{"paper_id":"24a58f88-2efd-4a73-8010-5ff6a1a21c4b","body_text":"Page | 1 \nMulti-Target In Silico Investigation of Withaferin A as a Potential Antiviral \nInhibitor Against Key Marburg Virus Proteins\nFarayed Ahamed Nabil a, Abu Darda a, Ehsanul Islam c, Ferdaus Mohd Altaf Hossain d, Kazi Md. Ali \nZinnah*a,b\naFaculty of Biotechnology and Genetic Engineering, Sylhet Agricultural University, Sylhet-3100,\nBangladesh\nbDepartment of Animal and Fish Biotechnology, Sylhet Agricultural University, Sylhet-3100, Bangladesh\ncDepartment of Pharmacy, Faculty of Biological Science, Jahangirnagar University, Savar, Dhaka-1342, \nBangladesh\ndDepartment of Dairy science, Sylhet Agricultural University, Sylhet-3100, Bangladesh\n*Corresponding Author \nDr. Kazi Md. Ali Zinnah, \nAssociate Professor\nDepartment of Animal and Fish Biotechnology, Faculty of Biotechnology and Genetic Engineering\nSylhet Agricultural University, Sylhet-3100.\nE-mail: zinnah.afb@sau.ac.bd\nMobile: +8801913208646\n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 2 \nAbstract\nMarburg virus (MARV) is a highly pathogenic filovirus that causes hemorrhagic fever with a high mortality \nrate, with very limited treatment options. The urgent need for targeted antiviral agents emphasizes the \nimportance of structure-based drug discovery approaches. The present study aimed to evaluate the antiviral \npotential of Withaferin A (PubChem CID-265237) against three key proteins of MARV: viral protein 35 \n(VP35), and nucleoproteins (NP). Three-dimensional structures of these proteins were retrieved from \nRCSB-Protein Data Bank and docked with Withaferin A using AutoDock Vina. The ligand demonstrated \nfavourable binding affinities towards all three viral targets, indicating strong interaction potential at \nfunctionally relevant sites. Drug-likeness and pharmacokinetic properties predicted using SwissADME and \npkCSM indicated acceptable ADMET profiles that comply with key drug-like criteria. To validate the \nstability of the docking, molecular dynamics simulations (GROMACS, 100 nanoseconds) were conducted. \nThe protein-ligand complexes exhibited stable root mean square deviation (RMSD), root mean square \nfluctuation (RMSF), and consistent hydrogen bonding patterns throughout the simulation. The MM-GBSA \nbinding free energy analysis further supported favorable binding energetics, predominantly driven by van \nder Waals and electrostatic interactions. Altogether, these findings demonstrate that Withaferin A exhibits \npromising multi-target inhibitory potential against key MARV proteins. This study provides molecular \ninsights into ligand-protein interactions and supports further experimental validation of Withaferin A as a \npotential therapeutic candidate against Marburg virus.\nKeywords: Marburg virus, Withaferin A, VP35, Nucleoprotein, Molecular docking, Structure-based drug \ndiscovery.\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 3 \n1. Introduction\nMarburg virus (MARV) is an extremely pathogenic filamentous, negative (negative sense) single-stranded \nRNA virus belonging to the Filoviridae family, the same virus family as Ebola. It causes Marburg virus \ndisease (MVD), a severe hemorrhagic fever with high fatality rates, systemic inflammation, multiorgan \nfailure, and immune dysregulation (1). Since its first outbreak in 1967 in Germany and Serbia, recurrent \noutbreaks have been reported in several African countries, with case fatality rates reaching up to 88% in \nsome epidemics (2). Although MARV poses a serious public health risk and has epidemic potential, \napproved antiviral therapies specific to this virus are still lacking. This gap highlights the urgent need for \neffective and rational drug discovery (3).\nThe Marburg virus (MARV) genome encodes seven structural proteins. Among them, viral protein 35 \n(VP35), and the nucleoproteins (NP) play essential roles in viral replication, immune evasion, and virion \nassembly. VP35 (PDB ID: 4GH9, Crystal structure of Marburg virus VP35 RNA binding domain) functions \nas a polymerase cofactor and a potent interferon antagonist, thereby suppressing the host innate immune \nresponse and promoting viral replication. Whereas, NP (PDB ID: 4W2Q and 4W2O, Anti-Marburgvirus \nNucleoprotein Single Domain Antibody C Complexed with Nucleoprotein C-terminal domain) \nencapsidates the viral RNA, which is critical for nucleocapsid formation and the regulation of viral \ntranscription (4, 5).  Owing to their essential roles in the viral life cycle, these proteins are considered \npromising molecular targets for antiviral intervention.\nStructure-based drug discovery (SBDD) provides a computational framework for screening potential \ninhibitors against viral protein targets (6). Molecular docking predicts ligand-protein binding orientations \nand estimates binding affinities (7). whereas molecular dynamics (MD) simulations provide insights into \nconformational flexibility, interaction stability, and time-dependent structural changes under near-\nphysiological conditions (8). In addition, binding free energy was estimated using the MM/GBSA approach \nto provide a more reliable assessment of interactions, incorporating solvation and entropic effects (9). \nCollectively, these computational strategies play an important role in antiviral drug discovery pipelines, \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 4 \nespecially for high-risk pathogens for which experimental investigations are restricted by biosafety \nrequirements (10). \nThere have been a number of computational research pursuing potential inhibitors against the individual \nMARV proteins through docking-based virtual screenings. While promising molecules have been reported, \na considerable number of studies have focused mainly on static docking results, lacking comprehensive \ndynamic validation and multi-target assessment. Considering the complex replication strategy of MARV \nand the coordinated roles of its structural proteins, a multi-target inhibition approach may offer greater \ntherapeutic potential than single-target strategies (11). Incorporation of extended MD simulations and the \nfree energy calculations can also contribute to increased robustness and predictive reliability of the \ncomputational analysis.