key: cord-1008577-pcyi20fd
authors: Ngo, Son Tung
title: 501Y.V2 spike protein resists the neutralizing antibody in atomistic simulations
date: 2022-01-19
journal: Comput Biol Chem
DOI: 10.1016/j.compbiolchem.2022.107636
sha: ac3c62766b25b43bd9f75a897851e2f6482391ca
doc_id: 1008577
cord_uid: pcyi20fd

SARS-CoV-2 outbreaks worldwide caused COVID-19 pandemic, which is related to several million deaths. In particular, SARS-CoV-2 Spike (S) protein is a major biological target for COVID-19 vaccine design. Unfortunately, recent reports indicated that Spike (S) protein mutations can lead to antibody resistance. However, understanding the process is limited, especially at the atomic scale. The structural change of S protein and neutralizing antibody fragment (FAb) complexes was thus probed using molecular dynamics (MD) simulations. In particular, the backbone RMSD of the 501Y.V2 complex was significantly larger than that of the wild-type one implying a large structural change of the mutation system. Moreover, the mean of [Formula: see text] , CCS, and SASA are almost the same when compared two complexes, but the distributions of these values are absolutely different. Furthermore, the free energy landscape of the complexes was significantly changed when the 501Y.V2 variant was induced. The binding pose between S protein and FAb was thus altered. The FAb-binding affinity to S protein was thus reduced due to revealing over steered-MD (SMD) simulations. The observation is in good agreement with the respective experiment that the 501Y.V2 SARS-CoV-2 variant can escape from neutralizing antibody (NAb).

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) initially reported from Wuhan, China in December 2019 (WHO, 2021) . The virus rapidly spreads worldwide that caused the human coronavirus disease 2019 (COVID-19) pandemic worldwide (WHO, 2021) . Despite the huge efforts of the international community to limit the spread of SARS-CoV-2, more than 180 million people were infected within one and half years (Worldometrics, 2021) . The viral outbreak caused more than 3 million deaths and several global issues. The virus is a single-positive-strand ribonucleic acid (RNA) virus that viral sequence is similar to severe acute respiratory syndrome coronavirus (SARS-CoV) and Middle East respiratory syndrome coronavirus (MERS-CoV) (Barnes et al., 2020; de Wit et al., 2016) . SARS-CoV-2 genomes, which are contained from 26 to 32 kb in size, are encryption of more than 20 structural and non-structural proteins (Francés-Monerris et al., 2020) . These proteins arrange into four groups including spike, envelope, membrane, and nucleocapsid Schoeman and Fielding, 2019) . In particular, the viral S protein is used to bind to human angiotensin-converting-enzyme 2 (ACE2) , SARS-CoV-2 thus using the receptor to infect human cells . It should be noted that ACE2 is available in various tissues involving the human lung, heart, and liver . Therefore, S protein is a target for neutralizing by human antibodies (Barnes et al., 2020; Ju et al., 2020; Kim et al., 2021) and the protein interestingly is the biological target for the COVID-19 vaccine design (COVID-19 Vaccines, 2021) .

The SARS-CoV-2 S trimer is folded by three monomers (cf. Fig. 1 ) (Yang and Du, 2021) , in which a monomer consists of two subunits including S1 and S2. S1 subunit is the receptor binding region, which contains the receptor binding domain (RBD) and N-terminal domain (NTD) (Fig. 1) . RBD facilitates the binding of S protein to ACE2 . S1B is thus called the receptor binding domain (RBD). RBD can identify and bind to the ACE2 when it is in the "up" shape (Yang and Du, 2021) . During the binding process, the conformation of S2 subunit is changed, resulting in SARS-CoV-2 being able to fuse with the cell membrane and insert host cells (Barnes et al., 2020; Lan et al., 2020) . The S2 subunit is thus called the membrane fusion region (Yang and Du, 2021) .

