Discovery of bioactive inhibitors targeting onion yellow dwarf virus coat protein based on molecular docking and simulation Authors: Amina Irfan, Mudassar Fareed Awan, Qudsia Naz, Muhammad Farhan Sarwar, Somdutt Mujwar, Iqra Mubeen, Safia Obaidur Rab Published: 2025 ABSTRACT AbstractBackgroundOnion yellow dwarf virus (OYDV), a devastating pathogen affecting garlic (Allium sativum), causes significant agricultural losses worldwide. Current management strategies rely on chemical pesticides that pose environmental and toxicity concerns, hence emphasizing the urgent need for eco-friendly alternatives.ResultsThis study aimed to discover plant-derived inhibitors of OYDV by targeting its coat protein through several computational approaches. The CP of OYDV was characterized, and a high-quality 3D model was validated. Through high-throughput molecular docking of 46 plant-based metabolites, four compounds were selected: hypericin, celastrol, sanguinarine, and asiatic acid with superior binding affinities (below − 9 kcal/mol). These four had exceptionally high docking scores, outperforming the commercial antivirals ribavirin and ningnanmycin. These top ligands have been proven to be less hazardous via in silico toxicity profiling, which makes them suitable for the development of biopesticides. DFT analysis and MD simulations validated the stability and structural flexibility of the hypericin and coat protein, respectively.ConclusionThese computational findings suggest hypericin could be a promising lead compound for developing sustainable OYDV management strategies.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12870-025-07745-7.Keywords: OYDV, Plant metabolites, Virtual screening, Molecular dynamics simulations, Antiviral agents Introduction Plant viruses pose a serious threat to global agriculture, affecting crop yield and quality and resulting in significant economic losses in ornamental, fruit, vegetable, and grain crops [1]. Among these, onion yellow dwarf virus, a member of the Potyvirus genus, predominantly infects Allium species, i.e., onion and garlic [2, 3]. It has been demonstrated that the virus can induce slight to severe leaf distortions and up to a 60% decrease in bulbs of garlic [4]. About eleven mature proteins, i.e., P1, HCPro, VPg, P3, CI, 6K1, 6K2, NIaPro, NIb, CP, and P3N-PIPO, are produced by proteolytic processing of a large polyprotein encoded by the Potyvirus genome. To ensure a successful infection, these proteins work together to coordinate viral movement, replication, and host interactions [5, 6]. To effectively control the spread of the corresponding diseases, it is crucial to neutralize the effects of any of these proteins or block their functional domains. Targeting the coat protein of OYDV is critical for developing novel management strategies. It plays an essential role not only in encapsidating the viral genome but also in systemic movement and aphid transmission [5]. Inhibiting the CP could therefore offer a potent strategy to control the virus. Although ribavirin [7] and ningnanmycin [8] are used to control several plant viruses, their efficacy against many viral diseases has not yet been fully validated. Given the current situation, evaluating a suitable molecule or compound is essential due to the huge demand at this moment. Research on several plant-based metabolites, i.e., alkaloids, phenolics, terpenoids, flavonoids, peptides, lignans, quinones, amines, and anthocyanins, has revealed their potential to fight off bacterial, viral, and fungal phytopathogens, which are essential to the defense response of plants [9]. However, it takes years to synthesize novel molecules and test them in vitro. Therefore, as a preliminary investigation, phytochemicals can be tested in silico to cut down on labor-intensive and expensive work. Computational methods like molecular docking and simulation studies provide effective and affordable alternatives for the development of antiviral molecules. For instance, anonaine, a plant alkaloid, has been recognized through large-scale molecular docking of 2851 metabolites, finding anonaine as a potential inhibitor for begomovirus βC1 protein, with further confirmation provided through MD simulations [10]. Similarly, another study employed molecular docking and MM–GBSA to evaluate 61 metabolites of cucurbits, discovering amentoflavone and quercetin as dual-target inhibitors that bind to both the Cucumber Mosaic Virus coat protein (CP) and the aphid vector’s cuticle protein (MPCP4) simultaneously, thereby potentially obstructing the viral function and insect transmission [11]. A targeted computational investigation for the identification of plant-based inhibitors against the OYDV coat protein has not been conducted. This study was designed to bridge this critical gap. In the context of this study, before molecular docking, a comprehensive characterization of the OYDV coat protein was performed. Molecular docking was employed to screen candidate phytochemicals against the CP, to find a suitable one that would protect against the target onion yellow dwarf virus infecting various Allium species. The stability of the protein–ligand complexes was then verified by a molecular dynamics analysis. However, thorough in vitro tests to evaluate the compound’s effectiveness against viral replication and infection are necessary to further validate its antiviral activity. These tests should be followed by further experiments to assess the safety and efficacy in the host plant under normal conditions. This work establishes a computational foundation for developing specific eco-friendly and sustainable biopesticides against OYDV. Materials and methods Retrieval and sequence analysis of coat protein The CP nucleotide sequence of OYDV (NCBI accession ID: MK336983.1) was available in FASTA format by the NCBI GenBank (https://www.ncbi.nlm.nih.gov/). The amino acid sequence of the CP was then ascertained by submitting the nucleotide sequence to NCBI ORFfinder (https://www.ncbi.nlm.nih.gov/orffinder/). Additionally, after the blast analysis tool was used to confirm the amino acid sequence, the NCBI GenBank database was used to obtain related CP sequences. Multiple sequence