Binding Prediction Tools Research Articles

An automated benchmarking platform for MHC class II binding prediction methods.

Computational methods for the prediction of peptide-MHC binding have become an integral and essential component for candidate selection in experimental T cell epitope discovery studies. The sheer amount of published prediction methods-and often discordant reports on their performance-poses a considerable quandary to the experimentalist who needs to choose the best tool for their research. With the goal to provide an unbiased, transparent evaluation of the state-of-the-art in the field, we created an automated platform to benchmark peptide-MHC class II binding prediction tools. The platform evaluates the absolute and relative predictive performance of all participating tools on data newly entered into the Immune Epitope Database (IEDB) before they are made public, thereby providing a frequent, unbiased assessment of available prediction tools. The benchmark runs on a weekly basis, is fully automated, and displays up-to-date results on a publicly accessible website. The initial benchmark described here included six commonly used prediction servers, but other tools are encouraged to join with a simple sign-up procedure. Performance evaluation on 59 data sets composed of over 10 000 binding affinity measurements suggested that NetMHCIIpan is currently the most accurate tool, followed by NN-align and the IEDB consensus method. Weekly reports on the participating methods can be found online at: http://tools.iedb.org/auto_bench/mhcii/weekly/. mniel@bioinformatics.dtu.dk. Supplementary data are available at Bioinformatics online.

An Approach for a Synthetic CTL Vaccine Design against Zika Flavivirus Using Class I and Class II Epitopes Identified by Computer Modeling.

The threat posed by severe congenital abnormalities related to Zika virus (ZKV) infection during pregnancy has turned development of a ZKV vaccine into an emergency. Recent work suggests that the cytotoxic T lymphocyte (CTL) response to infection is an important defense mechanism in response to ZKV. Here, we develop the rationale and strategy for a new approach to developing cytotoxic T lymphocyte (CTL) vaccines for ZKV flavivirus infection. The proposed approach is based on recent studies using a protein structure computer model for HIV epitope selection designed to select epitopes for CTL attack optimized for viruses that exhibit antigenic drift. Because naturally processed and presented human ZKV T cell epitopes have not yet been described, we identified predicted class I peptide sequences on ZKV matching previously identified DNV (Dengue) class I epitopes and by using a Major Histocompatibility Complex (MHC) binding prediction tool. A subset of those met the criteria for optimal CD8+ attack based on physical chemistry parameters determined by analysis of the ZKV protein structure encoded in open source Protein Data File (PDB) format files. We also identified candidate ZKV epitopes predicted to bind promiscuously to multiple HLA class II molecules that could provide help to the CTL responses. This work suggests that a CTL vaccine for ZKV may be possible even if ZKV exhibits significant antigenic drift. We have previously described a microsphere-based CTL vaccine platform capable of eliciting an immune response for class I epitopes in mice and are currently working toward in vivo testing of class I and class II epitope delivery directed against ZKV epitopes using the same microsphere-based vaccine.

Genomic analysis of 63,220 tumors reveals insights into tumor uniqueness and targeted cancer immunotherapy strategies

BackgroundThe integration of genomics with immunotherapy has potential value for cancer vaccine development. Given the clinical successes of immune checkpoint modulators, interest in cancer vaccines as therapeutic options has been revived. Current data suggest that each tumor contains a unique set of mutations (mutanome), thus requiring the creation of individualized cancer vaccines. However, rigorous analysis of non-individualized cancer immunotherapy approaches across multiple cancer types and in the context of known driver alterations has yet to be reported. We therefore set out to determine the feasibility of a generalizable cancer vaccine strategy based on targeting multiple neoantigens in an HLA-A/B subtype-directed manner.MethodsA cancer gene-focused, hybrid capture-based genomic analysis was performed on 63,220 unique tumors. Neoantigens were predicted using a combined peptide processing and MHC-I binding prediction tool (IEDB) for all recurrent (>10 tumors) missense alterations and non-frameshift indels for the two most common HLA-A/B subtypes in North American/European populations.ResultsDespite being overwhelmingly unique overall, many mutanomes (~45%) contain at least one mutation from a set of ten mutations chosen to maximize the number of unique tumors. This held true for tumors driven by KRAS G12C (n = 1799), PIK3CA E545K (n = 1713), or EGFR L858R (n = 478) alterations, which define distinct sample subsets. We therefore hypothesized that sets of carefully selected mutations/neoantigens may allow the development of broadly applicable semi-universal cancer vaccines. To test the feasibility of such an approach, antigen processing and MHC-I binding prediction was applied for HLA subtypes A*01:01/B*08:01 and A*02:01/B*44:02. In tumors with a specific HLA type, 0.7 and 2.5% harbored at least one of a set of ten neoantigens predicted to bind to each subtype, respectively. In comparison, KRAS G12C-driven tumors produced similar results (0.8 and 2.6% for each HLA subtype, respectively), indicating that neoantigen targets still remain highly diverse even within the context of major driver mutations.ConclusionsThis “best case scenario” analysis of a large tumor set across multiple cancer types and in the context of driver alterations reveals that semi-universal, HLA-specific cancer vaccine strategies will be relevant to only a small subset of the general population. Similar analysis of whole exome/genome sequencing, although not currently feasible at scale in a clinical setting, will likely uncover further diversity.