NAbs, which can be obtained from the patient plasma or immunized laboratories (Barnes et al., 2020) , mainly target RBD. These NAbs can be coarsely arranged into four classes. Classes 1 and 2 antibodies bind to RBD epitopes overlapping with the ACE2-binding site (Barnes et al., 2020) . Directly competing with ACE2 is suggested as the neutralization mechanism of these antibodies. Immunoglobulin V-gen segment with heavy chain complementarity determining regions (CDRH) including CDRH1, CDRH2 and a short CDRH3 encodes antibodies in class 1, which are typically produced by SARS-CoV-2 infection (Chen et al., 2021) . Class 2 antibodies aim to epitopes of class 1 antibodies (Barnes et al., 2020; Wibmer et al., 2021) . Moreover, class 3 antibodies bind outside ACE2, and class 4 antibodies only aim to RBD in 'up' shape (Barnes et al., 2020) and cannot prevent ACE2. Besides that, NTD of S protein is also a common target for NAbs and other monoclonal antibodies . These antibodies directly bind to NTD showing a large potential in COVID-19 treatment (Wibmer et al., 2021) . The major targets of these antibodies are N1-loop (NTD N-terminus), N3-loop (supersite β-hairpin), and N5-loop (supersite loop) (Wibmer et al., 2021) .

Numerous viral variants have been recently reported that can be escaped from NAbs (Chen et al., 2021; Weisblum et al., 2020; Hoffmann et al., 2021 Hoffmann et al., , 2021 . The issues were raised because the mutations mostly appeared in the S protein sequence, especially in RBD. For example, in the South African (B.1.351 or 501Y.V2) variant, the mutations appeared in both NTD (L18F, D80A, D215G, Δ242-244, and R246I) and RBD (K417N, E484K, and N501Y) . The structural change under mutation effects probably decreases the binding affinity between S protein and NAbs (Wibmer et al., 2021; Hoffmann et al., 2021; Wang et al., 2021; Cele et al., 2021) as well as RDB and antibodies . The rapid transmission among communities probably generates more variants and some of them have higher toxicity, larger transmission rate, and are able to escape from NAbs, etc (CDC, 2021a (CDC, , 2021b WHO, 2020; Abdool Karim and de Oliveira, 2021; Reuters, 2021) . Decreasing NAbs efficacy is probably associated with the reduction of the Covid-19 vaccine efficacy (CDC, 2021b). Therefore, it is an emergency to understand the physical insight at the atomic level into the escape from NAbs of a new SARS-CoV-2 variant.

Studying conformations, interactions, and association/dissociation of protein-protein complexes are fundamental problems (Pan et al., 2019) . It should be noted that structures of many complexes persist difficult to experimentally evaluate (Lupardus et al., 2014; Shan et al., 2014) . Moreover, investigating the protein-protein interactions/associations needs influential experimental tools (Tang et al., 2006; Frisch et al., 2001) , but the obtained results are popularly indirect or limited. Directly determining physical insights into the protein-protein binding at an atomic level remains open problems (Pan et al., 2019) . Atomistic simulations emerge as appropriate methods for evaluating the structural changes and interactions between biomolecules (Ngo et al., 2019 . Therefore, in this work, a fragment of NAb bind to 501Y. V2/wildtype (WT) S proteins was revealed using molecular dynamics (MD) and steered-MD (SMD) simulations. Furthermore, the obtained results indicated that 501Y.V2 complex is less stable compared with the WT one. The FAb-binding affinity to S protein is significantly decreased when the South African variant was induced. In addition, it should be noted that although glycans play an important role in the modulation of the spike conformational dynamics (Casalino et al., 2020; Woo et al., 2020; Turoňová et al., 2020) , glycosylation of S protein was neglected to clarify the natural interaction between S protein + antibodies.

The three-dimensional structure of SARS-CoV-2 S protein and S2M11 NAb fragment was downloaded from the Protein Data Bank (PDB) with code 7K43 (Tortorici et al., 2020) . The complexed structure was obtained using electron microscopy method with a resolution of 2.60 Å. The South African SARS-CoV-2 variant adopts eight changes in the S protein. In particular, four substitutions and deletion in NTD appeared as L18F, D80A, D215G, Δ242-244, and R246I. RBD had three substitutions including K417N, E484K, and N501Y. In order to generate the structure of 501Y.V2 S protein, PyMOL mutagent tool (Schrödinger LLC, 2010) was thus employed to perform six substitutions including L18F, D80A, D215G, K417N, E484K, and N501Y. Residues L242 and A243 were also deleted using PyMOL. However, deletion Δ244 and substitution R246I were not available since the sequence 244-257 was missed in the 7K43 structure.