alignment was used to identify the conserved amino acids in CP sequences. The evolutionary relationships between the retrieved CP sequences were then investigated using MEGA 11 for phylogeny construction [12]. Analysis of physicochemical parameters Physicochemical parameters, i.e., molecular weight, extinction coefficient, theoretical amino acid composition, instability index, isoelectric point, aliphatic index, and grand average of hydropathicity were determined using the ExPASy Protparam server (https://web.expasy.org/protparam/) [13]. Prediction of functional sites in coat protein Online servers were used to find functional sites in coat protein, such as phosphorylation and glycosylation sites [14]. NetNGlyc 1.0 server (https://services.healthtech.dtu.dk/services/NetNGlyc-1.0/) [15] with a threshold of 0.5 was used to predict the glycosylation sites, while phosphorylation sites were predicted using NetPhos 3.1 server (https://services.healthtech.dtu.dk/services/NetPhos-3.1/) [16], selecting the score >0.5, considering all three residues (Tyr, Thr, Ser) for both generic and kinase-specific predictions. Gene ontology, secondary structure, and motif prediction At the biological, molecular, and cellular levels, the coat protein’s gene ontology or function was predicted using the PANZZER2 web server (http://ekhidna2.biocenter.helsinki.fi/sanspanz/) [17], with positive predicted values. Alpha helices, random coils, beta strands, beta turns, and other secondary structural features of the coat protein were all analyzed in detail using the SOPMA web server (https://npsa.lyon.inserm.fr/cgibin/npsa_automat.pl?page=/NPSA/npsa_sopma.html.) [18]. The results are displayed on this server in a variety of styles (bars and peaks). By choosing a maximum of 10 motifs that are crucial to protein structure and function, MEME suite version 5.5.1 (https://meme-suite.org/meme/) [19] was used to investigate motifs, which are the small sequences of amino acids in coat protein sequences. 3D structure prediction, quality assessment, and active site prediction First, trRosetta (https://yanglab.qd.sdu.edu.cn/trRosetta/) [20], an algorithm for protein structure prediction based on direct energy minimizations, was used to build the three-dimensional structure of the CP. Additionally, for comparative structure analysis, the coat protein structure was also predicted using ColabFold (v1.5.5), i.e., AlphaFold2 using MMseqs2 [21]. Ramachandran plots of CP predicted by trRosetta and ColabFold (v1.5.5) were generated with PROCHECK from the SAVES 6.0 server (https://saves.mbi.ucla.edu/) [22] and were used to assess the structure’s quality. ERRAT [23] was used to confirm protein structures determined via crystallography by using PROCHECK, which was previously used to build a Ramachandran plot. For active site prediction in the CP, the CASTpFold server (https://cfold.bme.uic.edu/castpfold/compute) [24] was used. This server recognizes and analyzes surface pockets and internal cavities within a protein structure. Retrieval of plant secondary metabolites The ZINC 15 database (https://zinc15.docking.org/) [25] provided a total of 46 plant secondary metabolites in a structured data file (SDF) format. Alkaloids, napthoquinones, amino acids, polyketides, flavonoids, tannins, skimate, terpenoids, phenypropanoids, and other miscellaneous compounds reported from different plant species are among the many phytochemical groups represented by the plant secondary compounds included in this extensive database. The PubChem database (https://pubchem.ncbi.nlm.nih.gov/) [26] was used to retrieve the 3D structures of ribavirin and ningnanmycin, two well-known antiviral drugs, as a positive control. Preparation of receptors, ligands, and virtual screening The CP (receptor) was submitted to BIOVIA Discovery Studio (https://discover.3ds.com/discovery-studio-visualizer-download) [27] in PDB format. Briefly, the protein was primed for docking by removing water molecules. PyRx software [28], which provides a handy feature that permits simultaneous energy minimization with the help of the OpenBabel GUI, was also used to minimize the energy of plant secondary metabolites (ligands). The ligands were finally converted into PDBQT format. The protein–ligand docking investigation was started by creating a grid box for the centers at the x, y, and z coordinates: 25.7107 Å, 27.9591 Å, and − 28.1848 Å, respectively. The exhaustiveness was set to be 8. The docking analysis was carried out in PyRx software, which used AutoDock Vina as a docking machine, a virtual screening tool that enables extensive screening of ligands against a target protein. Among the docked poses included in the output, the one with the top binding affinity (kcal/mol) value and the least root mean square deviation was deemed the best pose. It was then further visualized in BIOVIA Discovery Studio and PyMOL [29]. Toxicity prediction Regular monitoring of the effects of plant secondary metabolites on human health and the environment in relation to feed and food safety is necessary for the evaluation and assessment of these chemicals as antiviral agents in green agriculture. In silico investigations, which use computational techniques to provide scientific data quickly, offer a practical and effective substitute for experimental procedures. The evaluation of physicochemical properties and toxicity of the chosen secondary metabolites, including targeted pathways, toxicological endpoints, and median lethal dosage (LD50), was ascertained using the ProTox 3.0 server (https://tox.charite.de/protox3/) [30]. FMOs and DFT analyses The electronic structure of hypericin was investigated using frontier molecular orbital and density functional theory analyses using ORCA 6.1 [31]. Subsequent DFT calculations were performed using the BP functional with the def2-SVP basis set, employing the RI approximation and def2/J auxiliary basis under TightSCF convergence criteria. With the optimized structure, frontier orbital energies (HOMO and LUMO) were calculated to determine the energy gap (ΔE = ELUMO − EHOMO) and associated molecular reactivity descriptors. Using the calculated orbital energies, additional molecular properties were