key: cord-0789693-i7oi7mfi
authors: Mitra, Debarghya; Pandey, Janmejay; Swaroop, Shiv
title: Multi-epitope-based peptide vaccine design against SARS-CoV-2 using its spike protein
date: 2020-07-01
journal: bioRxiv
DOI: 10.1101/2020.04.23.055467
sha: eb51da16793bd0dc824a9d8f79d6391749f68927
doc_id: 789693
cord_uid: i7oi7mfi

SARS CoV-2 has particularly been efficient in ensuring that many countries are brought to a standstill. With repercussions ranging from rampant mortality, fear, paranoia and economic recession, the virus has brought together countries in order to look at possible therapeutic countermeasures. With prophylactic interventions possibly months away from being particularly effective, a slew of measures and possibilities concerning the design of vaccines are being worked upon. We attempted a structure-based approach utilizing a combination of epitope prediction servers to develop a multi-epitope-based subunit vaccine that involves the two major domains of the spike glycoprotein of SARS CoV-2 (S1 and S2) coupled with a substantially effective chimeric adjuvant to create stable vaccine constructs through MD simulations. The designed constructs were evaluated based on their docking with Toll Like Receptor (TLR) 4. Our findings provide an epitope-based peptide fragment; which can be a potential candidate for the development of a vaccine against SARS-CoV-2. Recent experimental studies based on determining immunodominant regions across the spike glycoprotein of SARS-CoV-2 indicate the presence of the predicted epitopes included in this study.

between S1 and S2 post-fusion has not been considered due to previous apprehensions on the hypersensitivity of the immune system considering the entire spike glycoprotein of SARS-CoV.

The utilization of these epitopes and the vaccine construct across an experimental setting will help provide essential evaluation and aid in the development of a vaccine to generate a robust immunological prophylactic response against SARS-CoV-2.

The SARS-CoV-2 spike glycoprotein sequence (PDB ID: 6VSB) [6] was checked for specific domains and the identified S1 domain (residues) comprising the Receptor Binding Domain(RBD) along with the S2 domain(residues) were edited from the cryo-EM structure of the surface glycoprotein using UCSF Chimera [12] .

Based on the sequences of S1 and S2 domains, they were separately used to determine linear B cell epitopes through different servers. The Artificial Neural Network (ANN) approach-based server ABCPred [13] , Bcepred [14] and the Immune Epitope Database and Analysis Resource (IEDB) based linear B cell epitope prediction tools [15] [16] [17] [18] [19] [20] were utilized to predict probable epitope sequences across the query sequences. The latter two servers utilize physicochemical parameters like hydrophilicity, polarity, surface accessibility and flexibility to predict a B cell epitope as has been previously evidenced. The predicted linear B cell sequences from ABCPred were matched with the probable epitopes predicted over the different parameters inclusive of surface accessibility area, hydrophilicity, polarity and flexibility in that order over the IEDBbased Bepipred 2.0 based linear prediction of epitopes and the results from the two servers were corroborated with these parameters. These same parameters were then again checked with the predicted epitopes by Bcepred, which predicts epitopes on the basis of the same parameters.

Consensus sequences which emerged from all three were considered to be the probable epitopes in this regard. Since conformational or discontinuous B cell epitopes could be predicted through ElliPro [21] and Discotope [22] , they were utilized for analysis of the consensus sequences arrived upon earlier. Consensus 16-mer epitopes were predicted using these servers.

The probable MHC II binding epitopes on S1 and S2 domains were predicted and again a consensus approach was utilized to arrive at the conclusive sequences. The prediction of helper T lymphocytes was made using MHCPred [23] , SYFPEITHI [24] , NetMHCIIpan 3.2 server [25] and the IEDB server [26, 27] . A consensus selection of epitopes from the four different servers allowed us to improve upon and circumvent the limitations associated with prediction of MHC II binders due to their polymorphic nature, peptide length variation and also determination of the appropriate peptide binding core. Hence, looking into the limitations these servers were determined to be among the best available that can be utilized in this avenue. Each of these prediction servers was compared with experimental datasets to assess their performance.

