ACME: pan-specific peptide-MHC class I binding prediction through attention-based deep neural networks

MOTIVATION Prediction of peptide binding to the major histocompatibility complex (MHC) plays a vital role in the development of therapeutic vaccines for the treatment of cancer. Algorithms with improved correlations between predicted and actual binding affinities are needed to increase precision and reduce the number of false positive predictions. RESULTS We present ACME (Attention-based Convolutional neural networks for MHC Epitope binding prediction), a new pan-specific algorithm to accurately predict the binding affinities between peptides and MHC class I molecules, even for those new alleles that are not seen in the training data. Extensive tests have demonstrated that ACME can significantly outperform other state-of-the-art prediction methods with an increase of the Pearson correlation coefficient between predicted and measured binding affinities by up to 23 percentage points. In addition, its ability to identify strong-binding peptides has been experimentally validated. Moreover, by integrating the convolutional neural network with attention mechanism, ACME is able to extract interpretable patterns that can provide useful and detailed insights into the binding preferences between peptides and their MHC partners. All these results have demonstrated that ACME can provide a powerful and practically useful tool for the studies of peptide-MHC class I interactions. AVAILABILITY ACME is available as an open source software at https://github.com/HYsxe/ACME. SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

[1]  James McCluskey,et al.  A Naturally Selected Dimorphism within the HLA-B44 Supertype Alters Class I Structure, Peptide Repertoire, and T Cell Recognition , 2003, The Journal of experimental medicine.

[2]  Hiroaki Tanaka,et al.  Multipeptide immune response to cancer vaccine IMA901 after single-dose cyclophosphamide associates with longer patient survival , 2012, Nature Medicine.

[3]  Morten Nielsen,et al.  Automated benchmarking of peptide-MHC class I binding predictions , 2015, Bioinform..

[4]  Charles H. Yoon,et al.  An immunogenic personal neoantigen vaccine for patients with melanoma , 2017, Nature.

[5]  O. Lund,et al.  NetMHCpan, a method for MHC class I binding prediction beyond humans , 2008, Immunogenetics.

[6]  Jun Liu,et al.  Structural basis for the differential classification of HLA-A*6802 and HLA-A*6801 into the A2 and A3 supertypes , 2013, Molecular Immunology.

[7]  Morten Nielsen,et al.  NetMHCcons: a consensus method for the major histocompatibility complex class I predictions , 2011, Immunogenetics.

[8]  Catherine J. Wu,et al.  Towards personalized, tumour-specific, therapeutic vaccines for cancer , 2017, Nature Reviews Immunology.

[9]  V. Engelhard,et al.  Structure of peptides associated with class I and class II MHC molecules. , 1994, Annual review of immunology.

[10]  Yoshua Bengio,et al.  Neural Machine Translation by Jointly Learning to Align and Translate , 2014, ICLR.

[11]  D. Madden.  The three-dimensional structure of peptide-MHC complexes. , 1995, Annual review of immunology.

[12]  Jianyang Zeng,et al.  Analysis of Ribosome Stalling and Translation Elongation Dynamics by Deep Learning. , 2017, Cell systems.

[13]  M. Nielsen,et al.  NetMHCpan-4.0: Improved Peptide–MHC Class I Interaction Predictions Integrating Eluted Ligand and Peptide Binding Affinity Data , 2017, The Journal of Immunology.

[14]  J. Yewdell,et al.  Immunodominance in major histocompatibility complex class I-restricted T lymphocyte responses. , 1999, Annual review of immunology.

[15]  Rainer Blasczyk,et al.  Peptide-binding motif of HLA-A*6603 , 2004, Immunogenetics.

[16]  Rainer Blasczyk,et al.  Residue 81 confers a restricted C-terminal peptide binding motif in HLA-B*44:09 , 2012, Immunogenetics.

[17]  Conrad C. Huang,et al.  UCSF Chimera—A visualization system for exploratory research and analysis , 2004, J. Comput. Chem..

[18]  S. Henikoff,et al.  Amino acid substitution matrices from protein blocks. , 1992, Proceedings of the National Academy of Sciences of the United States of America.

[19]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[20]  O. Lund,et al.  NetMHCpan, a Method for Quantitative Predictions of Peptide Binding to Any HLA-A and -B Locus Protein of Known Sequence , 2007, PloS one.

[21]  M. Nielsen,et al.  NetMHCpan-3.0; improved prediction of binding to MHC class I molecules integrating information from multiple receptor and peptide length datasets , 2016, Genome Medicine.

[22]  Dongsup Kim,et al.  Deep convolutional neural networks for pan-specific peptide-MHC class I binding prediction , 2017, BMC Bioinformatics.

[23]  Alessandro Sette,et al.  Generating quantitative models describing the sequence specificity of biological processes with the stabilized matrix method , 2005, BMC Bioinformatics.

[24]  J. Sidney,et al.  Bolstering the Number and Function of HSV-1–Specific CD8+ Effector Memory T Cells and Tissue-Resident Memory T Cells in Latently Infected Trigeminal Ganglia Reduces Recurrent Ocular Herpes Infection and Disease , 2017, The Journal of Immunology.

[25]  Xiaohui Xie,et al.  HLA class I binding prediction via convolutional neural networks , 2017, bioRxiv.

[26]  S. Lemieux,et al.  MHC class I-associated peptides derive from selective regions of the human genome. , 2016, The Journal of clinical investigation.

[27]  Geoffrey E. Hinton,et al.  Rectified Linear Units Improve Restricted Boltzmann Machines , 2010, ICML.

[28]  Daniel Jurafsky,et al.  A Hierarchical Neural Autoencoder for Paragraphs and Documents , 2015, ACL.

[29]  Morten Nielsen,et al.  Gapped sequence alignment using artificial neural networks: application to the MHC class I system , 2016, Bioinform..

[30]  Morten Nielsen,et al.  Dataset size and composition impact the reliability of performance benchmarks for peptide-MHC binding predictions , 2014, BMC Bioinformatics.

[31]  Sarah Rowland-Jones,et al.  Structures of Three HIV-1 HLA-B*5703-Peptide Complexes and Identification of Related HLAs Potentially Associated with Long-Term Nonprogression12 , 2005, The Journal of Immunology.

[32]  O. Lund,et al.  novel sequence representations Reliable prediction of T-cell epitopes using neural networks with , 2003 .

[33]  E. Mardis,et al.  A dendritic cell vaccine increases the breadth and diversity of melanoma neoantigen-specific T cells , 2015, Science.

[34]  Morten Nielsen,et al.  NetMHC-3.0: accurate web accessible predictions of human, mouse and monkey MHC class I affinities for peptides of length 8–11 , 2008, Nucleic Acids Res..

[35]  Deborah Hix,et al.  The immune epitope database (IEDB) 3.0 , 2014, Nucleic Acids Res..