Structural change of S protein + FAb complexes was investigated by using atomistic simulations with the GROMACS version 5.1.5 (Abraham et al., 2015) . The protein, FAb, and neutralized ions were parameterized via the Amber99SB-iLDN force field (Aliev et al., 2014) , according to the previous works ) . Besides that, the complex was inserted into a water box, in which water molecules were parameterized via the TIP3P water model (Jorgensen et al., 1983) . In particular, the dodecahedron periodic boundary conditions (PBC) and rectangular PBC boxes were used in simulating the solvated complex with MD and SMD simulations, respectively. The size of the dodecahedron PBC box is 6311.26 nm 3 with a box vector of (20.74, 20.74, 20.74) nm and the size of the rectangular PBC box is 5397.98 nm 3 with a box vector of (15.59, 14.54, 23.80) nm. 10 and 1 Na + ions were added to neutralize the WT and 501Y.V2 S protein + FAb complexes, respectively. More details about systemic configurations were shown in Fig. 2.   Fig. 1 . The conformation of the S protein + FAb complex. Three monomers were highlighted by yellow, sky blue, and cyan colors (right). The FAb was noted by green color. The N-terminal domain (NTD) and RBD were denoted by orange and red colors, respectively.

MD simulations were performed using parameters, which were referred to the prior studies . However, the integral steps were attempted every 3 femtoseconds. A non-bonded pair between two atoms was available if the spacing was smaller than 0.9 nm. The electrostatic interaction was mimicked using the fast Particle-Mesh Ewald approach (Darden et al., 1993) . The van der Waals (vdW) interaction was calculated via the cut-off scheme.

Initially, the SARS-CoV-2 S protein + FAb in solution was minimized by using the steepest descent approach. The energy minimized complex was positionally restrained by NVT and NPT simulations in a short interval (100 ps each). During relaxed simulations, the integral steps were performed every 1 femtosecond. The equilibrium conformations, which were obtained via NPT simulations, were used as initial shapes of MD simulations. The MD simulations were carried out 4 independent times with a length of 100 ns.

MD-refined structures of S protein + FAb complexes, which were obtained via free energy landscape and clustering analyses, were used as starting conformations of SMD simulations. The systems were reinserted into a rectangular PBC box (Fig. 2) , which size was reported above, to reduce the computing times. During SMD simulations, the FAb was dissociated from S protein under the effects of a harmonic pulling force as mentioned in Fig. 2 . The pulling force was applied on the FAb center of mass with an external force using cantilever k = 1000 kJ mol -1 nm -2 and constant velocity v = 0.001 nm ps -1 referring to the recent work . It should be noted that the C α atoms were weakly constrained via a harmonic force. The data was recorded every 33 integral steps.

The MD-refined structures, as well as representative structures, were obtained using a combination of the free energy landscape (FEL) and clustering method (Papaleo et al., 2009) . FEL was constructed using the principal component analysis (PCA) approach, in which the first and second eigenvectors were computed using "gmx anaeig" tool. The solvent accessible surface area (SASA) and gyration of radius R g were estimated via using "gmx sasa" and "gmx gyrate" tools, respectively. The collision cross section (CCS) were determined via ion mobility projection approximation calculation tool (IMPACT) protocol (Marklund et al., 2015) . The hydrogen bond (HB) contact was predicted as the default option of PyMOL (Schrödinger LLC, 2010) .

As mentioned above, the protein-protein association is a fundamental problem (Pan et al., 2019) . Protein-protein association or dissociation processes are hard to directly determine in experiments (Lupardus et al., 2014; Shan et al., 2014; Tang et al., 2006; Frisch et al., 2001) . In this work, atomistic MD simulations were employed to clarify the physical insights into the binding process between S protein and FAb. The structural change of the complexes was easily monitored over simulation trajectories (Casalino et al., 2020; Sikora et al., 2021) . The backbone root-mean-square deviation (RMSD) of the complexes suggested that systems mostly reached equilibrium states after 40 ns of MD simulations (Figs. S1-S2 of the Supporting information -SI file). Moreover, the superposition of calculated metrics also confirms the convergence of simulations (Fig. S3 of the SI file). Structural analyses were then carried out over equilibrium intervals of S protein + FAb complexes.

