Dynamics-Based Peptide-MHC Binding Optimization by a Convolutional Variational Autoencoder: A Use-Case Model for CASTELO.

An unsolved challenge in the development of antigen-specific immunotherapies is determining the optimal antigens to target. Comprehension of antigen-major histocompatibility complex (MHC) binding is paramount toward achieving this goal. Here, we apply CASTELO, a combined machine learning-molecular dynamics (ML-MD) approach, to identify per-residue antigen binding contributions and then design novel antigens of increased MHC-II binding affinity for a type 1 diabetes-implicated system. We build upon a small-molecule lead optimization algorithm by training a convolutional variational autoencoder (CVAE) on MD trajectories of 48 different systems across four antigens and four HLA serotypes. We develop several new machine learning metrics including a structure-based anchor residue classification model as well as cluster comparison scores. ML-MD predictions agree well with experimental binding results and free energy perturbation-predicted binding affinities. Moreover, ML-MD metrics are independent of traditional MD stability metrics such as contact area and root-mean-square fluctuations (RMSF), which do not reflect binding affinity data. Our work supports the role of structure-based deep learning techniques in antigen-specific immunotherapy design.

[1]  Laxmikant V. Kalé,et al.  Scalable molecular dynamics with NAMD , 2005, J. Comput. Chem..

[2]  Andrew K. Sewell,et al.  Why must T cells be cross-reactive? , 2012, Nature Reviews Immunology.

[3]  Aristides Gionis,et al.  Clustering aggregation , 2005, 21st International Conference on Data Engineering (ICDE'05).

[4]  J. Strominger,et al.  Conformational variants of class II MHC/peptide complexes induced by N- and C-terminal extensions of minimal peptide epitopes. , 1999, Proceedings of the National Academy of Sciences of the United States of America.

[5]  Guojing Cong,et al.  CASTELO: clustered atom subtypes aided lead optimization—a combined machine learning and molecular modeling method , 2020, BMC Bioinformatics.

[6]  Serena H. Chen,et al.  Graphene-extracted membrane lipids facilitate the activation of integrin αvβ8. , 2020, Nanoscale.

[7]  CASTELO: clustered atom subtypes aided lead optimization—a combined machine learning and molecular modeling method , 2021, BMC Bioinform..

[8]  James McCluskey,et al.  T cell antigen receptor recognition of antigen-presenting molecules. , 2015, Annual review of immunology.

[9]  T. Lybrand,et al.  Polymorphic Amino Acid Variations in HLA-DQ Are Associated With Systematic Physical Property Changes and Occurrence of IDDM , 1995, Diabetes.

[10]  Irini Doytchinova,et al.  Peptide binding prediction for the human class II MHC allele HLA-DP2: a molecular docking approach , 2011, BMC Structural Biology.

[11]  B. Walker,et al.  T cell receptors for the HIV KK10 epitope from patients with differential immunologic control are functionally indistinguishable , 2018, Proceedings of the National Academy of Sciences.

[12]  E. Unanue,et al.  Register shifting of an insulin peptide–MHC complex allows diabetogenic T cells to escape thymic deletion , 2011, The Journal of experimental medicine.

[13]  Yi Wang,et al.  Scalable molecular dynamics on CPU and GPU architectures with NAMD. , 2020, The Journal of chemical physics.

[14]  B. Baker,et al.  Increased Immunogenicity of an Anchor-Modified Tumor-Associated Antigen Is Due to the Enhanced Stability of the Peptide/MHC Complex: Implications for Vaccine Design1 , 2005, The Journal of Immunology.

[15]  A. Sewell,et al.  Structural basis for ineffective T-cell responses to MHC anchor residue-improved “heteroclitic” peptides , 2014, European journal of immunology.

[16]  Jayvee R. Abella,et al.  Structure-based Methods for Binding Mode and Binding Affinity Prediction for Peptide-MHC Complexes. , 2019, Current topics in medicinal chemistry.

[17]  Carl Doersch,et al.  Tutorial on Variational Autoencoders , 2016, ArXiv.

[18]  Yuko Tsuchiya,et al.  Autoencoder-Based Detection of Dynamic Allostery Triggered by Ligand Binding Based on Molecular Dynamics , 2019, J. Chem. Inf. Model..

[19]  John Gounley,et al.  How Distinct Structural Flexibility within SARS-CoV-2 Spike Protein Reveals Potential Therapeutic Targets , 2021, 2021 IEEE International Conference on Big Data (Big Data).

[20]  Ron Elber,et al.  Comprehensive analysis of sequences of a protein switch , 2016, Protein science : a publication of the Protein Society.

