Structural Dynamics of SARS-CoV-2 Spike Protein: Computational Insights into Immune Evasion

Introduction

The SARS-CoV-2 spike glycoprotein mediates viral entry into host cells through its receptor-binding domain (RBD) interacting with angiotensin-converting enzyme 2 (ACE2) [1, 2]. This interaction is the primary determinant of host tropism and a major target for neutralizing antibodies [3, 4]. The spike protein undergoes extensive conformational changes during fusion, making its structural dynamics a critical area of study for understanding immune evasion [5, 6]. Computational approaches have become indispensable for characterizing these dynamics at atomic resolution, complementing experimental structural biology [7, 8]. This review examines how molecular dynamics simulations, free energy calculations, and machine learning models provide mechanistic insights into spike protein mutation effects on host receptor binding and antibody neutralization.

Molecular Dynamics Simulations of Spike Protein Conformational Landscapes

Molecular dynamics (MD) simulations are a cornerstone of computational virology. They enable the investigation of spike protein motions on multiple timescales, from side-chain rearrangements to large domain movements [2, 9]. All-atom MD simulations of the SARS-CoV-2 spike trimer have revealed the intrinsic flexibility of the RBD, which adopts "up" and "down" conformations relative to the core [9]. Furin cleavage at the S1/S2 boundary reshapes the conformational landscape, increasing the population of the RBD-up state required for ACE2 engagement [9]. This observation underscores the importance of proteolytic priming in modulating receptor accessibility.

Extended simulations have been used to map the free energy surface of the spike trimer, identifying metastable states that may serve as targets for therapeutic intervention [5, 10]. Computational studies have further demonstrated that N-glycosylation at specific sites alters the local conformational dynamics of the RBD, affecting both ACE2 binding and antibody recognition [11, 6]. The glycan shield imposes steric constraints that limit access to conserved epitopes, a phenomenon termed "glycan shielding" [12, 6]. MD simulations combined with hydrogen-deuterium exchange mass spectrometry have provided integrated views of antibody-induced allosteric pathways that drive spike disassembly [13].

Free Energy Calculations for Binding Affinity and Escape Prediction

Free energy perturbation (FEP) and MM/GBSA (Molecular Mechanics Generalized Born Surface Area) methods have been applied to quantify the impact of specific RBD mutations on ACE2 binding affinity [1, 14]. The RBD Y453F mutation, for instance, was shown to increase binding affinity by optimizing van der Waals contacts with ACE2 residues [1]. Conversely, mutations such as N501Y in emerging variants enhance binding through pi-pi stacking interactions with Y41 of ACE2 [1, 14]. Markov state models built from MD trajectories have elucidated the kinetic pathways of ACE2-RBD association and dissociation, revealing that certain mutations alter the rate-limiting conformational transitions [14].

Computational alanine scanning and energetic decomposition analyses have identified hotspot residues at the ACE2-RBD interface. These analyses are critical for predicting how mutations in the RBD affect overall binding stability [3, 15]. Mutations that reduce the binding free energy of neutralizing antibodies while preserving or enhancing ACE2 binding represent key immune evasion events [16, 15]. Frustration analyses of antibody-RBD interfaces have revealed that broadly neutralizing antibodies target conserved epitopes with low conformational frustration, whereas escape-prone antibodies engage regions with higher energetic frustration [3, 16].

Machine Learning Models for Antibody Escape Prediction

Machine learning (ML) approaches have been integrated with structural data to predict which RBD mutations are most likely to escape a given monoclonal antibody [3, 8]. Deep mutational scanning datasets provide training labels, while structural features such as residue depth, solvent accessibility, and local dynamics serve as input descriptors [17, 14]. Graph neural networks and attention-based models can capture long-range epistatic interactions within the spike trimer, which are essential for predicting combination mutations [18]. Intra-host recombination events that produce epistatic interactions have been shown to drive temperature-dependent adaptation and escape [18].