Structural metrics of complexes including backbone RMSD, R g , SASA, and CCS were analyzed in detail. The distribution of these values was reported in Fig. 3 . In particular, the 501Y.v2 variant enlarged the backbone RMSD of the complex (cf. Fig. 3A ), in which the corresponding values of WT and 501Y.V2 systems formed averages of 0.31 ± 0.03 and 0.38 ± 0.05 nm, respectively. The averaged R g of two complexes are almost similar with amounts of 5.13 ± 0.02 (WT) and 5.12 ± 0.03 (501Y.V2), but the 501Y.V2 R g is more diffusion than WT one (Fig. 3B) . The mean CCS of WT and 501Y.V2 systems are 134.47 ± 0.80 and 134.37 ± 1.64 nm 2 , respectively. Although the average value of CCS is not different, the form of CCS curve was significantly altered as well as RMSD curve (cf. Fig. 3C ). Moreover, it is consistent with the larger RMSD of 501Y.V2 were observed, the mean of total SASA is of 1400 ± 9 (WT) and 1403 ± 20 (501Y.V2) nm 2 (Fig. 3D) . The observations suggested that the complex structure of 501Y.V2 S protein + FAb was significantly altered in comparison with the WT system.

The C α atoms root-mean-square fluctuation (RMSF) of the complexes were investigated over equilibrium MD domains. The results are shown in Fig. 4 , in which the fluctuation of S protein was averaged over three chains. Two chains of FAb are asymmetric to each other, so the C α RMSF results were presented to all chains. The C α RMSF of S protein can roughly arrange into two domains corresponding to S1 and S2 domains. As shown in Fig. 4 , C α RMSF of S1 domain is significantly larger than that of S2 domain. Moreover, the 501Y.V2 S protein is more flexible than WT system since forming a larger C α RMSF along with the sequences. Especially, the RBD region of the South African variant much changed under the effects of three substitutions. In particular, residue K417 adopted a C α RMSF of 0.22 ± 0.02 nm, which is significantly smaller than that of residue K417N (0.33 ± 0.08 nm). While the residue E484 fluctuates with a C α RMSF of 0.28 ± 0.04 nm, the residue E484K changes with a C α RMSF of 0.55 ± 0.06 nm. In similar, N501Y formed a larger C α RMSF (0.43 ± 0.10 nm) compared with N501 (0.25 ± 0.04 nm). Furthermore, the fluctuation of 501Y.V2 S protein also turns the structure of FAb to becomes more flexible with the mean C α RMSF of 0.39 ± 0.08 nm in comparison with the mean WT C α RMSF of 0.26 ± 0.06 nm.

The stable conformation of the complexes can be probed using twodimensional FEL analysis, which was constructed using "gmx sham" tool (Papaleo et al., 2009; Mu et al., 2005) with two coordinates were first and second eigenvectors. Complexes eigenvectors were calculated using the principal component analysis (PCA) method (Papaleo et al., 2009; Mu et al., 2005) . The PCA method was used to generate 2D FEL in order to probe the conformational alterations of S protein + FAb complexes. The obtained FEL were displayed in Fig. 5 . Although the minima amounts are different with RBD + NAb complexes , the general picture is almost similar that the 501Y.v2 variant still increases the number of FEL minima of S protein + FAb complex. It probably implies that the 501Y.V2 complex is less stable than the WT one. This is in good agreement with a comparison of the backbone RMSD distribution of WT and 501Y.V2 systems above (cf. Fig. 3) . Therefore, it may argue that the binding affinity of FAb to S protein probably decreases when the 501Y.V2 variant was induced.

FEL of S protein + FAb formed more minima when 501Y.V2 variant was induced. In particular, FEL of WT S protein + FAb formed three minima denoted as WT1, WT2, and WT3 in Fig. 5 , which is located at (CV1; CV2) coordinates of (− 12.81; 0.63), (8.31; − 10.63), and (14.81; 10.00), respectively. The populations of three minima were of 50%, Table 1 , in which the minima MT3 formed the largest populations and the minima MT6 adopted the smallest populations.

The representative structures of reported minima were then obtained by using the clustering method with a non-hydrogen cut-off of 0.2 nm (Fig. S14 of the SI file) (Papaleo et al., 2009; . Moreover, the structural information as R g , CCS, and SASA was described in. Table 1 In good consistent with the metric analyses above, the WT and 501Y.V2 complexes formed a similar gyration of radius and CCS, but the SASA of the WT system is slightly larger than that of 501Y.V2 one. HB contacts between S protein and FAb of the complexed representative structures were reported in Figs. S15-S24 of the SI files. In particular, WT S protein formed 7.7 ± 1.1 HBs to FAb, while 501Y.V2 S protein adopted 5.4 ± 0.3 HBs to FAb. It may explain that NAb would form a larger affinity to WT S protein compared with the 501Y.V2 one. Furthermore, in order to reveal the binding affinity between FAb and S protein, the representative conformations of S protein + FAb were then employed as starting conformations for SMD simulations. However, it should keep in mind that the simulated PBC box was changed to a rectangular PBC in order to save computing resources as mentioned above.