SYFPEITHI predicts nonamer sequences based on the weighted contribution of each amino acid sequence present across a predicted epitope sequence. MHCPred allows prediction of 9-mer epitopes based on multivariate statistical methods and NetMHCIIpan 3.2 allows the determination of 15-mer epitope fragments but with limited allele-specific choices. The IEDBbased MHC II binding epitope sequence prediction server was also utilized because it employs a combination of methods ranging from ANN to SMM and ranks probable epitopes on the basis of a percentile score and an IC 50 score. The sequences predicted over the different servers were either 15 mers or 9 mers based on the server in use, but based on the consensus selection across these platforms and overlapping regions of the predicted epitope sequences, uniform 15-mer epitopes were selected.

The prediction of Cytotoxic T Lymphocyte cells (T CTL ) through MHC I binding servers involved the utilization of NetMHC 4.0 [28] , MHC-NP [29] , NetCTL 1.2 [30] and the IEDB-based T cell epitope tools [31] . All these servers predict a nonamer epitope sequence and with the use of a default dataset along with probable interacting human leukocyte antigen alleles with SARS CoV identified from literature. Employing a consensus selection of the predictions from the four servers, we were able to list the appropriate T CTL epitopes with relative confidence. The NetMHC 4.0 utilizes an Artificial Neural Network to predict epitopes, NetCTL 1.2 server allows for the identification of epitopes based on improved datasets with a sensitivity of 0.80 and specificity of 0.97 across the filtering threshold employed, MHC-NP employs a Machine Learning approach towards prediction of naturally processed epitopes by the MHC, whereas IEDB-based MHC I based prediction of T cell epitopes sorts epitopes based on the percentile score and low IC 50 values across a combination of ANN and SMM approaches based on an appropriate peptide library. The MHC I alleles specific for a SARS CoV 2 manifestation were based on the alleles confirmed during the outbreak of SARS CoV at the beginning of the millennium.

The consensus sequences thus arrived upon can suitably be considered as being capable of eliciting the necessary immune response. Additionally, each of the sequences were filtered based on their predicted antigenicity over Vaxijen [32] , allergenicity over AllerTop [33] and Algpred [34] , toxicity over ToxinPred [35] , capable of eliciting Interferon-gamma over IFN-G [36] servers. Also, the epitope sequences were matched with human proteins to prevent cases of antibody response against a self-antigen with the Multiple Peptide Match Tool. Each of the epitope sequences were matched with their corresponding secondary structure over the cryoEM structure (PDB ID: 6VSB) [6] and listed to be either alpha helices, coils, strands or beta sheets.

This was done to allow for sequence arrangement to allow for improved molecular modeling.

Another thing that we had not considered initially based on the conservation of the surface glycoprotein across the different strains deposited over the GISAID repository (https://www.gisaid.org/about-us/mission/ ) and the opinion that not much variability has been observed across these strains led us to check the coverage of the epitope sequences we have predicted with the sequences from the repository of surface glycoprotein and also the genome sequence deposited over NCBI from India [37] .

In a peptide-based subunit vaccine, the predicted B and T cell epitopes are not sufficient to elicit a strong immune response to generate the necessary prophylactic measures. Therefore, a suitable adjuvant must also be added to the vaccine design [38] . In most cases, such purification can be carried out separately as in the purification of the adjuvant and the peptide epitope sequence separately. Recently, several approaches have been made to utilize a chimeric adjuvant comprising two or more separately identified adjuvants [39, 40] . A similar approach was carried out and an in silico design comprising of three different adjuvants with evidence as being used as adjuvants in different vaccines or as agonists of TLR-4 receptors were utilized. A recent review on suitable adjuvants for different TLRs using this approach has been published [41] . It was backed up by experimental evidence of the identified adjuvants being capable of eliciting an immune response on interaction with toll-like receptors (TLRs) separately and downstream immune signaling generating the necessary prophylactic measures that can be expected from a vaccine against SARS CoV-2 [42] . The generation of neutralization antibodies by the peptide sequences of the surface glycoprotein has already been evidenced in previous instances and hence their linking with a triagonist chimeric adjuvant supports their utilization as a suitable prophylactic measure. A suitable rigid linker was utilized to join them with the peptide sequence and their position was rearranged at the N-terminal to design several constructs of the vaccine. A single construct with adjuvants at the N and C-terminal of the vaccine was also considered.