[21]  R. Zhou,et al.  Lanosterol Disrupts Aggregation of Human γD-Crystallin by Binding to the Hydrophobic Dimerization Interface. , 2018, Journal of the American Chemical Society.

[22]  William S. Lane,et al.  Predominant naturally processed peptides bound to HLA-DR1 are derived from MHC-related molecules and are heterogeneous in size , 1992, Nature.

[23]  Z. Weng,et al.  A flexible docking approach for prediction of T cell receptor–peptide–MHC complexes , 2013, Protein science : a publication of the Protein Society.

[24]  E. Unanue,et al.  Variations in MHC Class II Antigen Processing and Presentation in Health and Disease. , 2016, Annual review of immunology.

[25]  K Schulten,et al.  VMD: visual molecular dynamics. , 1996, Journal of molecular graphics.

[26]  Ruhong Zhou,et al.  A Public BCR Present in a Unique Dual-Receptor-Expressing Lymphocyte from Type 1 Diabetes Patients Encodes a Potent T Cell Autoantigen , 2019, Cell.

[27]  Morten Nielsen,et al.  NetMHCpan-4.1 and NetMHCIIpan-4.0: improved predictions of MHC antigen presentation by concurrent motif deconvolution and integration of MS MHC eluted ligand data , 2020, Nucleic Acids Res..

[28]  Frank Noé,et al.  MHC class II complexes sample intermediate states along the peptide exchange pathway , 2016, Nature Communications.

[29]  Oliver Kohlbacher,et al.  Immunoinformatics and epitope prediction in the age of genomic medicine , 2015, Genome Medicine.

[30]  Morten Nielsen,et al.  Improved prediction of MHC II antigen presentation through integration and motif deconvolution of mass spectrometry MHC eluted ligand data. , 2020, Journal of proteome research.

[31]  P. Marrack,et al.  Diabetogenic T cells recognize insulin bound to IAg7 in an unexpected, weakly binding register , 2010, Proceedings of the National Academy of Sciences.

[32]  Ricardo J. G. B. Campello,et al.  Density-Based Clustering Based on Hierarchical Density Estimates , 2013, PAKDD.

[33]  Hau-San Wong,et al.  TEPITOPEpan: Extending TEPITOPE for Peptide Binding Prediction Covering over 700 HLA-DR Molecules , 2012, PloS one.

[34]  E. Unanue,et al.  The Insulin-Specific T Cells of Nonobese Diabetic Mice Recognize a Weak MHC-Binding Segment in More Than One Form1 , 2007, The Journal of Immunology.

[35]  Jayvee R. Abella,et al.  APE-Gen: A Fast Method for Generating Ensembles of Bound Peptide-MHC Conformations , 2019, Molecules.

[36]  T. Lybrand,et al.  Polymorphic amino acid variations in HLA-DQ are associated with systematic physical property changes and occurrence of IDDM. Members of the Swedish Childhood Diabetes Study. , 1995 .

[37]  E. Bergseng,et al.  Inhibition of HLA-DQ2-mediated antigen presentation by analogues of a high affinity 33-residue peptide from alpha2-gliadin. , 2006, Journal of the American Chemical Society.

[38]  David R. Bell,et al.  Commensal bacteria stimulate antitumor responses via T cell cross-reactivity. , 2020, JCI insight.

[39]  Morten Nielsen,et al.  NetMHCpan 4.0: Improved peptide-MHC class I interaction predictions integrating eluted ligand and peptide binding affinity data , 2017, bioRxiv.

[40]  Kristin Ladell,et al.  Modification of MHC Anchor Residues Generates Heteroclitic Peptides That Alter TCR Binding and T Cell Recognition , 2010, The Journal of Immunology.

[41]  Alex Rubinsteyn,et al.  MHCflurry: Open-Source Class I MHC Binding Affinity Prediction. , 2018, Cell systems.

[42]  Jing Huang,et al.  CHARMM36 all‐atom additive protein force field: Validation based on comparison to NMR data , 2013, J. Comput. Chem..

[43]  Alessandro Laio,et al.  Predicting the affinity of peptides to MHC class II by scoring molecular dynamics simulations. , 2019, Journal of chemical information and modeling.

[44]  Zheng Pu,et al.  Amino-Terminal Flanking Residues Determine the Conformation of a Peptide–Class II MHC Complex1 , 2006, The Journal of Immunology.

[45]  P. Marrack,et al.  How C-terminal additions to insulin B-chain fragments create superagonists for T cells in mouse and human type 1 diabetes , 2019, Science Immunology.

