Computational Design of Antiviral Peptides Targeting Viral Envelope Proteins

Introduction

Antiviral peptides represent a promising class of therapeutic agents that can inhibit viral entry by binding to critical epitopes on viral envelope proteins [1, 2]. These peptides are designed to interfere with the viral fusion machinery, blocking receptor engagement or preventing conformational changes required for membrane fusion [3, 4]. The envelope proteins of enveloped viruses, including class I fusion proteins such as influenza hemagglutinin and coronavirus spike, present well-defined structural targets for such interventions [5, 6].

Computational design has become an indispensable approach for accelerating the discovery of these peptides [7, 8]. By leveraging atomic-resolution structural data of viral envelope proteins, bioinformatics tools can predict binding interfaces, evaluate binding free energies, and optimize peptide sequences for enhanced affinity and specificity [9, 10]. This article provides an exhaustive review of the computational strategies used to design antiviral peptides against viral envelope proteins, focusing on molecular docking, molecular dynamics simulations, free energy calculations, and machine learning-based screening. The focus remains on veterinary pathogens and comparative virology, drawing parallels where appropriate to host-range dynamics [11].

Strategy Overview: Structure-Based Peptide Design

The core principle of computational antiviral peptide design is to identify peptide sequences that can mimic natural binding partners or occupy conserved pockets on the viral envelope protein [12, 13]. The typical workflow begins with the selection of a high-resolution three-dimensional structure of the target envelope protein, often obtained from X-ray crystallography or cryo-electron microscopy [14]. This structure can be used for analyses that are discussed in greater detail within related articles such as those on Structural Prediction of Viral Envelope Glycoproteins Using AlphaFold2 and Cryo-EM Density Map Interpretation and Computational Structure Fitting.

The primary computational strategies can be categorized as follows:

1. Peptide Docking: Predicting the preferred orientation of a flexible peptide ligand within the binding site of a rigid or flexible protein receptor [15, 16].
2. Molecular Dynamics (MD) Simulations: Simulating the atomic-level motions of the peptide-protein complex to assess stability and refine binding poses [3, 7].
3. Free Energy Calculations: Quantifying the binding affinity between the peptide and the envelope protein using methods such as molecular mechanics generalized Born surface area (MM-GBSA) or molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) [10, 12].
4. Machine Learning (ML) Screening: Using trained models to predict peptide bioactivity based on sequence and structural features [2, 9, 13].

These approaches are not mutually exclusive and are often integrated into a hierarchical pipeline [17, 18].

flowchart TD
    A[Viral Envelope Protein Target], > B[Obtain 3D Structure]
    B, > C{Select Binding Interface}
    C, > D[Receptor-Binding Domain (RBD)]
    C, > E[Heptad Repeat 1/2 (HR1/HR2)]
    C, > F[Fusion Peptide Region]
    D, > G[Peptide Library Generation]
    E, > G
    F, > G
    G, > H[Molecular Docking (e.g., AutoDock, HADDOCK)]
    H, > I[Top Scoring Poses]
    I, > J[Molecular Dynamics Simulations]
    J, > K[Stability & Interaction Analysis]
    K, > L[Free Energy Calculation (MM-GBSA/PBSA)]
    L, > M[Lead Peptide Candidates]
    M, > N[Experimental Validation (e.g., SPR, Pseudovirus Neutralization)]
    N, > O[Optimized Antiviral Peptide]
    
    subgraph ML_Module
        P[Peptide Sequence Database], > Q[Feature Encoding]
        Q, > R[Training ML Model]
        R, > S[Predict Bioactivity]
        S, > T[Filter Candidates]
    end
    
    I, > T
    M, > T
    T, > U[Prioritized Virtual Hits]
    U, > J

Molecular Docking of Peptides to Envelope Proteins

Molecular docking is the initial computational filter used to identify peptide candidates that can bind to a defined site on the viral envelope protein [20, 21]. For viral entry inhibitors, the most common target sites include the receptor-binding domain (RBD), the heptad repeat (HR) regions that mediate six-helix bundle formation, and the fusion peptide (FP) that inserts into the host cell membrane [1, 4, 5].

Docking algorithms evaluate the complementarity between the peptide and the protein surface using scoring functions that account for van der Waals forces, electrostatic interactions, hydrogen bonding, and desolvation penalties [15, 16]. Rigid-body docking, where the protein is kept static and the peptide is flexible, is often used for initial screens due to its computational efficiency [20, 22]. More accurate results are obtained with semi-flexible or fully flexible docking protocols that allow side-chain and backbone rearrangements of the binding site residues [10, 17].

