key: cord-1005448-2a2z836v
authors: Baig, Abdul Mannan; Khaleeq, Areeba; Hira, Syeda
title: Elucidation of Cellular Targets and Exploitation of the Receptor Binding Domain of SARS‐CoV‐2 for vaccine and monoclonal antibody synthesis
date: 2020-06-23
journal: J Med Virol
DOI: 10.1002/jmv.26212
sha: a26f52bcf510ee82685cf0c9c7217ca3b1c2fe19
doc_id: 1005448
cord_uid: 2a2z836v

The pandemic caused by novel severe acute respiratory syndrome coronavirus (SARS‐CoV‐2) has resulted in over 452,822 deaths in the first twenty days of June 2020 due to the coronavirus virus disease 2019 (COVID‐19). The SARS‐CoV‐2 uses the host angiotensin‐converting enzyme 2 (ACE2) receptor to gain entry inside the human cells where it replicates by using the cell protein synthesis mechanisms. The knowledge of the tissue distribution of ACE2 in human organs is therefore important to predict the clinical course of the COVID‐19. Also important is the understanding of the viral receptor‐binding domain (RBD) a region within the spike (S) proteins that enable the entry of the virus into the host cells to synthesize vaccine and monoclonal antibodies (mAbs). We performed an exhaustive search of human protein databases to establish the tissues that express ACE2 and performed an in‐depth analysis like sequence alignments and homology modeling of the spike protein (S) of the SARS‐CoV‐2 to identify antigenic regions in the RBD that can be exploited to synthesize vaccine and mAbs. Our results show that ACE2 is widely expressed in human organs that may explain the pulmonary, systemic, and neurological deficits seen in COVID‐19 patients. We show that though the S protein of the SARS‐CoV‐2 is a homolog of S protein of SARS‐CoV‐1, it has regions of dissimilarities in the RBD and transmembrane segments. We show peptide sequences in the RBD of SARS‐CoV‐2 that can bind to the MHC alleles and serve as effective epitopes for vaccine and mAbs synthesis. This article is protected by copyright. All rights reserved.

taxonomically related to the betacoronaviridae group of the viral pathogens 9, 10 , the SARS-CoV-2 shares a significant similarity with other members of the coronaviridae group that are known to have caused similar if not identical cross-species viral diseases in humans in the past 5, 6, 7 . Examples of the latter include SARS-CoV- 1 (2002 to 2004) and MERS that affected the human population in 2012 14, 15 . Additionally, very little is known about the diversity of the expression of ACE2 in human organs and tissues. An investigation into the quantitative and qualitative distribution of ACE 2 in different tissues is needed that can predict the possible organs involved and hint towards the expected outcome of localized and systemic forms of COVID-19 presenting in the clinics. The COVID-19 genome (NCBI Reference Sequence: NC_004718.3) is known to encode several proteins, of which the surface protein (S) (NCBI Reference Sequence: NP_828851.1) is known to be essential to dock the virus to ACE2 receptor on host cells 6, 7, 8 . It was inferred that as both, SARS-nCoV-1 and SARS-CoV-2, taxonomically belong to the betacoronavirus 5, 6 they would have a similar if not identical composition of the S protein and the RBD that helps to dock the SARS-CoV-2 on to the ACE2 receptors 6, 8 . It is important to mention here that although we know the target molecular receptor that is needed for the virus to gain entry into human cells, the RBD of SARS-CoV-2 is very similar but not identical to the RBD of SARS-CoV-1 6 . The functional significance of the regions of disparity between the S protein of SARS-CoV-2 and SARS-CoV-1 remains to be determined and exploited for therapeutic gains. The structural model of SARS-CoV-1 is already studied in detail for the binding of neutralizing antibodies and therefore identification of the differences between the S proteins of both the SARS-CoV viruses can be of help in designing more specific neutralizing mAbs against SARS-CoV-2.

We at first focused on drawing an organ-based distribution of ACE2 receptor followed by an in-depth analysis of the S protein of SARS-CoV-2 and its RBD for identification of short sequences that can be used for vaccine development and synthesis of mAbs against SARS-CoV-2. By performing sequence alignments and homology modeling the understating of the RBD of SARS-CoV-2, it is expected that we will gain an in-depth understanding of the pathogenesis of the COVID-19 and assemble effective ways to counter the mode of infection of SARS-CoV-2.