determined, such as chemical hardness (η = ΔE/2), chemical softness (S = 1/2η), chemical potential (µ = (ELUMO + EHOMO)/2), and electrophilicity index (ω = µ²/2η). The ligand’s stability, reactivity, electron transfer, and electronic interaction potential with the target protein were described using these properties [32]. Molecular dynamics simulations Hypercin was chosen for MD simulation based on its docking score and confirmed chemical interactions with the target coat protein receptor. MD simulation of the macromolecular complex of coat protein with complexed Hypercin was performed for 100 ns employing the Desmond module of Schrodinger’s software. Utilization of explicit solvent molecules, subsequently neutralization with the corresponding ions. To minimize the system’s energy, the steepest-descent methodology was implemented to relax it and eliminate any steric conflicts or inadequate connections between atoms. A brief series of low-temperature constant-pressure (NPT) simulations was performed to bring the system into equilibrium. The system was subject to positioning constraints as well as a progressively raised temperature. This maximizes the possibility that the system will be stable and balanced before the simulation; to get the necessary results, the simulation continued for 100 ns while considering the system’s energies, atom positions, and RMSD values. This contributed to a better understanding of the system’s dynamic behaviour while providing long-term insights into the complex’s structure and functionality stability (Fig. 1). The representation of the schematic workflow followed in this study Results Retrieval of protein and its analysis A 212 amino acid-long ORF encoding CP was obtained from NCBI ORFfinder. The target CP sequence was aligned with 13 other CP sequences that were obtained from different potyviruses associated with several diseases, i.e., Watermelon, turnip, sugarcane, lettuce mosaic disease, and many more. The potyviruses associated with various crops across different locations and their CP sequences were obtained from NCBI GenBank and are summarized in Table 1 and TableS1. The multiple sequence alignment of the coat protein indicated conserved regions across the sequences. About 30.3% of the amino acids (i.e. amino acids present at location 48, 61, 64, 66, 68, 73, 77, 84, 92, 99, 107, 108, 109, 110, 111, 113, 114, 116, 121, 125, 126, 127, 131, 138, 143, 145, 148, 149, 150, 151, 153, 154, 155, 158, 159, 161, 162, 165, 171, 173, 174, 175, 179, 186, 188, 189, 191, 192, 193, 194, 200, 203, 204, 206, 207, 210, 211, 212, 213, 214, 215, 224, 225, 227, 228, 235, 237, 238, 239, 240, 241, 244, 250, and 254) present in the CP’s active sites were conserved across all the retrieved CP. The constructed phylogenetic tree revealed a close relationship of onion yellow dwarf virus coat protein with Passionfruit woodiness virus and Watermelon mosaic virus, and is represented in Fig. 2. Table 1Retrieved accession IDs and respective hosts from NCBI Swiss-Prot for coat protein analysisSr.no.NCBI Accession IDsOrganismHost 1 QHI06022Onion yellow dwarf virusGarlic 2 P20177Clover yellow vein virusCoriander, Wild carrot, Soybean, European yellow lupine, Garden pea, Clover 3 P89509Turnip mosaic virusGarlic mustard, Brassica, Alsike clover, Common chickweed, Shepherd’s purse 4 P17765Bean yellow mosaic virusPeanut, Soybean, Garden pea, Clover, French bean 5 P04517Tobacco etch virusCapsicum annuum, Common tobacco, Tomato, Jimsonweed 6 P19724Papaya ringspot virusWatermelon, Summer squash, Bitter gourd 7 Q89251Watermelon mosaic virus 2Muskmelon, Watermelon, Cucumber, Summer squash 8 P29152Pea seed-borne mosaic virusGarden pea 9 P31999Lettuce mosaic virusChickpea, Garden pea, Spinach, Bluebells, Chicory, Endive 10 Q02963Potato virus YPotato, Capsicum, Nicotiana, Tomato 11 P07993Pepper mottle virusCapsicum pepper, Datura inoxia, Solanum 12 P25242Sugarcane mosaic virus SCMaize, Eleusine, Panicum, Saccharum, Setaria 13 P32652Maize dwarf mosaic virusJohnson grass 14 P32575Passionfruit woodiness virusSoybean, Lentil, Passion fruit, French bean, Bell apple Retrieved accession IDs and respective hosts from NCBI Swiss-Prot for coat protein analysis Phylogenetic tree of amino acid sequences of CP associated with Onion yellow dwarf virus with other selected respective potyviruses using the Neighborhood Joining method and the Poisson model with 500 bootstrap replicates in MEGA 11 software Analysis of physicochemical properties The physicochemical properties of the coat protein were determined to predict its structural and functional characteristics. The molecular weight was calculated to be 29,175.93 Da, and the isoelectric point (pI) was 6.28. Analysis of the amino acid composition revealed 35 negatively charged and 33 positively charged residues, resulting in a slight net negative charge. The extinction coefficient was measured as 32,890 M− 1 cm− 1 (280 nm in water). The instability index (II) was computed to be 30.90, and the aliphatic index (AI) was 67.24. Additionally, the grand average of hydropathicity was determined to be − 0.665. Prediction of functional sites in coat protein Phosphorylation sites in the CP were determined using the NetPhos 3.1 server, revealing 30 potential sites for both specific and nonspecific kinases, involving serine, threonine, and tyrosine residues, represented in Fig. 3. Additionally, six N-glycosylation sites were identified in the protein sequence using the NetNGlyc 1.0 server, represented in Figs. 4 and 5. This figure illustrates the identified phosphorylation sites as predicted through computational analysis using the NetPhos 3.1 server This figure illustrates the identified N-glycosylation sites as predicted through computational analysis using the NetNGlyc 1.0 server This figure represents the Asn-Xaa-Ser/Thr sequons (Blue) and asparagines predicted to be N-glycosylated (Red) in the sequence Analysis of gene ontology, secondary structure, and motif identification Gene ontology of the coat protein was conducted using the PANZZER2 web server for the protein’s biological process, cellular component, and molecular function prediction. For the biological process classification, the protein was