Large language models (LLMs) trained on protein sequences have also been adapted to predict mutation effects on ACE2 binding and antibody neutralization. These models leverage evolutionary information embedded in sequence alignments to score variants [14]. Integration of MD-derived features with sequence-based models has improved the accuracy of free energy predictions for unseen mutations [8, 14]. Such hybrid approaches are now being used to prospectively identify escape mutations before they arise in circulating variants [4].

graph TD
    A[Experimental Structures (cryo-EM, X-ray)], > B[Molecular Dynamics Simulations]
    B, > C[Conformational Ensemble Sampling]
    C, > D[Free Energy Calculations (FEP, MM/GBSA)]
    D, > E[Mutation Effect on ACE2 Binding]
    D, > F[Mutation Effect on Antibody Binding]
    B, > G[Glycan Shield Dynamics]
    G, > H[Epitope Accessibility Profiles]
    C, > I[Markov State Models / Kinetic Analysis]
    I, > J[Pathway Identification for Drug Targeting]
    E, > K[Machine Learning Training Data]
    F, > K
    K, > L[Escape Mutation Prediction Models]
    L, > M[Vaccine / Therapeutic Design Guidance]

Figure 1: Integrated computational workflow for studying spike protein dynamics and immune evasion.

Key Mutations in the Receptor-Binding Domain and Their Functional Consequences

The RBD accumulates mutations that modulate both ACE2 binding and antibody recognition. Table 1 summarizes effects of several representative mutations characterized through computational studies.

Table 1: Functional impact of selected RBD mutations analyzed by computational methods.

The N481K mutation, located near the ACE2 binding ridge, modifies the local hydrogen bond network and reduces the flexibility of the receptor-binding motif [19]. This change leads to altered glycan shielding that diminishes recognition by some neutralizing antibodies [19]. The XBB.1.5 subvariant contains several mutations that remodel four conserved antigenic determinants, as shown by computational mapping and MD simulations [20].

Glycan Shielding and Immune Evasion

N-glycosylation of the spike protein is a key determinant of immune evasion. The dense glycan shield physically obstructs antibody access to peptide epitopes while allowing flexibility for ACE2 binding [12, 6]. MD simulations have demonstrated that removal of specific glycosylation sites, such as N331 and N343, reduces conformational stability of the RBD and increases antibody accessibility [11, 12]. However, such removal also decreases ACE2 binding affinity due to altered local folding [11].

Comparative studies between SARS-CoV-2 and related bat coronaviruses, such as WIV1, have revealed differences in glycan density that correlate with host range and immune evasion potential [21]. The bat spike glycoprotein shows a more open glycan shield, allowing greater access to conserved receptor-binding surfaces [21]. These structural differences inform the design of pan-coronavirus vaccines intended to target conserved epitopes across sarbecoviruses [22].

Antibody Escape Mechanisms

Monoclonal antibodies that target the RBD can be classified into several classes based on epitope location and binding angle [3, 23]. Class 1 antibodies typically block ACE2 binding directly by overlapping with the receptor-binding motif, whereas Class 2 and 3 antibodies bind to distinct surfaces [23, 15]. Computational studies using a hierarchical mutational profiling approach have identified residues that, when mutated, confer resistance to multiple antibody classes simultaneously [3, 15]. These "escape hotspots" include positions 417, 484, and 501 [1, 15].

Sotrovimab resistance in Omicron variants was attributed to combined effects of glycan reorganization at epitope sites and mutation-induced changes in backbone dynamics [12]. Similarly, bebtelovimab activity was shown to be sensitive to structural dynamics of the spike, with mutations altering the balance between RBD-up and RBD-down states [5]. Nanobody binding to spike variants has been examined using force spectroscopy simulations, revealing differences in rupture forces that correlate with neutralization potency [24].

Liquid-liquid phase separation (LLPS) of the RBD has been recently proposed as a mechanism that may challenge antibody binding by sequestering epitopes in dense protein clusters [25]. Computational predictions indicate that the intrinsically disordered regions of the RBD contribute to LLPS propensity, and that mutations can modulate this behavior [25]. This finding adds a new dimension to immune evasion, as phase separation could reduce effective antibody concentrations at the viral surface.

Implications for Vaccine and Therapeutic Design

Computational models are increasingly used to guide the design of spike-based vaccines and therapeutic antibodies. Immunoinformatics approaches integrate predicted MHC binding, epitope conservation, and structural accessibility to select antigen sequences that elicit broad protection [22]. Engineering of the glycan shield through site-directed mutagenesis of glycosylation motifs can expose conserved neutralizing epitopes while retaining native-like spike conformation [11, 6]. Such glycan engineering has been successfully applied in the development of stabilized spike immunogens.