Finally, we performed in silico methods by using automated epitope detecting databases that extract sequences of variable length with the potential of binding to class I and class II major histocompatibility complex (MHC) alleles on antigen-presenting cells and therefore capable of evoking a B/T-killer cell immune response. For the epitopes generation, the S protein and the RBD of SARS-CoV-2 were selected and the epitopes that had shown a strong binding (SB) prediction with MHC alleles were proposed for vaccination and mAbs synthesis. Online databases that are an integrative web resource on human and other mammalian tissue expressions 13, 14 were used to identify the ACE2 encoding mRNA. The tissue associations in these databases for ACE2 were retrieved from manually curated knowledge in UniProtKB ( https://www.uniprot.org/uniprot/?query=ACE2&sort=score) This article is protected by copyright. All rights reserved. 15 and via automatic text mining of the biomedical works of literature. The confidence of each association is signified by signs, where +++++ is the highest confidence and +++, ++, + as the lowest. Human tissues and Protein Atlas (HPA) 14 (www.proteinatlas.org/ENSG00000130234-ACE2/tissue) were used to retrieve the mRNA expression levels encoding ACE2 in forms of histograms. Both the servers have a user-friendly interface wherein protein (ACE2) and other protein molecules can be queried which results in a list of organs expressing the protein in question.

Online databases that are a huge resource of data of expression of human proteins in a healthy and diseased state like HPA was used to retrieve and study the expression levels of ACE2 in normal body tissues as compared to diseased states. Immunohistochemistry (IHC) of ACE 2 stained tissue was investigated for mild, moderate, and strong staining.

The genome of COVID-19 was obtained from the NCBI database 10, 16 and the proteins encoded by this virus were retrieved. NCBI automated server (https://www.ncbi.nlm.nih.gov/tools/msaviewer/) was used for multiple sequence alignment (MSA) and building an evolutionary tree of the genome of viruses and the proteins that they encode. Uniprot database 15 was also used for MSA and UniprotKB and NCBI were used to develop the coronavirus phylogenetic tree of SARS-CoV-2. For comparisons, the S protein sequence of both the viruses was retrieved from This article is protected by copyright. All rights reserved.

the NCBI database. Pfam and NCBI database 16 Bioinformatics computational tools were used to search for protein homologs of the RBD of S protein of SARS-CoV-1 and SARS-CoV-2. BLASTn and BLASTp tools were used to uncover nucleotide and protein homologs in between SARS-CoV and SARS-CoV-2 respectively. BLASTn was selected for nucleotide similarity searches and BLASTp was selected for the determination of S protein and RBD homologs. The align tool available in the Uniprot server 15 was used for MSA of different protein sequences of SARS-CoV-1 and SARS-CoV-2 in general and S protein and its RBD in particular. In sequence alignments, particular attention was directed towards identifying the mutations in amino acid residues of the RBD of SARS-CoV-2 that are known to engage with ACE2 receptor during the process of host-virus interaction and initiating viral entry into the cells. During comparisons of S protein and RBD, we specifically focused on spotting the occurrence of mutations in the transmembrane regions and motif segments of the S protein of the SARS-CoV-2 as compared to SARS-CoV-1.

Sequences of the SARS-CoV-2 S protein and RBD were submitted to the automated SWISS-MODEL database 17 to develop template-based models. Similarly, sequences of the SARS-CoV-1 S protein and RBD were submitted to the automated SWISS-MODEL This article is protected by copyright. All rights reserved.

database for the development of template-based models so that the differences and similarities between the proteins of the viruses can be identified. Previous models of SARS-CoV-1 S protein and RBD were downloaded and compared with a similar region in the SARS-CoV-2 virus. B 1.6-Identification of Epitopes predicted to bind the MHC -Class I and II allele.