annotated with RNA-templated transcription (GO: 0001172). In the molecular function category, the prediction identified RNA-dependent RNA polymerase activity (GO: 0003968). The cellular component analysis categorized the protein as a viral capsid (GO: 0019028). The secondary structure composition of the coat protein was predicted using the SOPMA online server. The analysis showed that random coils constitute the predominant structural element, accounting for 54.86% of the protein’s secondary structure. Alpha helices represented a significant structural component with 35.41%, while extended strands comprised a lower proportion of 9.73%. Based on comprehensive amino acid sequence analysis, 10 motifs were identified using the MEME suite version 5.5.1, all diverse at the sequence level. The motifs are presented in Fig. 6 with a color-coded representation for easier differentiation. This figure illustrates the identified motif sequences and their precise locations within the protein, as predicted through computational analysis using MEME suite (v5.5.1) Tertiary structure prediction, quality assessment, and active site identification The 3D structure of CP linked to onion yellow dwarf disease was predicted using trRosetta and ColabFold (v1.5.5), represented in Fig. 7. Model quality assessment of the CP structure predicted by trRosetta and ColabFold (v1.5.5) was performed through Ramachandran plot analysis. It revealed that 92.2% of the amino acid residues were in the most favored regions, while the remaining 7.8% were in additionally allowed regions in the trRosetta predicted CP model, as presented in Fig. 8; Table 2. For the CP model predicted by ColabFold (v1.5.5), the Ramachandran plot analysis revealed that 67.7% of the amino acid residues were in the most favored regions, 18.1% were located in additionally allowed regions, 8.2% of amino acid residues covered generously allowed regions, while the remaining 6.0% were located in disallowed regions Fig. 9; Table 2. Three-dimensional (3D) predicted structure of OYDV coat protein by (A) trRosetta and (B) ColabFold (v1.5.5) Model validation of the OYDV coat protein predicted by trRosetta. (A) Ramachandran plot generated by PROCHECK and (B) ERRAT plot showing regions of possible structural errors, where yellow regions indicate areas with potential structural errors at the 95% confidence level Table 2Statistical details of the Ramachandran plot for the OYDV coat protein structure predicted by trrosetta and ColabFold (v1.5.5)trRosettaColabFold (v1.5.5)PropertiesNo of residuesResidues percentageNo of residuesResidues percentageResidues in most favoured regions [A, B, L]21492.2%15767.7%Residues in additional allowed regions [a, b, l, p]187.8%4218.1%Residues in generously allowed regions [~ a, ~b, ~l, ~p]00%198.2%Residues in disallowed regions00%146.0%Number of non-glycine and non-proline residues232232Number of end-residues (excl. Gly and Pro)22Number of glycine residues (shown as triangles)1414Number of proline residues99Total number of residues257257 Statistical details of the Ramachandran plot for the OYDV coat protein structure predicted by trrosetta and ColabFold (v1.5.5) Model validation of the OYDV coat protein predicted by ColabFold (v1.5.5). (A) Ramachandran plot generated by PROCHECK and (B) ERRAT plot showing regions of possible structural errors, where yellow and red regions indicate areas with potential structural errors at the 95% and 99% confidence level, respectively Further validation using ERRAT yielded an overall quality score of 97.0443 and 81.395, well above the accepted threshold of > 50 for high-quality models, as can be observed in Figs. 8 and 9. The potential binding sites of the coat protein were predicted using the CASTpFold server, which identified 23 distinct binding pockets, out of which the top five were represented in Table 3 and Table S2. For docking studies, the largest binding site encompassed 244 amino acid residues was selected, covering nearly the entire protein structure except for 13 residues located at positions: ALA1, GLY2, SER14, ASP26, ALA27, GLY28, THR29, THR30, GLY31, LYS32, PHE33, THR34, and LYS44. Table 3The pocket information of the top five predicted active sites of the coat protein model generated by trrosetta was identified using the castp serverPocket IDArea (SA) (Å2)Volume (SA) (Å3)No. of Residues Pocket ID: 1 13232.90618311.947244 Pocket ID: 2 21.16527.8445 Pocket ID: 3 26.00026.9936 Pocket ID: 4 4.77224.8773 Pocket ID: 5 18.70313.4675 The pocket information of the top five predicted active sites of the coat protein model generated by trrosetta was identified using the castp server Molecular Docking of coat protein with plant secondary metabolites In the present investigation, an extensive virtual screening of 46 plant-derived secondary metabolites and established positive controls was conducted, targeting the coat protein of the viral pathogen using the PyRx software. The strength of protein–ligand interactions was quantitatively assessed by the calculated binding affinity (kcal/mol), as shown in Table S3. The reference antiviral compounds ningnanmycin and ribavirin demonstrated binding affinities of − 7.6 kcal/mol and − 6.8 kcal/mol, respectively. Among the screened phytochemicals, hypericin exhibited the highest binding affinity (− 9.9 kcal/mol), followed by celastrol (− 9.7 kcal/mol), sanguinarine (− 9.3 kcal/mol), and asiatic acid (− 9.1 kcal/mol), as shown in Table 4 and Table S3. Table 4Docking scores (binding affinity) and molecular interactions of top-ranked plant secondary compounds and positive controls docked against the coat protein encoded by onion yellow Dwarf virusSr. No.ZINC ID/PubChem IDName of the compoundBinding Affinity (kcal/mol)Molecular interactions1ZINC3780340Hypericin−9.9π–Sigma: LYS247π–Alkyl: LYS247π–cation: LYS247Conventional Hydrogen Bond: SER231, ARG239, THR241, ASP244Carbon–Hydrogen Bond: SER231Van der Waals: THR233, GLU234, VAL230, ASP227, HIS240, ALA242, ILE245, ASN2462ZINC19795938Celastrol−9.7π–π Stacked: HIS240Conventional Hydrogen Bond: GLN210, ASP236, ARG239, THR241, ASP244Carbon–Hydrogen Bond: ALA242Van der Waals: ASP227, ASN229, VAL230, SER231, THR232, THR237, GLU235, GLU234, ILE245, LYS247,3ZINC706Sanguinarine−9.3π–π