Broad neutralization is achieved by antibodies that target epitopes with low conformational frustration and high structural conservation [3, 16]. Computational energetic landscape analysis has identified several such epitopes in the stem helix and the fusion peptide region [13, 26]. IgG-bridging seeded aggregation of spikes, as described for a low-affinity antibody, represents a neutralization mechanism distinct from standard blocking, and can be optimized through structural modeling [26].

Peptide inhibitors designed to bind the fusion peptide region have been identified using mirrored combinatorial phage display and validated by MD simulations [27]. Macrocyclic peptides with broad-spectrum activity against Omicron variants have been developed using intranasal delivery, informed by computational docking and dynamics [28]. Small molecule inhibitors targeting the ACE2-spike interface have been discovered through integrated virtual screening and experimental validation [29]. Thiadiazole-based compounds showed promising inhibitory activity in cell-based assays, with MD simulations confirming stable binding to the RBD [29].

Cross-Species and Zoonotic Considerations

The structural dynamics of the spike protein are central to understanding zoonotic potential. Bat coronaviruses such as WIV1 utilize ACE2 orthologs from multiple mammalian species, and MD simulations have been used to predict binding affinities across host range [21, 30]. Camelid antibodies that target ACE2 have been shown to block SARS-CoV-2 binding and provide protection in animal models, an approach applicable to multiple zoonotic sarbecoviruses [30].

Intra-host recombination events in SARS-CoV-2 generate novel epistatic interactions that may alter host tropism [18]. Temperature-dependent adaptation mediated by spike mutations may affect viral fitness in different host species [18]. Computational models that incorporate both sequence evolution and structural dynamics can predict cross-species transmission risk for emerging coronaviruses [4, 7].

Resonant frequency analysis of viral particles, including spikes, has been proposed as a novel method for microwave-based detection and inactivation [31]. This approach relies on computational electromagnetics combined with structural models of the spike trimer. Although still in early stages, such technologies could provide rapid, label-free diagnostics for veterinary surveillance.

Challenges and Future Directions

Current computational methods face several limitations. The immense size of the spike trimer (over 600 kDa) requires significant computational resources for all-atom MD simulations, often necessitating coarse-grained models for long-timescale phenomena [7, 10]. Force field accuracy remains a concern for predicting binding free energies of highly glycosylated systems [6, 14]. Integration of multi-modal data from cryo-EM, HDX-MS, and computational models will improve reliability [13, 10].

Machine learning models trained on existing variant data may not generalize well to future mutations due to epistatic effects and changing sequence contexts [18]. Continual retraining with experimental data from new variants is essential. The combination of deep mutational scanning with MD-derived features holds promise for building more robust predictive models [3, 8].

Finally, the role of spike dynamics in alternative entry pathways, such as receptor-mediated endocytosis, is an active area of computational investigation [32]. Single-particle tracking combined with computational modeling has revealed the dynamic interplay between spike binding, endocytosis, and fusion [32]. These insights will be crucial for developing entry inhibitors that block the virus at multiple stages.

Conclusion

Computational virology has transformed our understanding of SARS-CoV-2 spike protein structural dynamics and immune evasion. Molecular dynamics simulations, free energy calculations, and machine learning models collectively provide mechanistic explanations for how mutations alter ACE2 binding affinity and antibody neutralization. Glycan shield rearrangements, epitope frustration landscapes, and liquid-liquid phase separation are emerging concepts that expand the repertoire of immune evasion strategies. The integration of these computational approaches with experimental validation continues to guide the rational design of vaccines and therapeutics against current and future zoonotic coronaviruses.

References

[1] Soliman OA, Shahine Y, Baecker D et al. Beyond the Mutation Abyss: Revisiting SARS-CoV-2 Receptor-Binding Domain Evolution from ACE2 Binding Optimization to Immune Epitope Remodeling. Pathogens. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41901725/

[2] Grassmann G, Miotto M, Alessandrini F et al. Insights into the role of dynamical features in protein complex formation: The case of SARS-CoV-2 spike binding with ACE2. Comput Struct Biotechnol J. 2025. URL: https://pubmed.ncbi.nlm.nih.gov/41245899/

[3] Alshahrani M, Gatlin W, Ludwick M et al. Mechanisms of Binding and Immune Escape Resistance for Broadly Neutralizing Antibodies Targeting Distinct Conserved SARS-CoV-2 Spike Epitopes: A Hierarchical Approach Integrating Mutational Profiling and Energy Landscape Analysis. Int J Mol Sci. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42123600/

