Computational Analysis of Viral Capsid Assembly: Insights from Cryo-EM and Molecular Dynamics Simulations

1. Introduction

Viral capsid assembly is a critical step in the replication cycle of many viruses, including those that cause disease in veterinary species. The process by which capsid proteins (CPs) self-assemble into a protective shell around the viral genome is governed by precise biophysical interactions and often involves conformational changes, protein-protein recognition, and nucleic acid packaging signals [1, 2]. Understanding these mechanisms at the atomic and molecular level can inform the development of antiviral interventions targeting assembly, and can also guide the design of capsid-based nanoparticles for antigen display or gene delivery [3, 4].

The advent of high-resolution cryo-electron microscopy (cryo-EM) has revolutionized structural virology by providing near-atomic density maps of intact viral capsids and their assembly intermediates [5, 6]. However, these static snapshots do not fully capture the dynamic pathways of assembly, which involve nucleation, growth, and often error correction [7, 8]. Molecular dynamics (MD) simulations have emerged as a complementary tool that can bridge the gap between static structures and dynamic processes [3]. When combined with cryo-EM data, simulations can explore the energetic landscape of capsid assembly, the role of solvent and ions, and the effects of mutations [9, 10, 11].

This review presents an exhaustive analysis of computational approaches used to study viral capsid assembly, with emphasis on the integration of cryo-EM and MD simulations. Examples are drawn from HIV-1, norovirus, hepatitis B virus (HBV), bacteriophages, and plant viruses such as brome mosaic virus (BMV) and cowpea chlorotic mottle virus (CCMV), many of which serve as models for veterinary viral systems. The article also discusses machine learning applications and uncertainty quantification in assembly modeling, and links these methods to practical outcomes such as antiviral drug design and capsid engineering. Readers are encouraged to explore the 3D Protein Viewer to examine the capsid protein structures discussed herein.

2. Structural Basis of Capsid Assembly from Cryo-EM

Cryo-EM has become the method of choice for determining the structures of icosahedral viral capsids [1, 5]. The technique yields three-dimensional reconstructions from thousands of particle images, allowing computational analysis of conformational variability and assembly intermediates [12, 6]. For example, asymmetric cryo-EM reconstruction of bacteriophage MS2 revealed the arrangement of the single-stranded RNA genome inside the capsid and its interactions with CPs, providing insights into genome packaging and assembly [12]. Similarly, cryo-EM of tubular arrays of HIV-1 Gag resolved essential structures for immature virus assembly [13]. A later study achieved 8.8 Å resolution of the immature HIV-1 capsid structure within intact virions, revealing the arrangement of hexameric and pentameric lattices [14].

Computational analysis of cryo-EM data can also extract dynamics from apparently static ensembles. Variance analysis of particle sets can map structural heterogeneity along the assembly pathway, as demonstrated for bacteriophage HK97 maturation intermediates [6]. This approach enables the identification of intermediate states and the order of conformational changes that occur during capsid maturation [6]. The integration of cryo-EM with computational modeling has led to the concept of "assembly landscapes" that map the stability and kinetics of different oligomeric states [7, 5].

A notable tool for representing cryo-EM-derived capsid structures in a computationally tractable manner is CapsidMesh, which generates atomic-detail structured mesh representations of icosahedral capsids and allows finite element analysis of their mechanical properties [15]. Such mechanical characterization is important for understanding how capsids withstand internal pressure from the packaged genome and external stresses during transmission [15, 16].

3. Molecular Dynamics Simulations of Capsid Assembly

MD simulations provide a computational microscope to observe the self-assembly of capsid proteins at atomic or coarse-grained resolution [3, 17]. Coarse-grained (CG) models are particularly useful for capturing the assembly of large complexes such as viral capsids, which involve thousands of protein copies and timescales beyond the reach of all-atom simulations [17, 2]. For instance, multiscale modeling of hepatitis B virus (HBV) capsid assembly and its dimorphism (T=3 and T=4 capsids) has shown that subtle changes in protein conformation and solution conditions can shift the equilibrium between different capsid sizes [9]. This work combined continuum and particle-based simulations to explain how the same protein can assemble into shells with different triangulation numbers [9].