Stacked: HIS240π–Alkyl: LYS247Conventional Hydrogen Bond: ARG239Van der Waals: ASP227, ASP244, ASN229, VAL230, SER231, GLU235, GLU234, ASP2364ZINC8221271Asiatic acid−9.1Conventional Hydrogen Bond: GLU234,GLN210Carbon–Hydrogen Bond: SER231Van der Waals: THR233, GLU235, LYS247, VAL230, ASP227, HIS240, THR241, ALA242, ASP244, ILE245, GLY228, ASN229, ARG239, ASP236,5PubChem CID44588235Ningnanmycin−7.6Unfavorable Donor–Donor: ARG223Conventional Hydrogen Bond: GLN198, HIS224, ASN168, ARG223Carbon–Hydrogen Bond: ASP220, ASP213, HIS224Van der Waals: GLU222, GLU219, LEU169, ALA226, ARG167, THR218, ASP227, ASN215, ASN231, LYS230, SER217, VAL216, THR2256PubChem CID37542Ribavirin−6.8Conventional Hydrogen Bond: LYS187, GLN12, GLU147, ARG176, TYR182, THR185Van der Waals: SER186, THR188, TYR177, THR14, PRO189, ILE62, VAL13 Docking scores (binding affinity) and molecular interactions of top-ranked plant secondary compounds and positive controls docked against the coat protein encoded by onion yellow Dwarf virus π–Sigma: LYS247 π–Alkyl: LYS247 π–cation: LYS247 Conventional Hydrogen Bond: SER231, ARG239, THR241, ASP244 Carbon–Hydrogen Bond: SER231 Van der Waals: THR233, GLU234, VAL230, ASP227, HIS240, ALA242, ILE245, ASN246 π–π Stacked: HIS240 Conventional Hydrogen Bond: GLN210, ASP236, ARG239, THR241, ASP244 Carbon–Hydrogen Bond: ALA242 Van der Waals: ASP227, ASN229, VAL230, SER231, THR232, THR237, GLU235, GLU234, ILE245, LYS247, π–π Stacked: HIS240 π–Alkyl: LYS247 Conventional Hydrogen Bond: ARG239 Van der Waals: ASP227, ASP244, ASN229, VAL230, SER231, GLU235, GLU234, ASP236 Conventional Hydrogen Bond: GLU234,GLN210 Carbon–Hydrogen Bond: SER231 Van der Waals: THR233, GLU235, LYS247, VAL230, ASP227, HIS240, THR241, ALA242, ASP244, ILE245, GLY228, ASN229, ARG239, ASP236, Unfavorable Donor–Donor: ARG223 Conventional Hydrogen Bond: GLN198, HIS224, ASN168, ARG223 Carbon–Hydrogen Bond: ASP220, ASP213, HIS224 Van der Waals: GLU222, GLU219, LEU169, ALA226, ARG167, THR218, ASP227, ASN215, ASN231, LYS230, SER217, VAL216, THR225 Conventional Hydrogen Bond: LYS187, GLN12, GLU147, ARG176, TYR182, THR185 Van der Waals: SER186, THR188, TYR177, THR14, PRO189, ILE62, VAL13 Hypericin formed conventional hydrogen bonds with SER231, ARG239, THR241, and ASP244, while LYS247 formed stabilizing π–alkyl, π–cation, and π–sigma interactions, as well as carbon–hydrogen bond with SER231 and van der Waals interactions. Celastrol made several hydrogen bonds with GLN210, ASP236, ARG239, THR241, ALA242, and ASP244, along with π–π stacking with HIS240. Sanguinarine exhibited π–π stacking with HIS240, π–alkyl contact with LYS247, and a conventional hydrogen bond with ARG239, in addition to van der Waals interactions. Asiatic acid formed hydrogen bonds with SER231, GLU234, and GLN210, along with van der Waals contacts. Among the positive controls, ribavirin formed conventional hydrogen bonds with residues LYS187, GLN12, GLU147, ARG176, TYR182, and THR185, along with van der Waals interactions. Ningnanmycin formed unfavorable donor–donor interaction with ARG223, conventional hydrogen bonds with GLN198, HIS224, ASN168, and ARG223, as well as carbon–hydrogen bonds with ASP220, ASP213, and HIS224. A comprehensive comparison of the interacting residues between these top-ranked phytochemicals and the reference compounds is presented in Table 4; Fig. 10, illustrating the molecular basis for their differential binding affinities. Selected 2D ligand–protein interaction plots as observed in BIOVIA Discovery Studio for the top four ligand molecules and positive controls. Panels. (A) Hypericin, (B) Celasbtrol, (C) Sanguinarine, (D) Asiatic acid, (E) Ningnanmycin, and (F) Ribavirin Toxicity prediction In the current investigation, the toxicological profiles of the top-ranked phytochemicals and positive controls were evaluated using the ProTox 3.0 prediction platform. The server not only generates comprehensive predictions for multiple toxicological endpoints, encompassing acute toxicity (quantified through LD50 values), cytotoxicity, carcinogenic potential, mutagenicity, and immunotoxicity, but also provides information regarding physicochemical parameters. The profile of the physicochemical parameters included molecular weight, lipophilicity (logP), topological polar surface area (TPSA), number of hydrogen bond donors (HBD), and number of hydrogen bond acceptors (HBA), as shown in Table 5. All of the compounds that were tested had molecular weights below 500 Da except hypericin, and logP values ranged from − 3.43 (ningnanmycin) to 6.7 (celastrol). The number of HBDs ranged from 0 (sanguinarine) to 8 (ningnanmycin). The number of HBAs ranged from 4 (celastrol and sanguinarine) to 14 (ningnanmycin). The TPSA values ranged from 40.8 Å2 (sanguinarine) to 244.15 Å2 (ningnanmycin). The number of rotatable bonds ranged from 0 in hypericin and sanguinarine to 10 in ningnanmycin. Molar refractivity values varied between 51.06 (ribavirin) and 142.91 (hypericin). Table 5Physicochemical properties of the best identified ligands with higher binding affinities for the target CPName of the compoundMWlogPHBAHBDNo. of rotatable bondsMolar refractivityTopological polar surface area (TPSA) Å2Hypericin504.445.39860142.91155.52Celastrol450.616.7421131.374.6Sanguinarine332.333.4340094.6840.8Asiatic acid488.75.03542139.2497.99Ningnanmycin443.41−3.4314810101.21244.15Ribavirin244.21−2.3184351.06143.72 Physicochemical properties of the best identified ligands with higher binding affinities for the target CP In toxicity analysis, hypericin (LD50 1425 mg/kg), celastrol (LD502000 mg/kg), sanguinarine (LD50778 mg/kg), asiatic acid (LD502000 mg/kg), and ningnanmycin (LD501000 mg/kg) were classified under GHS Class IV. However, ribavirin (LD502700 mg/kg) was classified under GHS Class V. Notably, none of the evaluated compounds, including both positive controls and phytochemicals, demonstrated predicted cytotoxicity. Both sanguinarine and ribavirin were flagged as potential carcinogens, while three compounds, namely, celastrol, sanguinarine, and asiatic acid, exhibited predicted immunotoxicity. Regarding mutagenic potential, hypericin and sanguinarine were