[4] Uzer F, Erendor F, Sanlioglu S. SARS-CoV-2 Variants and Immune Evasion: Mapping the Future of Vaccine Design. Rev Med Virol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42047168/

[5] Mansur M, Hong H, Singh S et al. SARS-CoV-2 Spike Protein's Structural Dynamics Affect the Activity of the Bebtelovimab Antibody. J Chem Inf Model. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42202234/

[6] Osman SH, Haynes CA, Woolfitt AR et al. Role of N-Glycosylation in SARS-CoV-2 Spike Glycoprotein Conformational Dynamics, Structure, and Function. ACS Infect Dis. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42018455/

[7] Herzig JC, Magwira ML, Lovell SC. Structural Constraints Acting on the SARS-CoV-2 Spike Protein Reveal Limited Space for Viral Adaptation. Genome Biol Evol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41876430/

[8] Netsey EK, Naandam SM, Asante Jnr J et al. Structural and Functional Impacts of SARS-CoV-2 Spike Protein Mutations: Insights From Predictive Modeling and Analytics. JMIR Bioinform Biotechnol. 2025. URL: https://pubmed.ncbi.nlm.nih.gov/41359941/

[9] Mani N, Suresh R, Chakraborty S. Cleaved versus Uncleaved: How furin cleavage reshapes the conformational landscape of SARS-CoV-2 spike. Protein Sci. 2025. URL: https://pubmed.ncbi.nlm.nih.gov/41236412/

[10] Mancini T, Luchetti N, Macis S et al. Multimodal Structural Characterization of SARS-CoV-2 Spike Variants: Spectroscopic and Computational Insights. Int J Mol Sci. 2025. URL: https://pubmed.ncbi.nlm.nih.gov/41226381/ *** Disclaimer: This article is for educational and informational purposes only. It is not intended to substitute for professional veterinary advice, diagnosis, treatment, or regulatory guidance. Always consult a licensed veterinarian or qualified specialist regarding animal health, disease diagnosis, and therapeutic decisions.

[11] Wang H, Li X, He P et al. Investigating the effects and underlying mechanisms of glycosylation sites on the immunogenicity of COVID-19 vaccines. Signal Transduct Target Ther. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42129141/

[12] Kumar A, Yadav AJ, Tripathi T et al. Glycan shielding and epitope reorganization drive sotrovimab resistance in SARS-CoV-2 Omicron variants. Arch Biochem Biophys. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42128042/

[13] Luo Y, Liu B, Liu R et al. Hydrogen-Deuterium Exchange Mass Spectrometry Reveals an Antibody-Induced Allosteric Pathway Driving SARS-CoV-2 Spike Trimer Disassembly. J Am Chem Soc. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41784932/

[14] Zhou Y, Wang T. Dynamic insights into the structural evolution of ACE2-RBD interactions through molecular dynamics simulation, Markov state modeling, and large language model mutation prediction. J Chem Phys. 2025. URL: https://pubmed.ncbi.nlm.nih.gov/41268951/

[15] Alshahrani M, Parikh V, Foley B et al. Dissecting binding and immune evasion mechanisms for ultrapotent Class I and Class 4/1 neutralizing antibodies of SARS-CoV-2 spike protein using a multi-pronged computational approach: neutral frustration architecture of binding interfaces and immune escape hotspots drives adaptive evolution. Phys Chem Chem Phys. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41623222/

[16] Alshahrani M, Gatlin W, Ludwick M et al. Frustration Landscapes of Broadly Neutralizing SARS-CoV-2 Spike Antibodies Targeting Conserved Epitopes Reveal Energetic Logic of Escape-Proof and Escape-Prone Mechanisms. bioRxiv. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41959354/

[17] Dastorani S, Shariati M, Hasanzadeh Ghasemi R. Analysis of the binding affinity of SARS-CoV-2 spike RBD to novel DNA origami-heparin nanostructures: a steered molecular dynamics (SMD) study. J Biomol Struct Dyn. 2025. URL: https://pubmed.ncbi.nlm.nih.gov/41388771/

[18] Altaf M, Kin Cheng MT, González-Vázquez LD et al. SARS-CoV-2 intra-host recombination promotes epistatic spike interactions and temperature-dependent adaptation. Cell Rep. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42213788/