Epitope-Based Peptide Vaccine Design against Mokola Rabies Virus Glycoprotein G Utilizing In Silico Approaches

Background: Lyssavirus is considered as a neglected, zoonotic and tropical virus. Among all the Lyssavirus species known to exist today, Mokola virus is unique and appears to be exclusive to Africa. This virus is responsible for a meningoencephalomyelitis in mammals therefore; in silico prediction of epitopes of appropriate protein residues is important to produce a peptide vaccine with powerful immunogenic and minimal allergic effect. The aim of this study was to design a vaccine for Mokola virus using its glycoprotein peptides as an immunogen to stimulate protective immune response. Methods and materials: Glycoprotein G Sequences of Mokola was explored from NCBI then the sequences were aligned to obtain conserved regions. The nominees epitopes from Immune Epitope Database were analyzed by different prediction tools for B-cell, T-cell MHC class II and I. Then sequences aligned with the aid of ClustalW implemented in the BioEdit program. Results and conclusions: For Bepipred test of B-cell the total number of conserved epitopes was 85. For Emini surface accessibility prediction, 36 conserved epitopes were passing the default threshold 1.0. In Kolaskar and Tongaonkar antigenicity, 36 conserved epitopes gave score above the default threshold 1.045. However, there are only three epitopes that pass the three tests (LYTIPEK, LAHQK, YPSVPS). The reference glycoprotein strain was analyzed using IEDB MHC-I binding prediction tool to predict T cell epitope. Twenty conserved peptides were predicted to interact with different MHC-I alleles. For MHC-II binding prediction there were 47 conserved epitopes found to interact with MHC-II alleles. The peptides GQILIPEMQ, FRRLSHFRK and FVGYVTTTF had the affinity to bind the highest number of MHC-II alleles. World population coverage for MHC-I most promising 3 peptides FVDLHMPDV, FVGYVTTTF and RLFDGTWVS was 67.42%, while the world population coverage for most promising MHC-II peptides was 99.77%, for the binding to MHC-I and MHC-II, The peptide FVG TTTF world population coverage was 99.31%.

Bioinspired Design and Oriented Synthesis of Immunogenic Site-Specifically Penicilloylated Peptides.