A study targeting the Middle East Respiratory Syndrome coronavirus spike fusion machinery used in silico exploration of the heptad repeat 2 (HR2) domain to design antiviral peptides [1]. Docking simulations identified peptides that formed stable interactions with the conserved HR1 groove, a region critical for membrane fusion. Similarly, computational docking was applied to target the White Spot Syndrome Virus envelope protein in a study focused on shrimp aquaculture [5]. The identified peptides showed favorable binding energies against key envelope protein epitopes.

For influenza A virus, in silico designed peptides targeting hemagglutinin were evaluated using docking and later confirmed in in vitro infection assays [20]. The computational screen identified peptides that bound to the conserved stem region of hemagglutinin, preventing the low pH-induced conformational change required for fusion. These examples underscore the value of docking in generating testable hypotheses for diverse veterinary viral pathogens.

Molecular Dynamics Simulations for Refinement and Stability Assessment

Following docking, molecular dynamics (MD) simulations are employed to refine the binding pose and assess the dynamic stability of the peptide-protein complex [3, 7]. MD simulations model the physical movements of atoms over time, providing a realistic picture of how the complex behaves in a solvated environment [10, 12].

The process involves embedding the docked complex in a periodic water box, adding counterions to neutralize the system, and applying a molecular mechanics force field (e.g., CHARMM, AMBER, GROMACS) [3]. Simulations are typically run for tens to hundreds of nanoseconds. Key metrics extracted from the trajectory include the root-mean-square deviation (RMSD) of the peptide and protein backbone, which indicates structural stability [12]. A low and stable RMSD suggests a well-bound complex. Root-mean-square fluctuation (RMSF) can identify flexible regions in both the peptide and the target that may be important for binding [18].

MD simulations have been used to study the interaction of computationally optimized peptides targeting the SARS-CoV-2 spike protein [14]. The simulations revealed that the designed peptides maintained stable contacts with the RBD over the simulation time scale. Another study used MD to analyze the stability of stapled peptides designed to target the SARS-CoV-2 RBD [12]. The simulations showed that stapling, which cyclizes the peptide through a covalent linker, reduced conformational entropy and enhanced binding stability compared to linear counterparts.

For veterinary applications, MD simulations have been critical in refining peptide inhibitors against the severe fever with thrombocytopenia syndrome virus glycoprotein Gn [8]. The simulations validated that the designed peptides bound to a conserved pocket on Gn with high stability, suggesting a strong potential for viral entry inhibition. A detailed discussion of these techniques is available in the article Molecular Dynamics Simulations of Viral Envelope Proteins: Insights into Host Recognition and Drug Design.

Free Energy Calculations and Binding Affinity

Quantitative estimation of binding free energy is a major goal of computational design. Methods such as MM-GBSA and MM-PBSA are widely used to rank peptide candidates based on their predicted binding affinity to the envelope protein [10, 12]. These methods use a combination of molecular mechanics energies (electrostatic and van der Waals), solvation free energies (using generalized Born or Poisson-Boltzmann models), and entropy approximations (from normal mode analysis) to estimate the free energy of binding [3].

A study by de Campos et al. applied MM-GBSA calculations to evaluate the binding of stapled peptides to the SARS-CoV-2 RBD [12]. The results showed that specific stapled peptides had significantly more negative binding energies than non-stapled controls, correlating with their improved inhibitory potential. In a similar vein, Wang and colleagues used MM-PBSA to evaluate the binding of triterpenoid-nucleoside conjugates targeting coronavirus membrane fusion [3]. The calculations identified compounds with low nanomolar predicted affinities.

Free energy calculations are also used to predict the impact of mutations in viral envelope proteins on peptide binding [2]. As viral variants emerge, computational methods can rapidly assess whether designed peptides remain effective. This aspect is directly relevant to understanding viral evolution, as discussed in Computational Modeling of Viral Glycoprotein Evolution: Predicting Antigenic Drift Using Machine Learning and In Silico Profiling of Viral Receptor-Binding Domain Evolutionary Trajectories.

Machine Learning-Based Peptide Screening

Machine learning (ML) has emerged as a powerful tool to accelerate peptide screening [2, 9, 13]. ML models are trained on large datasets of known active and inactive peptides against specific viral targets or on general antiviral peptide databases [17]. Features used for training can include peptide physicochemical properties (hydrophobicity, charge, isoelectric point), structural descriptors (secondary structure propensity, solvent accessibility), and sequence composition [9, 13].