predicted to possess mutagenic properties. The predicted LD50 values ranged from 778 mg/kg for sanguinarine to 2700 mg/kg for ribavirin, as shown in Table 6. Table 6In-silico toxicity prediction of the top four plant secondary metabolites and positive controls having a better binding affinity with coat protein and their LD50 and toxicological endpointsSr. No.Name of the compoundLD50 (mg/kg)Tox ClassCarcinoImmunoMutaCyto 1 Hypericin14254−−−−++−− 2 Celastrol20004−++−−−− 3 Sanguinarine7784 + ++ + −− 4 Asiatic acid20004−−++−−−− 5 Ningnanmycin10004−−−−−−− 6 Ribavirin27005 + −−−−−− In-silico toxicity prediction of the top four plant secondary metabolites and positive controls having a better binding affinity with coat protein and their LD50 and toxicological endpoints FMOs and DFT analyses Density functional theory calculations of hypericin demonstrated a HOMO energy of − 5.865 eV and LUMO energy of − 5.724 eV, yielding a narrow energy gap, i.e., ΔE = 0.141 eV. The chemical hardness (η), chemical softness (S), chemical potential (µ), and electrophilicity index (ω) were 0.0705 eV, 7.09 eV− 1, − 5.795 eV, and 238.2 eV, respectively. MD simulations The coat protein complexes with hypercin have been chosen for molecular dynamics simulations to examine their thermodynamic stability over 100 ns. For the inhibitory response to be potent, the ligand-protein complex must have sufficient stability throughout the ns timeframe range. Consequently, the macromolecular complex performed a 100 ns molecular dynamics simulation with Schrodinger’s Desmond software version 2022.4. The monomeric chain of the coat protein comprised 257 amino acids and 2048 heavy atoms out of a total of 4059 atoms. Structural modifications and RMSD analysis of the macromolecular backbone were carried out during the 100 ns simulation to evaluate thermodynamic stability. The complex ligand hypercin had four flexible links, each containing 38 heavy atoms, resulting in a total of 48 atoms. The combination of coat protein and hypercin exhibited remarkable stability throughout the simulation. The RMSD value of the protein backbone ranged from 5.8 to 11.7 Å, whereas the bound ligand hypercin exhibited minor fluctuations in the receptor cavity, ranging from 7.5 to 13.4 Å. Figure 11 illustrates the identified RMSD between the coat protein complex and hypercin. RMSD observed for the coat protein complexed with hypercin during 100 ns MD simulation The atoms comprising a protein or ligand structure may deviate from their initial positions, which may be determined by their RMSF value. This is an essential variable that influences the flexibility and dynamic behavior of the macromolecular complex. Protein RMSF has significance since it may anticipate protein dynamics and evaluate stability by demonstrating the relative flexibility of multiple domains. The MD examination of the coat protein complexed with hypercin disclosed that the RMSF for the Cα backbone fluctuated between 1.4 and 7.5 Å, while the ligand hypercin fluctuated between 4.2 and 8.2 Å. Figure 12 illustrates the RMSF of the coat protein backbone and its associated ligand hypercin, as seen during MD analysis. RMSF observed for the coat protein complexed with hypercin during 100 ns MD simulation The establishment of hydrophobic interactions, ionic bonds, and hydrogen bonds during an MD simulation optimized the thermodynamic stability of a receptor-ligand complex, assessed by continuously monitoring their strengths across all macromolecular complexes throughout the simulation. In the simulation, the ligand hypercin appeared to interact with the coat protein through hydrophobic interactions involving the amino acids Val230, His240, Ala242, and Ile245, while Ser231, Asp236, Arg239, Thr241, and Lys247 via formation of hydrogen bonds, and amino acids Arg217, Gly225, Asp227, Asn229, Thr233, Glu234, Glu238, Ile245, and Asn246 are found to be interacting through water bridges. Figure 13 illustrates the interaction between the coat protein and its ligand, hypericin. Interactions observed between the COAT protein complexed with hypercin during 100 ns MD simulation. (Purple: Hydrophobic interactions, Green: Hydrogen bonds, Blue: Water bridges) Discussion Garlic, scientifically referred to as Allium sativum, is a widely cultivated plant across the globe. It is not only a culinary staple but also holds significant value as a medicinal herb, offering numerous health advantages. Garlic is rich in essential vitamins, minerals, and bioactive compounds, contributing to its extensive pharmacological properties, including antimicrobial, antioxidant, anti-inflammatory, and anticancer effects [33]. However, the cultivation and productivity of A. sativum are severely affected by several viral infections, leading to substantial declines in global crop yields. Of these viral diseases, the onion yellow dwarf virus (OYDV) is most detrimental to garlic [34]. Plant secondary metabolites (phytochemicals) can act as antiviral agents and counteract harmful viral pathogens. The plant uses these metabolites to protect itself by strengthening its defensive systems, helping it fight off a broad spectrum of viruses. The plant is better able to resist and fight the viral infections, leading to increased plant growth and ultimately increased crop productivity [35, 36]. Similarly, one of the reported studies utilized an in-silico strategy to analyze the therapeutic effectiveness of turmeric against potato virus Y. This natural compound demonstrated considerable inhibition against the potato virus [37]. Considering these important findings along with the documented therapeutic potential of natural compounds, the present study also incorporated such compounds from other identified plant sources in this case study. The coat protein of OYDV was the primary target of this study. By multiple sequence alignment and phylogenetic analysis, it was found that about 30.3% of the amino acids in the protein’s active sites were conserved among all investigated coat proteins from different potyvirus species. The neighbor-joining phylogenetic tree analysis illustrated a close evolutionary relationship among the coat proteins of onion yellow dwarf virus, pepper mottle