β-Lactam antibiotics allergy is recognized as a public health concern. By covalently binding to serum proteins, penicillins are known to form immunogenic complexes. The latter are recognized and digested by antigen-presenting cells into drug-hapten peptides leading to the immunization of treated persons and IgE-mediated hypersensitivity reactions encompassing anaphylaxis. If type I allergic reactions to drugs are often unpredictable, they are known to be dependent on CD4+ T-cells. This fundamental study revisits the chemical basis of the benzylpenicillin (BP) allergy with the aim of identifying immunologically relevant biomimetic benzylpenicilloylated peptides through the analysis of BP-conjugated human serum albumin (BP-HSA) profile and the evaluation of the naı̈ve CD4+ T-cell responses to candidate BP-HSA-derived peptides. The chemical structures of BP-HSA bioconjugates synthesized in vitro at both physiological and basic pH were investigated by mass spectrometry. From the ten most representative lysine residues grafted by BP-hapten, HSA-bioinspired 15-mer peptide sequences were designed and the potential T-cell epitope profile of each peptide was predicted using two complementary in silico approaches, i.e., HLA class II binding prediction tools from the Immune Epitope Database and Analysis Resource (IEDB) and computational alanine scanning mutagenesis. Twelve structurally diversified benzylpenicilloylated peptides (BP-Ps) were selected and synthesized with the aid of a flexible synthesis pathway using an original benzylpenicilloylated lysine monomer as common precursor. In order to corroborate their predicted "epitope" profile, the naı̈ve CD4+ T-cell response specific to BP was evaluated through a coculture approach. To our knowledge, this study showed for the first time the ability of bioinspired peptides structurally stemming from BP-HSA to be recognized by naı̈ve CD4+ T-cells thus identifying a pre-existing T-cell repertoire for penicillin molecules bound to proteins. It also established a promising model approach expandable to other most frequently used penicillin classes of antibiotics to reveal biomimetic drug-modified antigenic peptides relevant for qualitative and quantitative drug allergy studies.

TepiTool: A Pipeline for Computational Prediction of T Cell Epitope Candidates.

Computational prediction of Tcell epitope candidates is currently being used in several applications including vaccine discovery studies, development of diagnostics, and removal of unwanted immune responses against protein therapeutics. There have been continuous improvements in the performance of MHC binding prediction tools, but their general adoption by immunologists has been slow due to the lack of user-friendly interfaces and guidelines. Current tools only provide minimal advice on what alleles to include, what lengths to consider, how to deal with homologous peptides, and what cutoffs should be considered relevant. This protocol provides step-by-step instructions with necessary recommendations for prediction of the best Tcell epitope candidates with the newly developed online tool called TepiTool. TepiTool, which is part of the Immune Epitope Database (IEDB), provides some of the top MHC binding prediction algorithms for number of species including humans, chimpanzees, bovines, gorillas, macaques, mice, and pigs. The TepiTool is freely accessible at http://tools.iedb.org/tepitool/. © 2016 by John Wiley & Sons, Inc.

Interaction between IGFBP7 and insulin: a theoretical and experimental study.

Insulin-like growth factor binding protein 7 (IGFBP7) can bind to insulin with high affinity which inhibits the early steps of insulin action. Lack of recognition mechanism impairs our understanding of insulin regulation before it binds to insulin receptor. Here we combine computational simulations with experimental methods to investigate the interaction between IGFBP7 and insulin. Molecular dynamics simulations indicated that His200 and Arg198 in IGFBP7 were key residues. Verified by experimental data, the interaction remained strong in single mutation systems R198E and H200F but became weak in double mutation system R198E-H200F relative to that in wild-type IGFBP7. The results and methods in present study could be adopted in future research of discovery of drugs by disrupting protein–protein interactions in insulin signaling. Nevertheless, the accuracy, reproducibility, and costs of free-energy calculation are still problems that need to be addressed before computational methods can become standard binding prediction tools in discovery pipelines.

Functional analysis of autism-associated NRXN1β gene promoter

Neurexins are neuron-specific synaptic proteins, and abnormal structure of Neurexin1β is closely associated with autism. To characterize the minimal promoter of autism-associated NRXN1β gene and identify functional elements regulating its transcription, luciferase reporter plasmids containing different regulatory regions upstream of NRXN1β gene were constructed. After transfecting HEK293 cells with these plasmids, the minimal promoter region of NRXN1β gene was determined by detecting the transcriptional activity of luciferase reporter genes while the corresponding functional elements that significantly enhance or inhibit the activity of reporter genes were further screened out. To identify cis-acting elements, continuous nucleotide mutation within the functional regions and adjacent DNA sequences were generated using site-directed mutagenesis techniques and then transcriptional regulatory elements in corresponding regions were analyzed using transcription factor binding prediction tool. Our results showed for the first time that the minimal promoter region of human NRXN1β gene is located between positions -88 and +156 (-88/+156); two regions -88/-73 and +156/+149 enhance while the region +229/+419 inhibits promoter activity. The region -84/-63 significantly enhances promoter activity as cis-acting elements, suggesting the presence of DBP and ABF1 transcription factor binding sites in this region.