Appropriate linkers were utilized to join intra B cell epitopes, T-Helper and T-Cytotoxic epitopes and also between them in the vaccine construct [43, 44] . The construct design comprises of the B cell epitopes linked to the adjuvant through a rigid EAAAK linker (helix forming) toward the Nterminal, followed by a GPGPG linker linking the T HTL epitopes and an AAY linker for the T CTL epitopes toward the C-terminal. A DP linker was utilized to link the three adjuvants at the N terminal. The three adjuvants were rearranged each time and the vaccine constructs were analyzed for physicochemical properties using ProtPARAM tool [45] , allergenicity [33] , antigenicity [32] and scanned for probable stimulants of interferon gamma [36] . The ToxinPred database [35] was used to analyze each of the units of the vaccine construct and hence each of these constructs can be considered nontoxic.

The main challenge associated with the molecular modeling of the vaccine constructs was that the adjuvant-specific region and the epitope-specific region matched with two different templates. Although in both cases we had sufficient sequence coverage that calls for homology modeling of the vaccine constructs, the multi-template alignment led to a modeled structure comprising mainly of strands, which would have made the modeled constructs unstable. An initial approach included modeling the two parts of the vaccine construct separately using a single template, but linking both of these through loops and energy minimizations of these constructs fell through. The ROBETTA server [46] , which allows for comparative modeling, brought the most promising model of the vaccine constructs with sufficient secondary structure coverage. Each of these constructs was then deposited over GALAXY webserver [47] for refinement. Based on select parameters based on which each modeled construct was refined, the best structures were then carried forward for validation through Ramachandran Plot, Z-score over PROCHECK [48] and ProSA-web server [49] . An additional validation was carried out using ERRAT-3D server [50] . Each of the modeled vaccine constructs was also checked for sequence mismatches.

Each of these constructs was then assessed for their stability using the ProtParam server. The various markers for stability made available through the Instability index (II) [51] , PEST hypothesis [52] , in vivo half-life and the isoelectric point [53] . The assessments based on Ramachandran Plot, ERRAT 3D [50] , WHATCHECK [54] and the ProSA-web server [49] allowed us to select a single vaccine construct for a prolonged molecular dynamic simulation to determine the stability of the construct and probable use in in vivo settings.

MD simulations to obtain elaborate insights on the dynamic stability of the construct in physiological conditions were performed on Desmond from D E Shaw Research (DESRES) [55] .

The protein structures were subjected to minimization in a 10 X 10 X 20 Å 3 periodic boundary box with the TIP3P water model using a fast scalable algorithm utilizing a distributed memory parallel program. Counter-ions and salt were added to neutralize the system. Pressure and temperature were maintained using a Nose-Hoover thermostat and a Martyna-Tobias-Klein barostat at 310K and 1 bar, respectively. All the systems were run for 100ns in the NPT ensemble. Trajectories were analyzed with a simulation interaction diagram in Desmond [55] and VMD [56] .

The rationale behind selecting TLR-4 (PDB ID: 3FXI_A) [57] is the fact that in the case of SARS CoV, HIV, Influenzae and other RNA-based viruses, this Toll-like receptor has been experimentally evidenced to be implicated [58] . The binding sites over TLR-4 were determined through available PDB IDs that showed binding of an adjuvant lipopolysaccharide with the Toll like receptor and also through ElliPro [21] and Castp [59] . It was carried out through Patchdock [60] and Firedock [61] .