Though the SARS-CoV-2 is declared to be a novel variant of betacoronavirus, its S protein has homologs (evolutionarily related) to the taxon of beta-coronaviruses. The Immune Epitope Database and Analysis Resource (IEDB) 18 and DTU bioinformatic servers 19, 20 were used for epitope mapping in the S protein RBD sequence in SARS-CoV-2. Filters were applied to search within betacoronavirus, SARS, and MERS members of the taxa only. Other filters that were selected included epitope type, host, assays type, MHC restrictions, and infectious disease 19 . Also, the amino acid sequence of S protein and RBD of SARS-CoV-1 and SARS-CoV-2 was used to generate epitopes of sequence lengths between 5mer to 9mer where a peptide-MHC class I binding could occur for provoking antibody production in vivo. The NetMHC-4.0 automated server uses a proteome scan in search of potential T cell epitopes therefore peptides of optimal length for the alleles of interest are inherently prioritized. The user inserts a sequence, as in our case the S-protein FASTA sequence, followed by the selection of peptide length of (8mer-14mer) range, specie, and HLA Class type, and subtype selection. There are options to sort by predicted affinity and threshold for strong binders: % Rank is set to 0.5 by default. Output can be retrieved in XLS format or simple text. NetMHC-4.0 ranks the predicted affinity by comparing it to a set of 400.000 random natural peptides. This measure is not affected by the inherent bias of certain molecules towards higher or lower mean predicted affinities. Strong binders are defined as having % rank <0.5 and weak binders with % rank < 2. The selection of the candidate binders was based on % Rank rather than nM Affinity which is predicted by NetMHC-4.0. A parameter of BindLevel (SB: strong binder, WB: weak binder) is also attributed to the sequence predicted as an epitope. The peptide will be identified as a strong binder if the % Rank is below the specified threshold for the strong binders, by This article is protected by copyright. All rights reserved. IEDB is funded by NIAID and is a freely available resource for users worldwide. It catalogs experimental data on antibody and T cell epitopes studied in the context of infectious disease, allergy, autoimmunity, and transplantation. The IEDB also hosts tools to assist in the prediction and analysis of epitopes. We used the IEDB for the prediction of epitopes that could interact with the MHC class II allele on antigen-presenting cells

and T-lymphocytes. The database is interactive and user friendly as the user has to choose one of the radio buttons to select protein sequence (RBD of SARS-CoV-2) that contains the MHC class II epitopes for which predictions get generated. The sequence will be transferred to the MHC class II binding predictions when clicking the "Submit" button. These test-datasets are meant to demonstrate the functionality of the tools and are by no means considered equivalent to a formal performance evaluation. For the prediction of MHC Class II epitopes, it has a consensus approach that combines NNalign, SMM-align, and combinatorial library methods. An automated server from Immune Epitope Database Analysis Resource 18 was selected and predictions were generated for RBD of SARS-CoV-1 and SARS-CoV-2 and compared. The automated IEDB interface shows options like species/locus selection, prediction of linear B cell or T cell epitopes, choice of prediction methods, selection of MHC allele(s), select option for length(s), and output formats as XML or text. Details of the use and options can be found at the automated webserver site at https://www.iedb.org/ This article is protected by copyright. All rights reserved.

Results: C 1-ACE 2 is expressed in diverse organs with a predominance in Lungs, Endothelium, Heart muscle, Kidney, Gastrointestinal tract, Testicular tissue, Brain, and Adrenal glands.

The mRNA encoding ACE2 dominates in the endothelial cell, lungs, heart muscle, kidney, gastrointestinal tract, testicular tissue, adrenal glands, and nervous system when compared to the rest of the organs and tissues in the human body. The ACE2 (Fig.1A) protein is expressed in different organs and tissues ( Fig.1 B, C) with its structure (Fig.1 D) deposited in a protein database that can be a target of the SARS-CoV-2. The data showed ( Fig.1 B) was derived from the data retrieved from HPA, tissue expression database, and UniProtKB as detailed in methodology. The representative immunohistochemistry staining of ACE2 in different tissues (Fig. 2) was retrieved from (https://www.proteinatlas.org/ENSG00000130234-ACE2/tissue). The retrieved images show the differential staining of ACE2 in foregut derivatives like intestine (duodenum, ileum, colon, and rectum) and gall bladder which showed stain with maximum intensity (Fig.2) . Kidney and testicular cells and tubules also showed a high expression of ACE2 receptors. Cardiac muscle and adrenal glands stained with moderate intensity in the retrieved images (Fig.2 histogram and image) .

The BLASTn results showed that the genome of SARS-CoV-2 (GenBank ID QHD34416.1) had 100% sequence identities with related SARS-CoV-2 clinical isolates reported from other countries (with few exceptions -see below) and shared 82% -99% sequence identities with SARS-CoV-1 genome. The NBCI automated server 10, 16 was used to perform an MSA and build an evolutionary tree for the top matches with SARS-CoV-1 and the Bat coronaviruses (Fig.3A ). The phylogenetic cladogram shows the COVID-19 Wuhan CoV-Hu-1 A and SARS-CoV-1 share a common ancestor (Fig.3B- black arrow). The BLASTn search with coronaviruses selected in the NCBI server showed (Fig.3 B) the genomes of a diverse SARS-CoV-1 and Bat coronavirus (Fig.3 A black -star) as homologs which justifies it to be named as SARS-CoV-2 based on its taxonomy. With the recently deposited SARS-CoV-2 genomes from different clinical isolates from around the world, the occurrence of mutations in the genome of SARS-CoV-2 has been noticed, some of which have sequences identities of 99.97% (Fig 3C - top node and blue stars) and the 00.03% difference could be of pathogenic significance. 