Development and validation of a broad scheme for prediction of HLA class II restricted T cell epitopes

Computational prediction of HLA class II restricted T cell epitopes has great significance in many immunological studies including vaccine discovery. In recent years, prediction of HLA class II binding has improved significantly but a strategy to globally predict the most dominant epitopes has not been rigorously defined. Using human immunogenicity data associated with sets of 15-mer peptides overlapping by 10 residues spanning over 30 different allergens and bacterial antigens, and HLA class II binding prediction tools from the Immune Epitope Database and Analysis Resource (IEDB), we optimized a strategy to predict the top epitopes recognized by human populations. The most effective strategy was to select peptides based on predicted median binding percentiles for a set of seven DRB1 and DRB3/4/5 alleles. These results were validated with predictions on a blind set of 15 new allergens and bacterial antigens. We found that the top 21% predicted peptides (based on the predicted binding to seven DRB1 and DRB3/4/5 alleles) were required to capture 50% of the immune response. This corresponded to an IEDB consensus percentile rank of 20.0, which could be used as a universal prediction threshold. Utilizing actual binding data (as opposed to predicted binding data) did not appreciably change the efficacy of global predictions, suggesting that the imperfect predictive capacity is not due to poor algorithm performance, but intrinsic limitations of HLA class II epitope prediction schema based on HLA binding in genetically diverse human populations.

Automated benchmarking of peptide-MHC class I binding predictions.

Numerous in silico methods predicting peptide binding to major histocompatibility complex (MHC) class I molecules have been developed over the last decades. However, the multitude of available prediction tools makes it non-trivial for the end-user to select which tool to use for a given task. To provide a solid basis on which to compare different prediction tools, we here describe a framework for the automated benchmarking of peptide-MHC class I binding prediction tools. The framework runs weekly benchmarks on data that are newly entered into the Immune Epitope Database (IEDB), giving the public access to frequent, up-to-date performance evaluations of all participating tools. To overcome potential selection bias in the data included in the IEDB, a strategy was implemented that suggests a set of peptides for which different prediction methods give divergent predictions as to their binding capability. Upon experimental binding validation, these peptides entered the benchmark study. The benchmark has run for 15 weeks and includes evaluation of 44 datasets covering 17 MHC alleles and more than 4000 peptide-MHC binding measurements. Inspection of the results allows the end-user to make educated selections between participating tools. Of the four participating servers, NetMHCpan performed the best, followed by ANN, SMM and finally ARB. Up-to-date performance evaluations of each server can be found online at http://tools.iedb.org/auto_bench/mhci/weekly. All prediction tool developers are invited to participate in the benchmark. Sign-up instructions are available at http://tools.iedb.org/auto_bench/mhci/join. mniel@cbs.dtu.dk or bpeters@liai.org Supplementary data are available at Bioinformatics online.

Enhancing in silico protein-based vaccine discovery for eukaryotic pathogens using predicted peptide-MHC binding and peptide conservation scores.

Given thousands of proteins constituting a eukaryotic pathogen, the principal objective for a high-throughput in silico vaccine discovery pipeline is to select those proteins worthy of laboratory validation. Accurate prediction of T-cell epitopes on protein antigens is one crucial piece of evidence that would aid in this selection. Prediction of peptides recognised by T-cell receptors have to date proved to be of insufficient accuracy. The in silico approach is consequently reliant on an indirect method, which involves the prediction of peptides binding to major histocompatibility complex (MHC) molecules. There is no guarantee nevertheless that predicted peptide-MHC complexes will be presented by antigen-presenting cells and/or recognised by cognate T-cell receptors. The aim of this study was to determine if predicted peptide-MHC binding scores could provide contributing evidence to establish a protein’s potential as a vaccine. Using T-Cell MHC class I binding prediction tools provided by the Immune Epitope Database and Analysis Resource, peptide binding affinity to 76 common MHC I alleles were predicted for 160 Toxoplasma gondii proteins: 75 taken from published studies represented proteins known or expected to induce T-cell immune responses and 85 considered less likely vaccine candidates. The results show there is no universal set of rules that can be applied directly to binding scores to distinguish a vaccine from a non-vaccine candidate. We present, however, two proposed strategies exploiting binding scores that provide supporting evidence that a protein is likely to induce a T-cell immune response–one using random forest (a machine learning algorithm) with a 72% sensitivity and 82.4% specificity and the other, using amino acid conservation scores with a 74.6% sensitivity and 70.5% specificity when applied to the 160 benchmark proteins. More importantly, the binding score strategies are valuable evidence contributors to the overall in silico vaccine discovery pool of evidence.