Atomistic simulations are computationally expensive but can provide detailed information on specific interactions, such as hydrogen bonding, salt bridges, and hydrophobic contacts [3, 18]. For HIV-1, MD simulations have been used to study the capsid nucleation step, revealing that the C-terminal domain (CTD) of the CA protein forms critical stabilizing contacts [19]. A dynamical systems model for HIV-1 capsid nucleation identified key parameters that control the onset of assembly [20]. Furthermore, combined in vitro protease cleavage and computational simulations elucidated the HIV-1 capsid maturation pathway, showing how the viral protease cleaves the Gag polyprotein to trigger a structural rearrangement into the mature conical capsid [11].

Simulations can also examine the effect of explicit RNA or DNA on assembly kinetics and pathways. The presence of a genome can accelerate assembly by providing a scaffold for CP binding, or it can induce errors if the genome length or sequence is suboptimal [21, 2, 22]. Poudel et al. used MD simulations to study the impact of hydrogen bonding between MS2 CP and its hairpin RNA recognition element [18]. This study showed that specific RNA-CP interactions stabilize the initial complex and guide the assembly of the full capsid. Dykeman developed a model for viral assembly around an explicit RNA sequence that generates an implicit fitness landscape, explaining how sequence variations affect assembly efficiency [21].

4. Integrating Cryo-EM Data into Simulation Workflows

A powerful approach is to use cryo-EM density maps as restraints or starting configurations for MD simulations [3, 5]. The software tool ioNERDSS (Input/Output for Network-free, Rule-based, Dynamic, Stochastic Simulations) transforms macromolecular structures derived from cryo-EM into simulations of self-assembly, enabling multiscale modeling of capsid formation [23]. This tool bridges the gap between static structural data and dynamic assembly simulations, allowing researchers to test hypotheses about assembly intermediates and pathways [23].

Another integration strategy involves normal mode analysis (NMA) of cryo-EM structures to extract large-scale collective motions. Fast normal mode computations of capsid dynamics inspired by resonance have been developed to efficiently compute vibrational modes that may be relevant to assembly and stability [24]. Song performed a comparative study of viral capsids and bacterial compartments using NMA to reveal an enriched understanding of shell dynamics [25]. These analyses highlight that capsid proteins are not rigid bricks but rather dynamic entities that flex and breathe, which can facilitate conformational changes needed for assembly and maturation [3, 25].

The combination of cryo-EM with MD also enables the study of mechanical stress distributions within capsids. TensorCalculator is a computational tool that maps mechanical stress tensors onto the capsid surface, as demonstrated for CCMV [26]. This tool revealed that stress is localized near pentameric vertices, which are hotspots for disassembly and may be targeted by antiviral compounds [26]. Asymmetric cryo-EM reconstructions further enable the study of genome organization inside capsids, as shown for a nodavirus using graph-theoretical analysis of tomographic data [27]. Such analyses provide constraints for simulations that include the genome.

5. Machine Learning and Uncertainty Quantification in Assembly Studies

Machine learning (ML) has been applied to predict assembly outcomes and fitness landscapes. Dechant and He used ML to infer a virus assembly fitness landscape from experimental data, identifying which mutations in the CP are tolerated and which disrupt assembly [28]. Similarly, Marques et al. applied ML to predict the assembly of adeno-associated virus (AAV) capsid libraries, useful for designing vectors with altered tropism [4]. In a picornavirus system, Mattenberger et al. globally defined the effects of mutations in the capsid using a combination of deep mutational scanning and computational modeling [29]. These approaches are directly transferable to veterinary viruses such as foot-and-mouth disease virus (FMDV) or porcine circovirus type 2 (PCV2), where capsid mutations affect host range and vaccine efficacy [30].

Uncertainty quantification (UQ) is essential when computational models are used to predict assembly outcomes from limited experimental data. Clement et al. developed a UQ approach that rigorously estimates the confidence in predictions of capsid assembly energetics [31, 8]. By propagating uncertainties from input parameters (e.g., binding energies, CP concentrations) through the model, they could identify which parameters most strongly influence assembly predictions and thereby guide further experiments [8, 31]. This methodology is particularly valuable in veterinary contexts where experimental data may be scarce.

6. Antiviral Strategies Informed by Computational Assembly Analysis

Understanding the atomic details of capsid assembly opens doors for rational antiviral design. For HIV-1, a novel intersubunit interaction critical for core assembly was discovered through computational analysis, revealing a potentially targetable inhibitor binding pocket [10]. This pocket is located at the interface between CA hexamers and can be targeted by small molecules to prevent capsid maturation or stability [10]. Similarly, dynamic allostery governs the interplay between HIV-1 capsid and the host protein cyclophilin A, and MD simulations have identified druggable allosteric sites [32].