virus, and potato virus Y. MEGA X software has also been used in other studies, including one that used neighbor-joining techniques to check evolutionary relationships among begomovirus-associated betasatellite βC1 proteins [10]. The results of this study for the selected CP, with a molecular weight of 29,175.93 Da and consisting of 257 amino acids, showed that it is a stable protein with an isoelectric point (pI) of 6.28. These parameters are essential for establishing optimal purification conditions to maintain protein stability and solubility [38]. The extinction coefficient, measured at 32,890 M⁻¹ cm⁻¹ (280 nm in water), facilitated the quantitative assessment of protein–ligand and protein–protein interactions in solution [39]. The instability index (II) was computed to be 30.90, classifying the protein as stable, as values below 40 typically indicate structural stability [40]. The aliphatic index (AI) of 67.24 suggested moderate thermostability, implying tolerance to a range of temperatures [41]. Additionally, the grand average of hydropathicity (GRAVY) was determined to be − 0.665, confirming the hydrophilic nature of the protein and its strong interaction with aqueous environments [42]. By contrast, a recent study used the same ExPASy ProtParam server to predict the physicochemical properties of M. persicae cuticle protein (MPCP4) [11]. Post-translational modifications, including phosphorylation and glycosylation, regulate several biological processes by controlling protein folding, activity, and localization [14]. The prediction of functional sites indicated that the coat protein contains 30 phosphorylation sites, as determined using NetPhos 3.1. All were localized to serine, threonine, or tyrosine residues and targeted both specific and nonspecific kinases. Additionally, NetNGlyc 1.0 identified six N-glycosylation sites in the same protein. Similar servers were used in a recent study on Osmotin alleles (OsmSt1 and OsmSt2), where phosphorylation and N-glycosylation sites were detected in those sequences [43]. Gene ontology analysis of the coat protein confirmed its involvement in viral RNA replication and viral capsid formation. Secondary structure prediction indicated random coils as the dominant element (54.86%), with smaller proportions of alpha helices and extended strands, suggesting that the coat protein has a flexible backbone structure. Furthermore, ten distinct motifs were identified in the coat protein, each displaying unique sequence features. Analysis of the spatial distribution of these motifs revealed that they are situated in specific regions of the protein sequence. These analyses were performed using PANZZER2 (gene ontology), SOPMA (secondary structure), and MEME suite v5.5.1 (motif analysis), respectively. Notably, similar tools were employed in a previous study examining BRa.PDF1.1 proteins [44]. The 3D structure of OYDV’s coat protein was predicted using trRosetta and ColabFold (v1.5.5), while a previous study employed SWISS-MODEL to predict the 3D structure of the coat protein of Yam Bean Mosaic Virus [45]. Model quality assessments of trRosetta- and ColabFold (v1.5.5)- predicted models were conducted via Ramachandran plot analysis using the PROCHECK tool in SAVES 6.0, showing 92.2% and 67.7% of residues in favored regions, respectively. The ERRAT (SAVES 6.0) further validated both trRosetta- and ColabFold (v1.5.5)-predicted CP structures with high-quality scores of 97.0443 and 81.395, respectively. The quality assessment confirmed the reliability of the trRosetta CP model for further analysis. These similar validation tools were previously employed in a study on infection control in tea plants, assessing the quality of GRAPH and chitinase proteins [46]. Prior to molecular docking, the CASTpFold server predicted 23 potential binding pockets in the onion yellow dwarf virus coat protein. Among these, the largest binding site was selected for subsequent docking studies due to its extensive interaction surface. A similar computational approach was previously employed in a study to identify active sites in movement proteins of tobamoviruses [47]. Molecular docking was performed using PyRx software with AutoDock Vina, revealing several plant secondary metabolites with higher binding affinities than positive controls (ningnanmycin and ribavirin). Four compounds showed exceptional binding (< − 9 kcal/mol): hypericin (− 9.9), celastrol (− 9.7), sanguinarine (− 9.3), and asiatic acid (− 9.1). PyRx has been similarly employed in a recent study in which Nigellidine demonstrated the highest affinity against Potato Virus X [48]. The ligand–protein interactions were observed in BIOVIA Discovery Studio for the top four ligand molecules and positive controls. Residues LYS247, ARG239, HIS240, and ASP244 were found in all ligand–protein complexes and thus highlight their importance in stabilizing ligand binding in the active site pocket of the coat protein. Conventional hydrogen bonds, carbon–hydrogen bonds, van der Waals interactions, π–π stacking, π–alkyl, π-cation, and π-sigma interactions contributed to complex stabilization, while the unfavorable donor–donor interaction slightly decreased the overall binding stability. One of the published research works has also employed the same software to visualize dexamethasone interaction with coronavirus proteins [49]. In this study, the ProTox 3.0 server revealed distinct toxicity patterns and physicochemical properties of the best identified ligands. In toxicity analysis, hypericin, celastrol, sanguinarine, asiatic acid, and ningnanmycin were classified under GHS Class IV, indicating that they may cause harm if ingested. However, ribavirin was classified under GHS Class V, suggesting it is potentially less harmful than the Class IV compounds. Notably, none of the evaluated compounds, including both positive controls and phytochemicals, demonstrated predicted cytotoxicity, suggesting an encouraging safety profile regarding this specific endpoint. However, the predictions identified several compounds requiring further toxicological scrutiny. Both sanguinarine and ribavirin were flagged as potential carcinogens, while three compounds (celastrol, sanguinarine, and