Optimized Potential of Mean Force Calculations for Standard Binding Free Energies

The prediction of protein-ligand binding free energies is an important goal of computational biochemistry, yet accuracy, reproducibility, and cost remain a problem. Nevertheless, these are essential requirements for computational methods to become standard binding prediction tools in discovery pipelines. Here, we present the results of an extensive search for an optimal method based on an ensemble of umbrella sampling all-atom molecular simulations tested on the phosphorylated tetrapeptide, pYEEI, binding to the SH2 domain, resulting in an accurate and converged binding free energy of -9.0 ± 0.5 kcal/mol (compared to an experimental value of -8.0 ± 0.1 kcal/mol). We find that a minimum of 300 ns of sampling is required for every prediction, a target easily achievable using new generation accelerated MD codes. Convergence is obtained by using an ensemble of simulations per window, each starting from different initial conformations, and by optimizing window-width, orthogonal restraints, reaction coordinate harmonic potentials, and window-sample time. The use of uncorrelated initial conformations in neighboring windows is important for correctly sampling conformational transitions from the unbound to bound states that affect significantly the precision of the calculations. This methodology thus provides a general recipe for reproducible and practical computations of binding free energies for a class of semirigid protein-ligand systems, within the limit of the accuracy of the force field used.

A Systematic Assessment of MHC Class II Peptide Binding Predictions and Evaluation of a Consensus Approach

The identification of MHC class II restricted peptide epitopes is an important goal in immunological research. A number of computational tools have been developed for this purpose, but there is a lack of large-scale systematic evaluation of their performance. Herein, we used a comprehensive dataset consisting of more than 10,000 previously unpublished MHC-peptide binding affinities, 29 peptide/MHC crystal structures, and 664 peptides experimentally tested for CD4+ T cell responses to systematically evaluate the performances of publicly available MHC class II binding prediction tools. While in selected instances the best tools were associated with AUC values up to 0.86, in general, class II predictions did not perform as well as historically noted for class I predictions. It appears that the ability of MHC class II molecules to bind variable length peptides, which requires the correct assignment of peptide binding cores, is a critical factor limiting the performance of existing prediction tools. To improve performance, we implemented a consensus prediction approach that combines methods with top performances. We show that this consensus approach achieved best overall performance. Finally, we make the large datasets used publicly available as a benchmark to facilitate further development of MHC class II binding peptide prediction methods.

Identification of HLA-A24-restricted epitopes with high affinities to Hsp70 using peptide arrays

Heat shock protein 70 (Hsp70) family members are known as facilitators of immune responses by interacting with receptors on antigen-presenting cells leading to Hsp70-peptide uptake and antigen cross priming. Here, identification of human leukocyte antigen (HLA)-A24-restricted epitopes was achieved using peptide arrays for evaluation of their affinities to Hsp70 and HLA-A24 binding prediction tools. Using Hsp70 as the model antigen, the GYPVTNAVI and VFQHGKVEI peptides were identified as antigens. These peptides actually bound to HLA-A24 in the stabilization assay using T2-A*2402 cells, and induced a strong peptide-reactive cytotoxic T lymphocyte response in HLA-A24 transgenic mice after vaccination.

A consensus strategy for combining HLA-DR binding algorithms

Diverse approaches have generated algorithms for predicting the binding of peptides and human leukocyte antigen DR (HLA-DR) molecules. As prediction tools appear online, interesting questions arise. Are the various predictions accurate? Do the predictions agree? Is there a value to combined results over individual predictions? Three directly accessible online HLA-DR binding prediction tools are evaluated here to address these questions. The three algorithms generate predictions for high affinity binding that partially agree. No one algorithm is consistently superior or inferior. When a consensus strategy provides a fourth prediction, it is consistently best or second best, regardless of the performance measure being utilized.