[19] Smatti MK, Al-Khatib H, Suleman M et al. Functional and structural characterization of the SARS-CoV-2 spike N481K mutation. Arch Virol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42243594/

[20] Akbulut E, Yildirim M, Kahraman H. SARS-CoV-2 Spike Protein XBB.1.5 Mutations Altered Four Conserved Antigenic Determinants. Int J Mol Sci. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41752076/

[21] Liu C, Zheng J, Wang Y et al. Cryo-EM structure of locked spike glycoprotein from bat SARS-like coronavirus WIV1, molecular dynamics and biophysics across host range. Proc Natl Acad Sci U S A. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41706884/

[22] Rahimnahal S, Yousefizadeh S, Mohammadi Y. Designing a novel vaccine against COVID-19 based on spike SARS-Cov-2 notable mutations using immunoinformatics approaches. PLoS One. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41746934/

[23] Harit A, Mor M, Yefet R et al. Monoclonal antibodies from COVID-19 convalescent patients target cryptic epitopes for broad SARS-CoV-2 neutralization. Proc Natl Acad Sci U S A. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41880581/

[24] Cofas-Vargas LF, Olivos-Ramirez GE, Marrink SJ et al. A comparative nanomechanical study of antibody and nanobody binding to SARS-CoV-2 variants. Phys Chem Chem Phys. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41885290/

[25] Sodasani M, Grandhi AVKS, Mukala N et al. Receptor binding domain of SARS CoV2 spike protein exhibits in vitro liquid-liquid phase separation due to structural disorderedness that may challenge the vaccine-generated antibody binding. Biochim Biophys Acta Gen Subj. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41270841/

[26] Lv N, Chen P, Dai X et al. IgG-Bridging-Seeded Synergistic Aggregation of SARS-CoV-2 Spikes Underlies Potent Neutralization by a Low-Affinity Antibody. Adv Sci (Weinh). 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41355083/

[27] Pal A, Roy NS, Angeliadis M et al. Screening for Peptides to Bind and Functionally Inhibit SARS-CoV-2 Fusion Peptide Using Mirrored Combinatorial Phage Display and Human Proteomic Phage Display. Molecules. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41599331/

[28] Wang M, Yang J, Tan Y et al. Intranasal administration of broad-spectrum macrocyclic peptide inhibitor protects against SARS-CoV-2 Omicron variants. Nat Commun. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41587975/

[29] Gul Z, Shahid SA, Abid OU et al. Discovery of thiadiazole-based small-molecule inhibitors of SARS-CoV-2 spike-ACE2 interaction through integrated computational prediction and experimental validation. Sci Rep. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41484143/

[30] Blachier S, Vaney MC, Conquet L et al. Targeting ACE2 with a camelid antibody inhibits SARS-CoV-2 binding and has protective effects in vivo. Nat Commun. 2025. URL: https://pubmed.ncbi.nlm.nih.gov/41271706/

[31] Kuang Z, Luginsland J, Hung CS et al. Identifying resonant frequencies of viruses for microwave-based detection and inactivation of pathogenic viruses. Sci Rep. 2025. URL: https://pubmed.ncbi.nlm.nih.gov/41315828/

[32] Ma X, Zhang Z, Yang D et al. Beyond ACE2: unveiling the receptor-mediated endocytic dynamic mechanism of SARS-CoV-2 at the single-particle level. J Mater Chem B. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42300082/

[33] Navhaya LT, Monama MZ, Matsebatlela TM et al. In-Depth Molecular Dynamics Simulations Reveal Ligand-Induced Modulations of the HSPA8-SARS-CoV-2 Spike Protein Interaction. Int J Mol Sci. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42196270/

[34] Huang CY, Cheng CC, Chen SH et al. SARS-CoV-2 Spike Protein-Mediated Cardiac Dysfunction: Structural Abnormalities, Impaired Calcium Dynamics, and Gene Expression Changes in Human Stem Cell-Derived Cardiomyocytes. J Med Virol. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41493418/

[35] Torabfam M, Celebi Torabfam G, Osonga F et al. Improving quercetin solubility via structural modification enhances dual-target coronavirus entry: an integrated in-vitro and in-silico study. Sci Rep. 2025. URL: https://pubmed.ncbi.nlm.nih.gov/41309904/