asiatic acid) exhibited predicted immunotoxicity that may warrant consideration in therapeutic applications. Regarding mutagenic potential, hypericin and sanguinarine were predicted to possess mutagenic properties. The predicted LD50 values demonstrated considerable variation across compounds, ranging from 778 mg/kg for sanguinarine (indicating pronounced toxicity) to 2700 mg/kg for ribavirin (suggesting relatively low toxicity). Lipinski’s rule of five (RO5) is regarded as the standard for determining the physicochemical and structural characteristics of a compound for estimating oral drug bioavailability, which is also relevant for assessing the agrochemical characteristics of any bioactive component. All of the compounds that were tested had molecular weights below 500 Da, except hypericin, and logP values for all the top ligands were greater than 5, except sanguinarine and positive controls. Both HBD and HBA indicated the reactivity of the ligand. The number of HBDs was less than 10 for all the ligands, except ningnanmycin and hypericin. The number of HBAs was less than 10 for all the ligands, except ningnanmycin. The TPSA values of sanguinarine, celastrol, and asiatic acid were below the threshold of 140 Ų. The number of rotatable bonds showed 10 in ningnanmycin, indicating its flexibility. Molar refractivity values of hypericin (142.9) suggested the strongest polarizability out of the rest of the compounds. One of the published research works has employed the ProTox 2.0 to check the toxicity patterns as well as physicochemical properties of phytosterol dendrosterone [50]. Afterward, the density functional theory calculations of hypericin were performed, which showed a small ΔE and associated molecular reactivity descriptors, signifying enhanced chemical reactivity and charge transfer capacity, facilitating molecular recognition and biological interactions. Another reported study also employed the DFT analysis of four isolated compounds from the leaves of Coccinia grandis [51]. The results of this study for MD simulations presented the stability and flexibility of the CP-hypercin complex. It was found that the protein underwent significant fluctuations in order to attain a stable behavior. This also signified a higher extent of structural flexibility in the receptor’s structure. However, in the end of the simulations the complex showed a stable interacting behavior which indicated towards favorable results. As a whole, these analyses showed the complex to be highly dynamic, stable, and compact, one of the published studies in which MD simulations were performed using a similar server in order to investigate complexes NS2B/NS3–taraxerol and NS5–apigenin has been reported [52]. The current strategic work was mainly aimed at not only identifying and designing a potential biopesticide targeting the coat protein of OYDV, but also to provide new directions for eco-friendly biopesticide development. However, the real-time efficacy of hypericin against the coat protein of OYDV could only be attained in in planta experiments. Conclusion The study combines molecular docking and simulations with a comprehensive analysis of plant-derived bioactive compounds to assess their potential in antiviral applications against onion yellow dwarf infection. Hypericin showed the most favorable results out of the rest of the compounds analyzed, making it a strong candidate. Moreover, celastrol, sanguinarine, and asiatic acid had binding conformations similar to hypericin, which reinforces their consideration as potential antiviral agents. Overall, the study reflects a significant breakthrough in the use of metabolites with antiviral activity from plants for the development of antiviral drugs and also paves the way for more studies as well as potential commercialization. For future studies, the analogues of these top bioactive compounds can be designed and evaluated to discover which amendment increases the antiviral efficacy. In addition, exploring appropriate methods to deliver these bioactive compounds to infected plants is necessary. Such investigations will validate the in-silico predictions and also accelerate the practical application of these eco-friendly compounds in sustainable plant disease management. Supplementary Information Supplementary Material 1. Acknowledgements The authors are thankful to the Deanship of Scientific Research, King Khalid University, Abha, Saudi Arabia, for supporting this work through the Large Research Group Project under Grant no. R.G.P.2/480/46. Authors’ contributions A, I. did original research and wrote the original manuscript. A.I. did the formal analysis, and Q.N. helped in collecting research data. M.F.A. supervised the whole research work and gave his expertise in conducting research by doing bioinformatics analysis, and I.M helped in the correct interpretation of the final results. The S.O.R. provided resources for the work. All authors approved the final manuscript. Funding No funding was available for this work. Data availability All data used in this study are provided in the manuscript in different forms and in various sections. The online sequence data and their information are already provided as accession numbers within the manuscript. The used plant materials were obtained from the University of Management and Technology, Sialkot, Pakistan. Declarations Ethical approval and consent to participate The study was conducted according to institutional and national guidelines. It does not involve the physical usage of any living organisms, plants, animals, and humans; therefore, ethical approval is not applicable. This study was based on computational analysis, and no informed consent was needed. Consent for publication All authors have given their full consent for the publication of this research work. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement All data used in this study are provided in the manuscript in different forms and in various sections. The online sequence data and their information are already provided as accession numbers within the manuscript. The used plant materials were obtained from the University of Management and Technology, Sialkot, Pakistan.