Post docking, the top pose was evaluated through Molecular Mechanics/Generalized Born Surface Area (MM/GBSA) calculations over the HawkDock Server [62] . The server allows the docked pose to be assessed on a per-residue basis across Van der Waal potentials, electrostatic potentials, polar solvation free-energy and solvation free energy through empirical models. The docked pose is minimized for 5000 steps, including 2000 cycles of steepest descents and 3000 cycles of conjugate gradient minimizations based on an implicit solvent model, the ff02 force field.

The sequence of the SARS CoV 2 surface glycoprotein was retrieved from PDB ID: 6VSB in FASTA format. The UCSF Chimera visualization software [12] was utilized to edit the structure and include the S1 and the S2 domains as separate entities. The use of S1 and S2 domains separately without including the cleavage site, which falls in between the two domains ( Figure   1 ) for the development of an immunogen is due to previously evidenced cases of increased immune hypersensitivity with concerns over utilizing the entire length of spike glycoprotein as a vaccine immunogen. [63] Figure 1: The S1 and S2 domains considered in the investigation are reproduced from PDB ID: 6VSB. 

The selected epitopes were based on multiple servers and since this rationale is not heuristic but based on the role of the different algorithms that lead each server to arrive at the predicted epitope.Based on the stringent filters applied in each case, the epitopes predicted over the different servers were manually curated to arrive upon the consensus sequences in each case. All the epitope sequences specific for S1 and S2, separately predicted by the servers, are listed. In the case of B cell and MHC II epitopes 15 mers epitope sequences were generated, while in the case of MHC I epitope sequences, nonamer fragments were generated.

The predicted sequences were matched with the data generated by [64] through IEDB on significant epitopes present on the surface glycoprotein, which did mention the need for analysis of the predicted sites and hence the utility of the stringent filters over the different servers followed by validation. The utilization of specific alleles which have been identified in the previous manifestation of the SARS CoV has helped us sort the predicted T cell epitopes appropriately on the different servers, which otherwise would have made prediction biased [65] [66] [67] [68] . The selected epitopes are listed in Table 1 . 

The various parameters against which the selected epitopes and adjuvants were filtered against included antigenicity, allergenicity, toxicity, capability of generating interferon gamma response (Tables 1A,B and C) and chances of being identified as self-antigen by the human immune system. Among the epitope sequences hence filtered, 100 percent sequence coverage was detected in all but single residue mutations in 4 of them. These residues were not considered in the vaccine constructs and the aberrations were generally isolated to the S1 domain only ( Figure   2B ). Hence, sequence coverage based validation of the selected epitopes was also carried out.

The list of epitopes considered for design of the vaccines did not include any sequences that did not conform to the 100% sequence coverage. The percent identity matrix utilized to indicate that the Indian sequence shows percent identity of more than 90% with the sequence retrieved from PDB ID:6VSB and the epitopes utilized show 100% sequence coverage across the sequence submitted from India of the spike glycoproteins [37] (Figure 2 ). 

The triagonist chimeric adjuvant has the propensity to elicit immune responses to allow for its use as a suitable adjuvant against SARS CoV 2. RS09 [69] , TR-433 [70] and a Mtb based Hsp70 partial sequence [71, 72] were utilized in this endeavor. An appropriate rigid polar linker (DP) was utilized in this regard. Among the designed constructs all of the rearrangements were validated based on their antigenicity score predicted by Vaxijen. The adjuvants considered were all verified as TLR-4 agonists. The use of fragments of Mycobacterial hsp70 towards the generation of cytokines and natural killer cells has been verified and also for antigen-specific CTL responses [73] [74] [75] . The initial constructs and the parameters across which they were A B assessed are mentioned in Supplementary Table 1A and 1B. The use of lipopolysaccharide based adjuvants has been abrogated in this study due to constraints of utilizing modeling, docking and simulation-based studies from their perspective.

The arrangement and validation of antigenicity of the vaccine constructs based on the linkers and the adjuvants led to the terminal selection of 5 different vaccine constructs, which were then subsequently characterized based on the ProtParam parameters, which determined the molecular weight and isoelectric point and predicted each to be a stable construct. Also, the constructs were found to be suitably antigenic, nontoxic and non-allergens. Regions of the vaccine constructs were predicted to have sufficient B cell epitopes and capable of generating interferon gamma response and have 100% sequence coverage (Table 2 and Figure 3 ). 