between the 306 th -527 th amino acids (aa) is located within the S protein sequence of the protein length of 1255 aa. While the RBD (319 th -541 st aa) of S protein in SARS-CoV-2, which is composed of 1273 aa has now been elucidated (21, 22) . We aligned the RBD (319 st -541 st aa) of SARS-CoV-2 (YP_009724390.1) with the RBD (306 th -527 th aa) of SARS-CoV-1 (NP_82885.1.1) (Fig.4 A) and pairwise alignments showed 73.09% sequence identities. Interestingly, we found that the RBD of SARS-CoV-2 had mutations at five (Fig. 4A red-stars) of the six amino acids that are needed to engage with the human ACE2 receptors, as has also been reported by others (23).

The SARS-CoV-2 RBD (319 th -541 st aa) was subjected to template-based model developments. The FASTA sequence of the above amino acids of RBD of SARS-CoV-2 was submitted to the SWISS-MODEL automated server 17 The entire sequence of S protein of SARS-CoV-2 (1273 aa) (YP_009724390.1) was submitted to the healthtech dtu dk server 18, 19, 

This article is protected by copyright. All rights reserved.

Analysis Resource -IEDB 20 for the generation of epitopes. The results from the healthtech dtu dk server fetched 4 epitopes within the RBD (319-541 aa) with strong binding (SB) prediction to the HLA-A and HLA-B class I allele (Fig.6 A-stars) . The IEDB automated server was used to fetch sequences in the RBD of SARS-CoV-2 for predicted binding to the MHC-class II allele. Shown are the sequences that fall between 319-541 aa of the RBD of SARS-CoV-2 as a result (Fig. 7 A and B) . The RBD regions of the SARS-CoV-2 that contain these sequences are shown on the model of SARS-CoV-2 spike protein (Fig. 7 1 and 2) .

The knowledge of the pathogenesis of the SARS-CoV-2 and the understanding of the diversity of its clinical presentation is still sketchy and much remains to be elucidated 24, 25 . We know very little about the translational significance of the genomic, transcriptomic, and proteomic discoveries on SARS-CoV-2 that have emerged recently.

Though we know the tissue distribution of the ACE2, a molecular target of RBD of SARS-CoV-2 7 , the cascade of host and virus protein interaction in human organs and tissue remains obscure therefore an understanding of the pathogenesis of SARS-CoV-2 is urgently needed 25 . It is noteworthy that though the SARS-CoV-2 uses the same ACE2 receptor for human cell infection in various organs and tissues that express this receptor subtype (Fig.1, 2) , the RBD of SARS-CoV-2 is different when compared to SARS-CoV- showed the origin of SARS-CoV-2 from the linage of beta coronaviruses (Fig.3 A, B) 23, 27 , and its strong association with different clinical isolates that have been identified during the recent pandemic (Fig.3 C) . Regarding the mutations occurring in proteins encoded by SARS-CoV-2, we show that the in clinical isolates from the US mutations have emerged (Fig.3 C -blue stars) as compared to the original isolate in a COVID-19 patient (GenBank ID YP_009724390.1) identified and archived in NCBI 27 (Fig. 3 C red star). Others have reported mutations involving RNA dependent RNA polymerase (RdRp) 28 .