In order to probe the binding affinity between two biomolecules, numerous approaches were developed such as perturbation free energy (Zwanzig, 1954) , thermodynamics integration (Kirkwood, 1935; Leonis et al., 2013) , molecular mechanics/Poisson-Boltzmann surface area (Srinivasan et al., 1998; Kollman et al., 2000) , umbrella sampling (US) , a combination of perturbation simulations and US (Siebenmorgen and Zacharias, 2019) , and enhanced sampling implementation of perturbation simulations . However, unfortunately, the S protein + FAb complex is a large system comprising more than a half-million atoms, it would be thus required huge computing resources if these approaches were applied. In that context, SMD simulations emerge as a potential approach to complete task force since requiring a small among of computing resources Pham et al., 2020) .

The antibody FAb was dissociated from the S protein using SMD simulations according to the recent work . The binding affinity of FAb to S protein was thus probed with an assumption that a stronger binder corresponds to a larger rupture force or pulling work system. The representative conformations of S protein + FAb provided by MD simulations were used as initial structures for SMD investigations. Because the complexes were reinserted to the new water box, these systems were first optimized using energy minimization and then relaxed using short NVT and NPT simulations. The protein coordinates were positionally retrained during these simulations. FAb was forced to mobilize from bound to unbound states via 2.5 ns of SMD simulations (cf. Figs. S4-S13 of the SI file). The obtained rupture force and pulling work were shown in Table 1 . In particular, the form of pulling force curve is in good consistent with the previous works Pham et al., 2020; Tam, Pham, et al., 2021) , in which the pulling force is linearly increased to the maximum value, F Max , then rapidly decreased to zero. The average of F Max and W was then calculated upon the population of the representative structures. The WT system adopted a mean value of 〈F Max 〉 = 1102.9 ± 27.6 pN and 〈W〉 = 149.1 ± 4.9 kcal mol -1 , which is significantly larger than that of the 501Y.V2 system forming a value of 〈F Max 〉 = 1038.7 ± 31.0 pN an 〈W〉 = 134.6 ± 5.0 kcal mol -1 . The obtained results imply that FAb formed a poorer binding affinity to 501Y. V2 S protein is in good agreement with the recent experiments Hoffmann et al., 2021; Wang et al., 2021) .

The structural conformations of the WT and 501Y.V2 S protein + FAb complexes were probed using atomistic MD simulations. The 501Y.V2 S protein + FAb structure is more flexible than the WT system. In particular, backbone RMSD of the 501Y.V2 complex was significantly larger than that of the WT implying a large structural change of the mutation system. The mean of other metrics as well as R g , CCS, and SASA are almost the same when compared two complexes, but the distribution of these values are clearly different. Moreover, the structural changes are clearly reflected in the FEL analysis, in which the number of FEL minima of the 501Y.V2 complex is much larger than that of the WT system. The population of the minima was also changed under the effects of the mutations. Furthermore, the binding affinity of FAb to S protein is then revealed using SMD simulations, which is starting from the representative structure of the complexes. The 501Y.V2 S protein was found to be formed a smaller binding affinity to FAb compared with the WT one that is in good agreement with the recent observations Hoffmann et al., 2021; Wang et al., 2021) . In addition, it may argue that the SMD simulations would be a potential approach to reveal the binding process between two proteins.

Son Tung Ngo: Conceptualization, Visualization, Methodology, Investigation, Validation, Formal analysis, Resources, Writingreview & editing, Funding acquisition. a The computed error is the standard error of the mean.

New SARS-CoV-2 variants -clinical, public health, and vaccine implications

GROMACS: high performance molecular simulations through multi-level parallelism from laptops to supercomputers

Motional timescale predictions by molecular dynamics simulations: case study using proline and hydroxyproline sidechain dynamics

Structures of human antibodies bound to SARS-CoV-2 spike reveal common epitopes and recurrent features of antibodies

SARS-CoV-2 variant B.1.617 is resistant to Bamlanivimab and evades antibodies induced by infection and vaccination