For veterinary viruses, in silico analysis of PCV2 capsid surface structure variations resulting from loop mutations has guided the design of more stable vaccine antigens [30]. The ability to predict how mutations alter capsid assembly or antigenicity is critical for vaccine strain selection. Computational models of assembly also contribute to the design of virus-like particles (VLPs) for vaccine delivery. By analyzing the dimorphism of HBV capsids, researchers can engineer VLPs that preferentially form a specific size to optimize immune presentation [9]. The computational design of capsid-like nanoparticles is further discussed in a related article: Computational Design of Viral Capsid-Like Nanoparticles for Antigen Display.

The mechanical properties of capsids, studied through mesh-based models and MD, can inform the development of antiviral compounds that increase or decrease capsid rigidity, leading to premature disassembly or failure to uncoat [15, 16]. For example, DNA bending-induced phase transitions in bacteriophage lambda capsids have been modeled, and the principles may apply to veterinary bacteriophage therapies [16].

7. Comparative Analysis of Capsid Assembly across Viral Families

The following table summarizes key computational studies on capsid assembly from the provided literature, highlighting the virus, computational method, and main findings.

8. Workflow for Computational Capsid Assembly Analysis

The following Mermaid diagram illustrates a typical workflow that integrates cryo-EM and MD simulations to study capsid assembly and inform antiviral design.

flowchart TD
    A[Purified viral capsid or VLP], > B[Cryo-EM data collection]
    B, > C[3D reconstruction & density map]
    C, > D[Atomic model building & refinement]
    D, > E{Molecular dynamics simulations}
    E, > F[All-atom MD for interaction details]
    E, > G[Coarse-grained MD for assembly pathways]
    C, > H[Variance analysis for conformational heterogeneity]
    H, > I[Identification of assembly intermediates]
    F, > J[Compute binding free energies & mutation effects]
    G, > J
    I, > K[Uncertainty quantification of assembly model]
    J, > K
    K, > L[Predict assembly fitness landscape]
    L, > M[Rational design of antiviral compounds or capsid mutants]
    M, > N[Experimental validation & iteration]

This workflow begins with sample preparation and cryo-EM data acquisition, leading to a density map that is used to build an atomic model [1, 5]. The model then serves as input for MD simulations at multiple resolutions [3]. Variance analysis of the cryo-EM dataset can reveal intermediate states [6]. Simulation outputs feed into uncertainty-quantified models [8, 31] to predict the effects of sequence changes or ligand binding, ultimately guiding antiviral or vaccine design.

9. Conclusion

Computational analysis of viral capsid assembly has matured into a discipline that seamlessly integrates cryo-EM and molecular dynamics simulations. The availability of high-resolution capsid structures from cryo-EM, combined with the dynamic insight from MD simulations, provides a comprehensive view of the assembly process from initial nucleation to final maturation. Machine learning and uncertainty quantification add predictive power to these models, enabling the rational design of antiviral strategies and engineered capsids for veterinary applications. Future developments in multiscale simulation methods, coupled with advances in cryo-EM resolution and speed, will continue to unveil the intricate mechanisms by which viruses build their protective shells. Veterinary virology stands to benefit greatly from these computational tools, particularly for emerging and neglected viral diseases of livestock and companion animals.

References

[1] Johnson JE, Olson AJ. Icosahedral virus structures and the protein data bank. J Biol Chem. 2021;:33744290. URL: https://pubmed.ncbi.nlm.nih.gov/33744290/

[2] Perlmutter JD, Perkett MR, Hagan MF. Pathways for virus assembly around nucleic acids. J Mol Biol. 2014;:25036288. URL: https://pubmed.ncbi.nlm.nih.gov/25036288/

[3] Pérez-Segura C, Perilla JR, Hadden-Perilla JA. Atomistic simulations of intact virus capsids: A computational challenge worth the scientific payoff. Curr Opin Struct Biol. 2025;:40544581. URL: https://pubmed.ncbi.nlm.nih.gov/40544581/