[46]  P. Jensen,et al.  Structural Characteristics of HLA-DQ that May Impact DM Editing and Susceptibility to Type-1 Diabetes , 2013, Front. Immunol..

[47]  M. Berger,et al.  Patient HLA class I genotype influences cancer response to checkpoint blockade immunotherapy , 2018, Science.

[48]  Laxmikant V. Kalé,et al.  Scalable molecular dynamics with NAMD on the IBM Blue Gene/L system , 2008, IBM J. Res. Dev..

[49]  P. Lackner,et al.  MHCII3D—Robust Structure Based Prediction of MHC II Binding Peptides , 2020, International journal of molecular sciences.

[50]  T. Huynh,et al.  Free-energy simulations reveal that both hydrophobic and polar interactions are important for influenza hemagglutinin antibody binding. , 2012, Biophysical journal.

[51]  P. Marrack,et al.  C-terminal modification of the insulin B:11–23 peptide creates superagonists in mouse and human type 1 diabetes , 2017, Proceedings of the National Academy of Sciences.

[52]  Ron Elber,et al.  The energy landscape of a protein switch. , 2014, Physical chemistry chemical physics : PCCP.

[53]  B. Walker,et al.  The complex and specific pMHC interactions with diverse HIV-1 TCR clonotypes reveal a structural basis for alterations in CTL function , 2014, Scientific Reports.

[54]  Eaman Alhassan,et al.  Novel Nondietary Therapies for Celiac Disease , 2019, Cellular and molecular gastroenterology and hepatology.

[55]  Irini Doytchinova,et al.  T-cell epitope vaccine design by immunoinformatics , 2013, Open Biology.

[56]  Claude Beazley,et al.  A Novel Predictive Technique for the MHC Class II Peptide-Binding Interaction , 2003, Molecular medicine.

[57]  Gabriella Sármay,et al.  Antigen-specific immunotherapies in rheumatic diseases , 2017, Nature Reviews Rheumatology.

[58]  Jayvee R. Abella,et al.  HLA-Arena: A Customizable Environment for the Structural Modeling and Analysis of Peptide-HLA Complexes for Cancer Immunotherapy , 2020, JCO clinical cancer informatics.

[59]  Shang Gao,et al.  Deep clustering of protein folding simulations , 2018, BMC Bioinformatics.

[60]  Charlotte M. Deane,et al.  Current status and future challenges in T-cell receptor/peptide/MHC molecular dynamics simulations , 2015, Briefings Bioinform..

[61]  Marcus D. Hanwell,et al.  Avogadro: an advanced semantic chemical editor, visualization, and analysis platform , 2012, Journal of Cheminformatics.

[62]  W. Langridge,et al.  Autoantigen based vaccines for type 1 diabetes. , 2011, Discovery medicine.

[63]  Serena H. Chen,et al.  Charging nanoparticles: increased binding of Gd@C82(OH)22 derivatives to human MMP-9. , 2018, Nanoscale.

[64]  David R. Bell,et al.  In silico design and validation of high-affinity RNA aptamers targeting epithelial cellular adhesion molecule dimers , 2020, Proceedings of the National Academy of Sciences.

[65]  Morten Nielsen,et al.  MHC Class II epitope predictive algorithms , 2010, Immunology.

[66]  B. Luan,et al.  Parameterization of Molybdenum Disulfide Interacting with Water Using Free Energy Perturbation Method. , 2019, The journal of physical chemistry. B.

[67]  Zhiping Weng,et al.  High-throughput modeling and scoring of TCR-pMHC complexes to predict cross-reactive peptides , 2020, Bioinform..

[68]  Frank Noé,et al.  Major Histocompatibility Complex (MHC) Class I and MHC Class II Proteins: Conformational Plasticity in Antigen Presentation , 2017, Front. Immunol..

[69]  Didier Devaurs,et al.  General Prediction of Peptide-MHC Binding Modes Using Incremental Docking: A Proof of Concept , 2018, Scientific Reports.

[70]  P. Petrone,et al.  MHC-peptide binding is assisted by bound water molecules. , 2004, Journal of molecular biology.

[71]  Weilong Zhao,et al.  Systematically benchmarking peptide-MHC binding predictors: From synthetic to naturally processed epitopes , 2018, PLoS Comput. Biol..

[72]  R. Zhou,et al.  Structural Basis of the Potential Binding Mechanism of Remdesivir to SARS-CoV-2 RNA-Dependent RNA Polymerase , 2020, The journal of physical chemistry. B.