Beyond shielding: the roles of glycans in the SARS-CoV-2 spike protein

SARS-CoV-2 Variant Classifications and Definitions

Science Brief: Emerging SARS-CoV-2 Variants

Escape of SARS-CoV-2 501Y.V2 from neutralization by convalescent plasma

Resistance of SARS-CoV-2 variants to neutralization by monoclonal and serum-derived polyclonal antibodies

COVID-19 Vaccines, 2021. FDA, FDA

Particle mesh Ewald: an N⋅log(N) method for Ewald sums in large systems

Molecular basis of SARS-CoV-2 infection and rational design of potential antiviral agents: modeling and simulation approaches

Experimental assignment of the structure of the transition state for the association of Barnase and Barstar

SARS-CoV-2 variants B.1.351 and P.1 escape from neutralizing antibodies

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

Structural and functional properties of SARS-CoV-2 spike protein: potential antivirus drug development for COVID-19

Comparison of simple potential functions for simulating liquid water

Human neutralizing antibodies elicited by SARS-CoV-2 infection

Stereotypic neutralizing V H antibodies against SARS-CoV-2 spike protein receptor binding domain in patients with COVID-19 and healthy individuals

Statistical mechanics of fluid mixtures

Calculating structures and free energies of complex molecules: combining molecular mechanics and continuum models

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

A contribution to the drug resistance mechanism of darunavir, amprenavir, indinavir, and saquinavir complexes with HIV-1 protease due to flap mutation I50V: a systematic MM-PBSA and thermodynamic integration study

Potent neutralizing antibodies against multiple epitopes on SARS-CoV-2 spike

Structure of the pseudokinase-kinase domains from protein kinase TYK2 reveals a mechanism for Janus kinase (JAK) autoinhibition

Collision cross sections for structural proteomics

Energy landscape of a small peptide revealed by dihedral angle principal component analysis

Probable transmembrane amyloid α-helix bundles capable of conducting Ca 2+ Ions

Cholesterol molecules alter the energy landscape of small Aβ1-42 oligomers

Stability of Aβ11-40 trimers with parallel and antiparallel β-sheet organizations in a membrane-mimicking environment by replica exchange molecular dynamics simulation

Thermodynamics and kinetics in antibody resistance of the 501Y.V2 SARS-CoV-2 variant

Oversampling free energy perturbation simulation in determination of the ligand-binding free energy

Fine tuning of the copper active site in polysaccharide monooxygenases

Computational determination of potential inhibitors of SARS-CoV-2 main protease

Benchmark of popular free energy approaches revealing the inhibitors binding to SARS-CoV-2 Mpro

Atomic-level characterization of protein-protein association

Free-energy landscape, principal component analysis, and structural clustering to identify representative conformations from molecular dynamics simulations: the myoglobin case

Rapid prediction of possible inhibitors for SARS-CoV-2 main protease using docking and FPL simulations

WHO classifies India variant as being of global concern

Coronavirus envelope protein: current knowledge

The PyMOL Molecular Graphics System, Version 1.3r1

Molecular basis for pseudokinase-dependent autoinhibition of JAK2 tyrosine kinase

Evaluation of predicted protein-protein complexes by binding free energy simulations

Computational epitope map of SARS-CoV-2 spike protein

Continuum solvent studies of the stability of DNA, RNA, and phosphoramidate-DNA helices

Binding of inhibitors to the monomeric and dimeric SARS-CoV-2 Mpro

Computational estimation of potential inhibitors from known drugs against the main protease of SARS-CoV-2

Visualization of transient encounter complexes in protein-protein association

situ structural analysis of SARS-CoV-2 spike reveals flexibility mediated by three hinges

Escape from neutralizing antibodies by SARS-CoV-2 spike protein variants

SARS-CoV-2 501Y.V2 escapes neutralization by South African COVID-19 donor plasma

SARS and MERS: recent insights into emerging coronaviruses

Developing a fully glycosylated fulllength SARS-CoV-2 spike protein model in a viral membrane

COVID-19 Coronavirus Pandemic

SARS-CoV-2 spike protein: a key target for eliciting persistent neutralizing antibodies

Molecular interaction and inhibition of SARS-CoV-2 binding to the ACE2 receptor

Force field benchmark of amino acids. 2. Partition coefficients between water and organic solvents

Force field benchmark of amino acids: I. Hydration and diffusion in different water models

High-temperature equation of state by a perturbation method. I. Nonpolar gases

This research is funded by Vietnam National Foundation for Science and Technology Development (NAFOSTED) Grant #103.01-2017.368.

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.compbiolchem.2022.107636.