The vaccine constructs thus designed were made through comparative modeling over ROBETTA and revealed a structure better than the other platforms utilized in this regard Figure 4 . E  p  i  t  o  p  e  s  H  e  l  p  e  r  T  C  e  l  l  E  p  i  t  o  p  e  s  C  y  t  o  t  o  x  i  c  T  C  e  l  l  E  p  i  t  o  p  e  s   D  P  K constructs were refined through GALAXY server (Table 3 ) and validated through Ramachandran Plots and Z-score from ProSA-webserver (Table 4C) . 

Attempts were made to assess the stability of the designed constructs before moving on to docking and simulation studies with TLRs. Since it comprises three different adjuvants, an ensemble of different sequences interspersed across the spike glycoprotein, it becomes imperative that the constructs be evaluated. In this endeavor, we utilized the Expasy ProtParam Tool and based our assumptions on the PEST hypothesis and the Instability Index determinants, which include analysis of the residues that make any biomolecule have lower half-life or increased instability ( includes Proline(P), Glutamic acid (E), Serine (S), Threonine (T), Methionine (M) and Glutamine (Q) ) in the designed constructs and including Guruprasad's basis [51] that the relative abundance of Asparagine (N), Lysine (K) and Glycine (G) has been found in stable proteins (Table 4B) . Moreover, basic or neutral isoelectric points also indicate a stable in vivo half-life for the modeled constructs. Since we cannot assess it entirely on these assumptions, we investigated the stability of the constructs through a 100 nanoseconds Molecular Dynamics Simulation. But due to constraints of computational time and effort we assessed the stability of Construct_4 since it gave the best estimate of stability based on our assumptions as listed in Table 4 (A,B,C) . The abovementioned parameters went into this assumption coupled with the Ramachandran Plot, ERRAT-3D and the Z-score. 

During the initial phase of MD simulation, major conformational changes in the vaccine construct were observed, as illustrated in Figure 5A . During the rest of the simulation time average Root Mean Square Deviation (RMSD) of the C-alpha atoms remained constant around 11.5 Å ( Figure 5A ). This suggests that after the initial structural rearrangements, the vaccine construct achieved stable three-dimensional structure, as evident in Supplementary Figure 2 .

Followed by the initial structural rearrangements, the vaccine construct achieves a stable three dimensional structure, as evidenced by the superimposed structures obtained at 0 ns and 10 ns in Figure 5D , followed by another at 10 ns and 100 ns in Figure 5E . Further, to identify the regions of the vaccine construct that contribute to structural rearrangements, Root Mean Square Fluctuations (RMSF) of C-alpha atoms were calculated at 0-10 ns and another from 10-100 ns in Figures 5B and 5C , respectively. The simulation indicates that Construct_4 conforms to the stability directives that are expected of it and can be reproduced under experimental conditions. Considering the MD, we may definitely extrapolate our results to include Constructs 2 and 5 as suitable vaccine constructs that can be considered in the development of vaccines since they scored on a similar index across the parameters on which we assessed and went ahead with the simulation study of Construct_4.

Based on the MD simulation and the observed shift in the modeled vaccine construct 4, molecular docking was carried out between TLR4 receptor and the vaccine construct 4 through Patchdock, which carries out docking of rigid molecules based on a protocol that employs molecular shape representation, surface patch mapping followed by filtering and scoring. The obtained results were refined and ranked using Firedock based on a global energy value (in kJ/mol) that helps determine the binding affinity of the molecules being considered. The results indicate that the post simulation Construct_4 does bind TLR-4 with high binding affinity (Supplementary Table 2 ) with a Global Energy Value (which is an analog of ranking based on binding affinity) of -26.36 kJ/mol. The interface between Construct_4 and TLR4 was analyzed post-docking through UCSF Chimera and DimPlot. These mainly include the non-covalent interactions that are determined through distance constraints between the two docked molecules and involve hydrogen bonds and hydrophobic interactions across the interface. Each of the residues have been specified that contribute to these interactions has been visualized as spheres and depicted on the basis of the corresponding non-covalent interaction across the proteinprotein interface in Figure 6 and in Supplementary Figure 3 . A similar docking was carried out between the modeled vaccine construct 4 (pre-MD simulation) with TLR-4 with a higher binding affinity value (-36.05 kJ/mol) and the residues implicated in both the dockings were found to be different as is represented in Supplementary Figure 4 The overall binding free energy of the TLR-4 and vaccine construct was predicted to be -16.91 kcal/mol (Supplementary Table 3 ).

