Fig. 2
MHCSeqNet achieves the best AUC and F1 scores on MHC class I binding dataset. a Bar plots showing the AUC value of each tool when evaluated on the set of MHC alleles it supports (Supported Type) or on the set of MHC alleles supported by all tools (Common Type). b Similar bar plots showing F1 values. c The ROC plot for all tools when evaluated on the set of MHC alleles supported by all tools. Vertical black line indicates the 5% false discovery rate (FDR). Inset shows the zoomed in ROC plot for the region with ≤5% FDR. d Similar ROC plot for the evaluation on MHC alleles supported by individual tools
Fig. 3
Fig. 3
MHCSeqNet achieves the best AUC and F1 scores on MHC class I ligand peptidome dataset. a The ROC plot for all tools. Vertical black line indicates the 5% FDR. Inset show the zoomed in ROC plot for the region with ≤5% FDR. b Bar plots showing the AUC (bars with solid face colors) and F1 (bars with stripes) scores of each tool
See this image and copyright information in PMC

References

    1. Castle J. C., Kreiter S., Diekmann J., Lower M., van de Roemer N., de Graaf J., Selmi A., Diken M., Boegel S., Paret C., Koslowski M., Kuhn A. N., Britten C. M., Huber C., Tureci O., Sahin U. Exploiting the Mutanome for Tumor Vaccination. Cancer Research. 2012;72(5):1081–1091. doi: 10.1158/0008-5472.CAN-11-3722. - DOI - PubMed
    1. Schumacher TN, Schreiber RD. Realising the Promise: Neoantigens in cancer immunotherapy. Sci Mag. 2015;348(6230):69–74. - PubMed
    1. Banchereau Jacques, Palucka Karolina. Immunotherapy: Cancer vaccines on the move. Nature Reviews Clinical Oncology. 2017;15(1):9–10. doi: 10.1038/nrclinonc.2017.149. - DOI - PubMed
    1. Sahin Ugur, Türeci Özlem. Personalized vaccines for cancer immunotherapy. Science. 2018;359(6382):1355–1360. doi: 10.1126/science.aar7112. - DOI - PubMed
    1. Engels Boris, Engelhard Victor H., Sidney John, Sette Alessandro, Binder David C., Liu Rebecca B., Kranz David M., Meredith Stephen C., Rowley Donald A., Schreiber Hans. Relapse or Eradication of Cancer Is Predicted by Peptide-Major Histocompatibility Complex Affinity. Cancer Cell. 2013;23(4):516–526. doi: 10.1016/j.ccr.2013.03.018. - DOI - PMC - PubMed