The significance of the mutations in RBD of the S protein of SARS-CoV-2 as compared to SARS-CoV-1 has just begun to be elucidated with the insight that the mutations in the RBD of S protein of SARS-CoV-2 has increased its affinity for the ACE2 by several folds 7 . In contrast, different ACE2 alleles could reduce the affinity of RBD of SARS-CoV-2 to infect human cells as mentioned above. We like the other 23 report mutations in five out of six key amino acids of the RBD of SARS-CoV-2 compared to SARS-CoV-1, which are cardinal for engaging the ACE2 receptors ( Fig.4 A red-stars). Sequence alignments showed other regions of mutagenesis in SARS-CoV-2 in regions like transmembrane, while the motif has shown to be remained conserved (Fig.4 B, arrows, and brown alphabets). The next step in our study was to explore the structural details of the S protein and its RBD in SARS-CoV-2 to see how it differs from SARS-CoV-1 which could help uncover the rate of transmission and infectivity of SARS-CoV-2 in human cells. As the structural details of the S protein and RBD of SARS-CoV-2 got resolved and deposited early after the outbreak of COVID-19 7, 21, 22 , we performed homology modeling of the structures of the RBD of SARS-CoV-1 and SARS-CoV-2 to compare them for homology and compute their interaction with ACE2. The RBD sequence of SARS-CoV-2 and SARS-CoV-1 were subjected to homology modeling (Fig.5) . The RBD of SARS-CoV-2 developed a template-based model of SARS-CoV-1 ( Fig.5 A) and the RBD of SARS-CoV-1 developed a template-based model of SARS-CoV-2 RBD in complex with human ACE2 (Fig.5 B) which showed them to be homologs (evolutionarily related) reinforcing the taxonomical relation between the two viruses. We show that in contrast to the A chain, the B and C chains of the template (5x5b.1) developed for SARS-CoV-2 RBD (319 th -541 st aa) have identical amino acids when compared with the model developed (Fig.5 C top row -Seqres) . After the identification of the RBD sequence of SARS-CoV-2 and their differences with SARS-CoV-1, we were curious to identify sequences in the RBD of SARS-CoV-2 that are capable of serving as epitopes to bind the HLA class I and class II allele in macrophages to mount an immune response against RBD of SARS-CoV-2 neutralizing its capability to infect human cells. As expected, any short or long sequence extracted from the S protein would not match any human protein as we checked (data not are shown). There is no certainty that sequences picked in this way would interact with the HLA class I and class II allele in macrophages to mount an immune response. The sequences that were identified to have the potential to bind the HLA class I (Fig-6 ) and class II allele (Fig-7) in macrophages for antigen presentation are shown. The DTU bioinformatics portal 19 , recognizes a minimal 5 to 9 amino acid length peptides with a binding core directly in contact with the MHC allele, predicted binding affinity is in nanomolar units, a rank of the predicted affinity ( Fig.6) with strong binders (SB-defined as having % rank <0.5) 19 .

The automated IEDB 20 server Bepipred-1.0 Linear Epitope Predictor and BepiPred-2.0: Sequential B-Cell Epitope Predictor identified the peptide sequence with the predictions of binding to the MHC-class II allele (Fig.7 ). An important observation that was needed Accepted Article to be taken into account was to spot an epitope(s) that would attack the ACE2 binding amino acid residues 23 in the RBD of the S protein of SARS-CoV-2. We show (Fig-7 A-B stars) the epitopes we present in this study can perhaps mount a T-cell response in particular by binding HLA class II allele(s), which was previously reported by using the IEDB server by us for SARS-CoV-2 in a preprint 29 and others studies on the Middle East Respiratory Syndrome Coronavirus (MERS-CoV) 30 . The mechanism of T-cell mediated cytotoxicity unlike IgG type humoral response is expected to cause a limited cytokine formation and therefore lesser tissue damage. We infer that the peptides we are reporting in this study have stronger MHC class I and II allele binding predictions with their affinity in nanomolar units that make them superior over other peptides that are proposed for vaccine development reported previously by us 29 . Also, reverse vaccinology and machine learnings have been applied to generate epitopes for a vaccine against SARS-CoV-2, SARS-CoV-1, and MERS-CoV 31 . It is obvious now that this is the right time to consider investing in vaccines against emerging viruses which if delayed could lead to loss of human lives 32 as has been seen in the ongoing pandemic. Though vaccines against other proteins and targets including the S protein have been detailed 31. 32 

The areas that need attention in our fight against COVID-19 are the differences in tissue distribution of ACE2 in different organs, the differential binding affinities of RBD of Sprotein to ACE2 in diverse ethnic populations 26 , the understanding of the RBD basis of This article is protected by copyright. All rights reserved.

the SARS-CoV-2 infection caused by mutations in this segment 7, 21, 22 and an urgent attempt to prepare vaccine 32 /mAbs to combat COVID-19 29 . It would be ideal if the mAbs generated against the epitopes could be produced in mass in vitro to be given by infusions, as this could lessen the need for the convalescent plasma of patients recovered from COVID-19 infections. Unraveling these pivotal components is likely to make us capable to fight the COVID-19. We implore to speed up the process of vaccine development and synthesis of mAbs against SARS-CoV-2 to fight the pandemic caused by COIVD-19, by testing the sequences submitted in this study and other similar reports 31, 32 . Also, much needed is a safe and efficacious antiviral agent that could eradicate COVID-19 with minimal adverse effects.