One of the most potent options that are being explored to curtail the spread of SARS-CoV-2

includes the design and development of appropriate vaccines. The entire process of vaccine development involves an extensive timeline ranging from experimental to clinical settings. In recent times, advancement in molecular immunology and development of several epitope mapping methods, several semi-empirical approaches at vaccine design have been introduced wherein, multi-epitope-peptide based subunit vaccines based on predicted B and T cell epitopes utilizing similar bioinformatics tools has made remarkable strides. Despite the cons of low immunogenicity and multiple doses, the trade-off with eliciting neutralizing antibodies, humoral immune response and relative safety when associated with attenuated or inactivated virus vaccines plays in favor of these approaches [77] .

Looking into the probable limitations, strategies have been employed to design an epitope based peptide vaccine utilizing the major subunits of the spike glycoprotein of SARS-CoV-2. In selecting the concerned immunogen, S1 and the S2 domains of the surface glycoprotein have been utilized separately, keeping in mind the limitations associated with utilizing the entire surface glycoprotein [63, 78] . The prediction of the three different types of epitopes ( B cells, T HTL and T CTL ) by utilizing multiple servers in each case based on different algorithms and selection based on a consensus involving all the servers, including experimental basis behind alleles (HLA Class I and Class II), helps add sustenance to the concluded list of epitopes [64] [65] [66] [67] [68] .

The predicted epitopes were found to be experimentally identified across a recent microarray study of mapped epitopes on the surface glycoprotein [79] . Each of the predicted epitopes was verified based on described parameters and were found to be antigenic, non-allergenic, non-toxic and share high coverage across the spike glycoprotein of SARS-CoV-2 sequences [37] . The epitope sequences were also matched with a recently deposited SARS-CoV-2 sequence from The adjuvant and the predicted epitope sequences were appropriately joined by rigid linkers based on their arrangement. Even the chimeric adjuvant sequences were linked by short rigid linkers (XP) n to aid in ensuring stability. GPGPG and AAY linkers were used as intra HTL and CTL linkers, respectively. Separation between the vaccine components has been arrived upon by the role of EAAAK as a linker between adjuvant and the peptide epitope sequences [43, 44] . The designed vaccine constructs based on the arrangement of the chimeric adjuvant were validated and concluded to be highly antigenic, non-allergenic and non-toxic. This makes up for the limitations of subunit vaccines as mentioned above.

The stability of the vaccine constructs was assessed based on physico-chemical parameters that make it suitable for purification across experimental settings. The presence of cysteine bonds (without introduction of any engineered mutations) points to the stability of the modeled constructs. The presence of defined secondary structure characteristics across the modeled construct despite the presence of mainly strand and loop regions due to predicted epitopes has led us to verify the stability dynamics through a 100ns MD simulation run. Moreover, the assessment of stability through amino acid composition [51, 52] , correlation with in vivo half-life periods and calculated isoelectric points [53] helps us identify three of the five modeled vaccine constructs to exhibit similar stability parameters. Based on the MD simulation of Construct 4, it can be extrapolated to include Construct 2 and Construct 5. The utilization of the ProSA web server [49] , SAVES server for Z-score and Ramchandran plot [48] , ERRAT-3D [50] scores, respectively helps validate the modeled constructs.