A high-throughput pipeline was developed by Wolfe et al. that combined computational screening with experimental validation to design peptides targeting the SARS-CoV-2 spike protein [7]. The pipeline used a machine learning classifier to prioritize peptides from a large virtual library for experimental testing. The top hits from the ML screen showed high affinity for the spike protein in surface plasmon resonance experiments.

Sakib and colleagues performed a computational screening of 645 known antiviral peptides against the SARS-CoV-2 RBD [9]. They used molecular docking and machine learning-based scoring to rank the peptides. Several peptides with known activity against other viruses were identified as strong candidates against SARS-CoV-2, demonstrating the potential for cross-reactive peptide therapeutics. Similarly, Egieyeh et al. used a drug repurposing strategy combined with machine learning to predict peptide-based drugs that could inhibit the spike-ACE2 interaction [13].

For veterinary virology, ML methods are being adapted to screen peptides against emerging swine and avian viral pathogens [17]. The ability to rapidly screen thousands of candidate sequences in silico is particularly valuable for responding to outbreaks of novel viruses where no pre-existing therapeutics exist.

Peptide Modification and Optimization Strategies

Once lead peptide candidates are identified, computational methods guide further optimization to improve their drug-like properties [8, 12]. Key modifications include:

• Cyclization (Stapling): Introducing a covalent bond between non-adjacent residues to stabilize the alpha-helical conformation [12]. This can increase binding affinity, proteolytic stability, and cell permeability. Computational tools can predict optimal stapling positions and bridge lengths [12, 18].
• Cholesterol Conjugation: Adding a cholesterol moiety to the peptide C-terminus to enhance membrane anchoring and pharmacokinetic properties [19]. This approach has been used to improve the in vivo half-life of peptide fusion inhibitors.
• Peptidomimetic Design: Replacing peptide bonds with non-natural isosteres to increase metabolic stability [2, 3]. Computational fragment-based design can identify peptidomimetic scaffolds that mimic the critical side-chain interactions of the original peptide [3].
• Mutation Scanning: In silico alanine scanning or saturation mutagenesis can identify which residues are most critical for binding [6, 21]. This information is used to design minimal active motifs or to introduce substitutions that enhance affinity [21].

These optimization strategies are essential for translating computational hits into viable therapeutic candidates. The design of self-inhibitory peptides derived directly from the envelope protein sequences is another computationally guided approach [21]. These peptides mimic conserved regions of the viral protein and compete for binding sites involved in oligomerization or receptor engagement [21, 22].

Targeting Specific Viral Envelope Proteins

The computational design principles are broadly applicable across different virus families. Examples from the literature include:

| Virus Family | Target Envelope Protein | Computational Approach | Key Reference(s) | | :-, | :-, | :-, | :-, | | Coronaviridae | Spike (RBD, HR1/HR2) | Docking, MD, MM-GBSA, ML | [1, 2, 3, 4, 6, 7, 9, 10, 11, 12, 13, 14, 15, 16, 17] | | Orthomyxoviridae | Hemagglutinin (stem, RBS) | Docking, MD | [20] | | Filoviridae | GP (primed form) | Docking, MD | [18] | | Flaviviridae | Envelope (E) protein | Docking, MD, optimization | [22] | | Bunyaviridae | Glycoprotein Gn | Docking, MD | [8] | | Nimaviridae | Envelope protein (VP28) | Docking | [5] |

For influenza, the in silico design of peptides against hemagglutinin has been explored, with peptides targeting both the receptor-binding site and the conserved stem region [20]. For the Dengue virus, de novo design and structural optimization of entry inhibitory peptides targeting the envelope protein dimer interface have been reported [22]. For Ebola virus, novel cyclo-peptides were designed in silico to target the primed GP protein [18]. Each of these studies demonstrates a workflow involving target identification, computational screening, and experimental validation.

Conclusions and Future Directions

Computational design of antiviral peptides targeting viral envelope proteins is a mature and rapidly evolving field. The integration of molecular docking, molecular dynamics simulations, free energy calculations, and machine learning provides a powerful framework for rapid candidate discovery and optimization [1, 2, 3, 7, 9, 10]. The application of these methods to veterinary pathogens is of great importance for controlling diseases in livestock, poultry, and aquaculture species [5, 8].