[4] Marques AD, Kummer M, Kondratov O, et al. Applying machine learning to predict viral assembly for adeno-associated virus capsid libraries. Mol Ther Methods Clin Dev. 2021;:33511242. URL: https://pubmed.ncbi.nlm.nih.gov/33511242/

[5] Doerschuk PC, Gong Y, Xu N, et al. Virus particle dynamics derived from CryoEM studies. Curr Opin Virol. 2016;:27085980. URL: https://pubmed.ncbi.nlm.nih.gov/27085980/

[6] Gong Y, Veesler D, Doerschuk PC, et al. Effect of the viral protease on the dynamics of bacteriophage HK97 maturation intermediates characterized by variance analysis of cryo EM particle ensembles. J Struct Biol. 2016;:26724602. URL: https://pubmed.ncbi.nlm.nih.gov/26724602/

[7] Spiriti J, Conway JF, Zuckerman DM. Should Virus Capsids Assemble Perfectly? Theory and Observation of Defects. Biophys J. 2020;:33113349. URL: https://pubmed.ncbi.nlm.nih.gov/33113349/

[8] Clement N, Rasheed M, Bajaj CL. Viral Capsid Assembly: A Quantified Uncertainty Approach. J Comput Biol. 2018;:29313735. URL: https://pubmed.ncbi.nlm.nih.gov/29313735/

[9] Mohajerani F, Tyukodi B, Schlicksup CJ, et al. Multiscale Modeling of Hepatitis B Virus Capsid Assembly and Its Dimorphism. ACS Nano. 2022;:36054910. URL: https://pubmed.ncbi.nlm.nih.gov/36054910/

[10] Craveur P, Gres AT, Kirby KA, et al. Novel Intersubunit Interaction Critical for HIV-1 Core Assembly Defines a Potentially Targetable Inhibitor Binding Pocket. mBio. 2019;:30862755. URL: https://pubmed.ncbi.nlm.nih.gov/30862755/

[11] Ning J, Erdemci-Tandogan G, Yufenyuy EL, et al. In vitro protease cleavage and computer simulations reveal the HIV-1 capsid maturation pathway. Nat Commun. 2016;:27958264. URL: https://pubmed.ncbi.nlm.nih.gov/27958264/

[12] Koning RI, Gomez-Blanco J, Akopjana I, et al. Asymmetric cryo-EM reconstruction of phage MS2 reveals genome structure in situ. Nat Commun. 2016;:27561669. URL: https://pubmed.ncbi.nlm.nih.gov/27561669/

[13] Bharat TA, Castillo Menendez LR, Hagen WJ, et al. Cryo-electron microscopy of tubular arrays of HIV-1 Gag resolves structures essential for immature virus assembly. Proc Natl Acad Sci U S A. 2014;:24843179. URL: https://pubmed.ncbi.nlm.nih.gov/24843179/

[14] Schur FK, Hagen WJ, Rumlová M, et al. Structure of the immature HIV-1 capsid in intact virus particles at 8.8 Å resolution. Nature. 2015;:25363765. URL: https://pubmed.ncbi.nlm.nih.gov/25363765/

[15] Alonzo-Velázquez JL, Botello-Rionda S, Herrera-Guzmán R, et al. CapsidMesh: Atomic-detail structured mesh representation of icosahedral viral capsids and the study of their mechanical properties. Int J Numer Method Biomed Eng. 2018;:29603677. URL: https://pubmed.ncbi.nlm.nih.gov/29603677/

[16] Lander GC, Johnson JE, Rau DC, et al. DNA bending-induced phase transition of encapsidated genome in phage λ. Nucleic Acids Res. 2013;:23449219. URL: https://pubmed.ncbi.nlm.nih.gov/23449219/ *** 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.

[17] Machado MAR, González HC, Pantano S. MD Simulations of Viruslike Particles with Supra CG Solvation Affordable to Desktop Computers. J Chem Theory Comput. 2017;:28876928. URL: https://pubmed.ncbi.nlm.nih.gov/28876928/

[18] Poudel L, Twarock R, Steinmetz NF, et al. Impact of Hydrogen Bonding in the Binding Site between Capsid Protein and MS2 Bacteriophage ssRNA. J Phys Chem B. 2017;:28581757. URL: https://pubmed.ncbi.nlm.nih.gov/28581757/