A rigid based docking of TLR-4 and Construct 4 was carried out after this and although there are no experimental data to insinuate any such interaction but the structural features of the vaccine construct and the functional characteristics of TLR-4, an accepted dock model can be achieved in this scenario. Moreover, the N terminal is composed of TLR-4 agonists which helps compound the interaction between the two proteins. Based on TLR-4 PDB structures [57] and their corresponding binding sites followed by Ellipro [21] and Castp [59] predictions of binding sites over the receptor, docking was carried out with the vaccine construct over PatchDock [60] The MD simulations of the docked pose were not carried out because of computational constraints. Moreover, it becomes imperative that since no experimental data is evidenced to verify the docking of the TLR4 and the vaccine construct, an actual determination of the proteinprotein interaction between them may be carried out by running extensive MD simulations involving both the proteins (tempered binding) and determining their association and dissociation profiles alongside residence time to assess the best docking pose of the two proteins in their energy minimized conditions circumventing the limitations associated with unavailability of experimental evidence to a certain extent. Therefore, the study does not include MD simulation of the docked pose since that would have made our results biased.

Authors declare no conflict of interest. 

Learning from the past: possible urgent prevention and treatment options for severe acute respiratory infections caused by 2019-nCoV

Receptor recognition by the novel coronavirus from Wuhan: an analysis based on decade-long structural studies of SARS coronavirus

A new coronavirus associated with human respiratory disease in China

Genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding

Structure, function, and antigenicity of the SARS-CoV-2 spike glycoprotein

Cryo-EM structure of the 2019-nCoV spike in the prefusion conformation

Structural basis of receptor recognition by SARS-CoV-2

Structural basis for the recognition of SARS-CoV-2 by full-length human ACE2

Therapeutic options for the 2019 novel coronavirus (2019-nCoV)

A SARS-CoV-2-Human Protein-Protein Interaction Map Reveals Drug Targets and Potential Drug-Repurposing

SARS-CoV-2 cell entry depends on ACE2 and TMPRSS2 and is blocked by a clinically proven protease inhibitor

UCSF Chimera-a visualization system for exploratory research and analysis

Prediction of continuous B-cell epitopes in an antigen using recurrent neural network

BcePred: prediction of continuous B-cell epitopes in antigenic sequences using physico-chemical properties

Induction of hepatitis A virus-neutralizing antibody by a virus-specific synthetic peptide

Prediction of the secondary structure of proteins from their amino acid sequence

Prediction of chain flexibility in proteins

A semi-empirical method for prediction of antigenic determinants on protein antigens

New hydrophilicity scale derived from high-performance liquid chromatography peptide retention data: correlation of predicted surface residues with antigenicity and X-ray-derived accessible sites

BepiPred-2.0: improving sequencebased B-cell epitope prediction using conformational epitopes

ElliPro: a new structure-based tool for the prediction of antibody epitopes

Reliable B cell epitope predictions: impacts of method development and improved benchmarking

Quantitative online prediction of peptide binding to the major histocompatibility complex

SYFPEITHI: database for MHC ligands and peptide motifs

Improved methods for predicting peptide binding affinity to MHC class II molecules

Peptide binding predictions for HLA DR, DP and DQ molecules

TepiTool: a pipeline for computational prediction of T cell epitope candidates

Gapped sequence alignment using artificial neural networks: application to the MHC class I system

MHC-NP: predicting peptides naturally processed by the MHC

Large-scale validation of methods for cytotoxic T-lymphocyte epitope prediction

A consensus epitope prediction approach identifies the breadth of murine T CD8+-cell responses to vaccinia virus

VaxiJen: a server for prediction of protective antigens, tumour antigens and subunit vaccines

AllerTOP v. 2-a server for in silico prediction of allergens

AlgPred: prediction of allergenic proteins and mapping of IgE epitopes

Designing of interferon-gamma inducing MHC class-II binders

Full-genome sequences of the first two SARS-CoV-2 viruses from India

Vaccine adjuvants: Understanding the structure and mechanism of adjuvanticity

Linked Toll-Like Receptor Triagonists Stimulate Distinct, Combination-Dependent Innate Immune Responses