Future developments will likely include the use of deep learning for de novo peptide generation, the integration of protein language models for predicting peptide-protein interactions, and the development of multi-target peptide designs that can neutralize diverse viral variants [2, 14]. The use of advanced sampling techniques in MD, such as metadynamics and replica exchange, will improve the estimation of binding free energies for particularly flexible targets [3]. Furthermore, linking computational predictions with high-throughput experimental platforms (e.g., phage display, peptide microarrays) will accelerate the validation pipeline [7].

Researchers are encouraged to explore the interactive 3D structures of peptide-protein complexes in public databases such as the Protein Data Bank (PDB) to visualize the binding modes discussed in this review. The related portal article Structural Bioinformatics of Viral Envelope Proteins and Entry Mechanisms provides additional context for understanding these molecular interactions.

References

[1] Alotaiq N, Dermawan D, Chtita S. Targeting Middle East Respiratory Syndrome Coronavirus Spike Fusion Machinery With Antiviral Peptides: In Silico Exploration of the Heptad Repeat 2 Domain. Microbiologyopen. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/42082899/

[2] Losery SPM, Laddha K, Martinek TA et al. Computational insights and impact of combinatorial peptidomimetics on immune escape SARS-CoV-2 variants. Expert Opin Drug Discov. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41731696/

[3] Wang M, Zhu X, Wang S et al. Predictive modeling and fragment-based design of triterpenoid-nucleoside conjugates targeting coronavirus membrane fusion. Eur J Med Chem. 2026. URL: https://pubmed.ncbi.nlm.nih.gov/41443084/

[4] Almabhouh S, Cecon E, Basubas F et al. Computational Design and Evaluation of Peptides to Target SARS-CoV-2 Spike-ACE2 Interaction. Molecules. 2025. URL: https://pubmed.ncbi.nlm.nih.gov/40333723/

[5] Panchalingam S, Kasivelu G. A computational approach to identifying peptide inhibitors against White Spot Syndrome Virus: Targeting the virus envelope protein. Microb Pathog. 2024. URL: https://pubmed.ncbi.nlm.nih.gov/39147215/

[6] Pourmand S, Zareei S, Shahlaei M et al. Inhibition of SARS-CoV-2 pathogenesis by potent peptides designed by the mutation of ACE2 binding region. Comput Biol Med. 2022. URL: https://pubmed.ncbi.nlm.nih.gov/35688710/

[7] Wolfe M, Webb S, Chushak Y et al. A high-throughput pipeline for design and selection of peptides targeting the SARS-Cov-2 Spike protein. Sci Rep. 2021. URL: https://pubmed.ncbi.nlm.nih.gov/34741099/

[8] Yuan SF, Wen L, Chik KK et al. In Silico Structure-Based Design of Antiviral Peptides Targeting the Severe Fever with Thrombocytopenia Syndrome Virus Glycoprotein Gn. Viruses. 2021. URL: https://pubmed.ncbi.nlm.nih.gov/34696477/

[9] Sakib MMH, Nishat AA, Islam MT et al. Computational screening of 645 antiviral peptides against the receptor-binding domain of the spike protein in SARS-CoV-2. Comput Biol Med. 2021. URL: https://pubmed.ncbi.nlm.nih.gov/34403938/

[10] Pei P, Qin H, Chen J et al. Computational design of ultrashort peptide inhibitors of the receptor-binding domain of the SARS-CoV-2 S protein. Brief Bioinform. 2021. URL: https://pubmed.ncbi.nlm.nih.gov/34180984/

[11] Bibilashvili RS, Sidorova MV, Dudkina US et al. [Peptide inhibitors of the interaction of the SARS-CoV-2 receptor-binding domain with the ACE2 cell receptor]. Biomed Khim. 2021. URL: https://pubmed.ncbi.nlm.nih.gov/34142531/

[12] de Campos LJ, Palermo NY, Conda-Sheridan M. Targeting SARS-CoV-2 Receptor Binding Domain with Stapled Peptides: An In Silico Study. J Phys Chem B. 2021. URL: https://pubmed.ncbi.nlm.nih.gov/34114829/

[13] Egieyeh S, Egieyeh E, Malan S et al. Computational drug repurposing strategy predicted peptide-based drugs that can potentially inhibit the interaction of SARS-CoV-2 spike protein with its target (human ACE2). PLoS One. 2021. URL: https://pubmed.ncbi.nlm.nih.gov/33417604/

[14] Chatterjee P, Ponnapati M, Kramme C et al. Targeted intracellular degradation of SARS-CoV-2 via computationally optimized peptide fusions. Commun Biol. 2020. URL: https://pubmed.ncbi.nlm.nih.gov