The author(s) declare that there are no conflicts of interest.

This paper includes no work on human and animals or their tissues. No trials are been reported, an ethical approval, therefore, was not needed for this work, Figure 2 . Immunocytochemistry data of quantitative expression, location, and staining intensity of ACE2 in human tissues. Protein expression overview of ACE2 showed it be expressed highly in intestines, gall bladder, kidney, and testicular tissue. HPA did not have neuronal tissue staining in its archive. [Data retrieved from The HPA database in accord with the policy of the database resource retrieval which enables the third parties to have access to the data shown www.proteinatlas.org/ENSG00000130234-ACE2/tissue] This article is protected by copyright. All rights reserved. Accepted Article Figure 6 . The peptides (3rd column) identified in the RBD (319-541 aa) region of the S of SARS-CoV-2 (stars) had strong binding (SB) predictions (7th column). The binding prediction to the MHC-class I allele (2nd column) with log-transformed binding affinity, nM affinity, and rank (4th, 5th, and 6th column respectively) are shown. The non-RBD resident sequences with SB and nM affinity are shown in the last four columns that were predicted to bind the MHC-class-I allele. Note the double stars (452-459 aa) of which 455 is a known residue in the RBD of S protein that engages with the ACE2 receptor 23 . Figure 7 . (A) BepiPred 1.0 prediction software in the automated IEDB server was used that predicted the location of linear B-cell epitopes using a combination of a hidden Markov model and a propensity scale method. Results are based on a large benchmark calculation containing close to 85 B cell epitopes. Note stars (494-511aa) which contain 501 and 505 amino acids that are known to engage with ACE2 receptor 23. (B) BepiPred-2.0: Sequential B-Cell Epitope Predictor automated serve was used to predict B-cell epitopes from RBD of the S of protein of SARS-CoV-2, using a Random Forest algorithm trained on epitopes and non-epitope amino acids that are determined from crystal structures. Note stars (457-492 aa) which contain 486, 501, and 505 amino acids that are known to engage with ACE2 receptor 23. The structure of SARS-CoV-2 RBD (PDB ID 6vsb.1) is shown (1) with the peptides predicted (A, B) to be located within the RBD (2-zoomed circle) of the template-based model of SARS-CoV-2.

Covid-19): situation summary

Preventing a COVID-19 pandemic

Responding to Covid-19 -A Once-in-a-Century Pandemic?

Covid-19: surge in cases in Italy and South Korea makes pandemic look more likely

A new coronavirus associated with human respiratory disease in China

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

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

Structure analysis of the receptor binding of 2019-nCoV

I-TASSER: a unified platform for automated protein structure and function prediction

Modeling protein quaternary structure of homo-and hetero-oligomers beyond binary interactions by homology

The Protein Data Bank Nucleic Acids Research

Lars Juhl Jensen, TISSUES 2.0: an integrative web resource on mammalian tissue expression

Tissue-based map of the human proteome

UniProt: a worldwide hub of protein knowledge

National Library of Medicine This article is protected by copyright. All rights reserved

SWISS-MODEL: homology modelling of protein structures and complexes

Immune epitope database analysis resource

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

Reliable prediction of T-cell epitopes using neural networks with novel sequence representations

Structure of the SARS-CoV-2 spike receptor-binding domain bound to the ACE2 receptor

Structural basis of receptor recognition by SARS-CoV-2

The proximal origin of SARS-CoV-2

An update on the status of COVID-19: a comprehensive review

Structural variations in human ACE2 may influence its binding with SARS-CoV-2 spike protein

Emerging SARS-CoV-2 mutation hot spots include a novel RNA-dependent-RNA polymerase variant

The Devil in its Details: Unravelling the Epitopes in COVID-19 Virus Surface Glycoprotein with the potential for Vaccination and Antibody Synthesis

Epitope-based peptide vaccine design and target site depiction against Middle East Respiratory Syndrome Coronavirus: an immune-informatics study

COVID-19 coronavirus vaccine design using reverse vaccinology and machine learning Edison Ong

Vaccines: Status Report. Immunity

Emergence of RBD mutations in circulating SARS-CoV-2 strains enhancing the structural stability and human ACE2 receptor affinity of the spike protein Junxian Ou