[19] Sha H, Zhu F. Parameter Optimization for Interaction between C-Terminal Domains of HIV-1 Capsid Protein. J Chem Inf Model. 2017;:28426204. URL: https://pubmed.ncbi.nlm.nih.gov/28426204/

[20] Sadre-Marandi F, Liu Y, Liu J, et al. Modeling HIV-1 viral capsid nucleation by dynamical systems. Math Biosci. 2015;:26596714. URL: https://pubmed.ncbi.nlm.nih.gov/26596714/

[21] Dykeman EC. A Model for Viral Assembly around an Explicit RNA Sequence Generates an Implicit Fitness Landscape. Biophys J. 2017;:28793206. URL: https://pubmed.ncbi.nlm.nih.gov/28793206/

[22] Dykeman EC, Stockley PG, Twarock R. Building a viral capsid in the presence of genomic RNA. Phys Rev E Stat Nonlin Soft Matter Phys. 2013;:23496558. URL: https://pubmed.ncbi.nlm.nih.gov/23496558/

[23] Ying YM, Sang M, Au G, et al. Transforming macromolecular structures into simulations of self-assembly with ioNERDSS. bioRxiv. 2026;:41659451. URL: https://pubmed.ncbi.nlm.nih.gov/41659451/

[24] Na H, Song G. Fast normal mode computations of capsid dynamics inspired by resonance. Phys Biol. 2018;:29557348. URL: https://pubmed.ncbi.nlm.nih.gov/29557348/

[25] Song G. A comparative study of viral capsids and bacterial compartments reveals an enriched understanding of shell dynamics. Proteins. 2018;:29114976. URL: https://pubmed.ncbi.nlm.nih.gov/29114976/

[26] Kononova O, Maksudov F, Marx KA, et al. TensorCalculator: exploring the evolution of mechanical stress in the CCMV capsid. J Phys Condens Matter. 2018;:29231176. URL: https://pubmed.ncbi.nlm.nih.gov/29231176/

[27] Geraets JA, Dykeman EC, Stockley PG, et al. Asymmetric genome organization in an RNA virus revealed via graph-theoretical analysis of tomographic data. PLoS Comput Biol. 2015;:25793998. URL: https://pubmed.ncbi.nlm.nih.gov/25793998/

[28] Dechant PP, He YH. Machine-learning a virus assembly fitness landscape. PLoS One. 2021;:33951035. URL: https://pubmed.ncbi.nlm.nih.gov/33951035/

[29] Mattenberger F, Latorre V, Tirosh O, et al. Globally defining the effects of mutations in a picornavirus capsid. Elife. 2021;:33432927. URL: https://pubmed.ncbi.nlm.nih.gov/33432927/

[30] Wang N, Zhan Y, Wang A, et al. In silico analysis of surface structure variation of PCV2 capsid resulting from loop mutations of its capsid protein (Cap). J Gen Virol. 2016;:27902320. URL: https://pubmed.ncbi.nlm.nih.gov/27902320/

[31] Clement N, Rasheed M, Bajaj C. Uncertainty Quantified Computational Analysis of the Energetics of Virus Capsid Assembly. Proceedings (IEEE Int Conf Bioinformatics Biomed). 2016;:28936368. URL: https://pubmed.ncbi.nlm.nih.gov/28936368/

[32] Lu M, Hou G, Zhang H, et al. Dynamic allostery governs cyclophilin A-HIV capsid interplay. Proc Natl Acad Sci U S A. 2015;:26553990. URL: https://pubmed.ncbi.nlm.nih.gov/26553990/

[33] Chakravarty A, Reddy VS, Rao ALN. Unravelling the Stability and Capsid Dynamics of the Three Virions of Brome Mosaic Virus Assembled Autonomously In Vivo. J Virol. 2020;:31996436. URL: https://pubmed.ncbi.nlm.nih.gov/31996436/

[34] Martin TD, Hill EH, Whitten DG, et al. Oligomeric Conjugated Polyelectrolytes Display Site-Preferential Binding to an MS2 Viral Capsid. Langmuir. 2016;:27464311. URL: https://pubmed.ncbi.nlm.nih.gov/27464311/

[35] Taylor WR, Stoye JP, Taylor IA. A comparative analysis of the foamy and ortho virus capsid structures reveals an ancient domain duplication. BMC Struct Biol. 2017;:28372592. URL: https://pubmed.ncbi.nlm.nih.gov/28372592/