Cutting edge: New chimeric NOD2/TLR2 adjuvant drastically increases vaccine immunogenicity

Receptor-ligand based molecular interaction to discover adjuvant for immune cell TLRs to develop nextgeneration vaccine

Understanding SARS-CoV-2-mediated inflammatory responses: from mechanisms to potential therapeutic tools

Fusion protein linkers: property, design and functionality

Construction of a linker library with widely controllable flexibility for fusion protein design

Protein identification and analysis tools on the ExPASy server

High-resolution comparative modeling with RosettaCM

Prediction of protein structure and interaction by GALAXY protein modeling programs

PROCHECK: a program to check the stereochemical quality of protein structures

ProSA-web: interactive web service for the recognition of errors in three-dimensional structures of proteins

Verification of protein structures: patterns of nonbonded atomic interactions

Correlation between stability of a protein and its dipeptide composition: a novel approach for predicting in vivo stability of a protein from its primary sequence

PEST sequences and regulation by proteolysis

Relationship between in vivo degradative rates and isoelectric points of proteins

Errors in protein structures

Scalable algorithms for molecular dynamics simulations on commodity clusters, SC'06: Proceedings of the 2006 ACM/IEEE Conference on Supercomputing

VMD: visual molecular dynamics

The structural basis of lipopolysaccharide recognition by the TLR4-MD-2 complex

Pathogen recognition and innate immunity

CASTp 3.0: computed atlas of surface topography of proteins

PatchDock and SymmDock: servers for rigid and symmetric docking

FireDock: a web server for fast interaction refinement in molecular docking

Assessing the performance of the MM/PBSA and MM/GBSA methods. 6. Capability to predict protein-protein binding free energies and re-rank binding poses generated by protein-protein docking

SARS CoV subunit vaccine: antibodymediated neutralisation and enhancement

A Sequence Homology and Bioinformatic Approach Can Predict Candidate Targets for Immune Responses to SARS-CoV-2

Comprehensive analysis of dengue virus-specific responses supports an HLA-linked protective role for CD8+ T cells

HLA loci and respiratory infectious diseases

Association of human-leukocyte-antigen class I (B* 0703) and class II (DRB1* 0301) genotypes with susceptibility and resistance to the development of severe acute respiratory syndrome

Functional classification of class II human leukocyte antigen (HLA) molecules reveals seven different supertypes and a surprising degree of repertoire sharing across supertypes

Synthetic Toll like receptor-4 (TLR-4) agonist peptides as a novel class of adjuvants

A TLR4-derived non-cytotoxic, self-assembling peptide functions as a vaccine adjuvant in mice

A truncated Cterminal fragment of Mycobacterium tuberculosis HSP70 enhances cell-mediated immune response and longevity of the total IgG to influenza A virus M2e protein in mice

The science of vaccine adjuvants: advances in TLR4 ligand adjuvants

Identification of stimulating and inhibitory epitopes within the heat shock protein 70 molecule that modulate cytokine production and maturation of dendritic cells

A 14-mer Hsp70 peptide stimulates natural killer (NK) cell activity

Biotechnology approaches to produce potent, self-adjuvanting antigenadjuvant fusion protein subunit vaccines

Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features

The outbreak of SARS-CoV-2 pneumonia calls for viral vaccines, npj Vaccines

Antibodies against trimeric S glycoprotein protect hamsters against SARS-CoV challenge despite their capacity to mediate FcγRII-dependent entry into B cells in vitro

SARS-CoV-2 proteome microarray for mapping COVID-19 antibody interactions at amino acid resolution, bioRxiv

 Supplementary Table 1B : The 2 nd , 6 th , 7 th and 9 th construct were found to be the most antigenic ( a stringent Vaxijen score of =>0.9 whereas threshold lies at 0.4) and were utilized to design the vaccine constructs Supplementary C+A+B NA 0.9 No **NA refers to a non-allergen and all adjuvants were predicted to be non-toxic in nature