\nNatural bioactive compounds have been extensively studied as potential antiviral agents owing to their \nstructural diversity, bioactivity, and traditional medicinal use (12). Withaferin A, a steroidal lactone derived \nfrom Withania somnifera, has demonstrated antiviral, anti-inflammatory, and immunomodulatory \nproperties in various experimental contexts. Its pharmacological versatility and reported safety profile \nrender it a suitable candidate for computational antiviral screening (13).\nIn this study, an integrated in silico study was conducted to examine the inhibitory potential of Withaferin \nA against three key proteins (VP35, two NP) of MARV. Their crystal structures were retrieved from the \nRCSB Protein Data Bank for subsequent computational analyses (14). Molecular docking was performed \nto predict binding affinities and interaction patterns, followed by 100 ns molecular dynamics simulations \nand MM/GBSA free energy analysis to evaluate dynamic stability and thermodynamically feasible binding. \nAdditionally, pharmacoinformatics and toxicity profiling are also being performed in order to evaluate \ndrug-likeness and safety parameters. Using a combination of multi-target docking, extended validation \nthrough MD simulations, and estimation of binding free energy in a unified computational approach, this \nwork is aimed towards a comprehensive evaluation of Withaferin A as a potential lead compound against \nMarburg virus. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 5 \n2. Materials and Methods  \nThe stepwise methodology of the entire study is illustrated in Figure 1.  \n2.1 Selection and Collection of Protein Sequences  \nThe amino acid sequences of proteins VP35, and the nucleoproteins (NP) of Marburg virus were chosen. \nThese proteins have been found to play vital roles in viral reproduction, immune response, and in forming \na viral particle; therefore, they are a promising therapeutic target. Corresponding structural data were \nretrieved from the RCSB Protein Data Bank, ensuring the availability of experimentally resolved crystal \nstructures appropriate for structure-based drug design (14). \n2.2 Retrieval and Preparation of Target Proteins  \nThe three-dimensional crystal structures of VP35 (PDB ID: 4GH9), and NP (PDB ID: 4W2Q and 4W2O) \nwere downloaded from the RCSB Protein Data Bank (15). Protein preparation involved removal of \ncrystallographic water molecules, co-crystallized ligands, and heteroatoms. Polar hydrogens were \nincorporated and Kollman charges were assigned using AutoDock tools. Energy minimization was \nperformed to eliminate steric clashes for structural stability before docking. \n2.3 Prediction of 3D Structures of the target Proteins  \nSince high-resolution crystal structures of the selected proteins were available in the Protein Data Bank, no \nadditional homology modeling was required. Structural validation was performed using Ramachandran plot \nanalysis to assess stereochemical quality and backbone dihedral angle distribution. Most residues have been \nfound in favored and allowed regions, confirming the structural reliability for computational analysis.\n2.4 Enlistment and Collection of Ligand Structure and Preparation  \nAn initial library of 50 bioactive compounds was compiled based on reported antiviral and \nimmunomodulatory properties. These compounds were subjected to preliminary virtual screening, from \nwhich Withaferin A (PubChem CID: 265237) was selected as the lead candidate based on its favorable \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 6 \nbinding performance and pharmacological relevance. The three-dimensional structure of Withaferin A was \nretrieved from the PubChem database (16). The downloaded SDF file was converted to PDB format using \nOpen Babel (17). Energy minimization was conducted using the Universal Force Field (UFF) to obtain a \nstable, low-energy conformation suitable for subsequent docking analysis. \n2.5 Molecular Docking  \nMolecular docking simulations were performed using AutoDock Vina (7). Grid boxes were configured to \nencompass the functional domains of VP35, and NP. The default exhaustiveness setting was used to ensure \nadequate conformational sampling. The resulting docked conformations were ranked based on predicted \nbinding affinity values (kcal/mol). The lowest energy complexes were chosen for further analysis.  \n2.5.1 Calculating the Binding Affinity of the Proteins and Ligands  \nAutoDock Vina's scoring function was used for estimating the free energy of binding. Lower (more \nnegative) binding energy values indicate more strongly predicted interactions between the ligand and \nprotein. The most favorable docking conformation for each protein-ligand complex was identified based \non its minimum binding energy and favorable interaction geometry.  \n2.5.2 Study of the Preferred Binding Site of the Superior Metabolites  \nThe preferred binding sites were analyzed by examining hydrogen bonding, hydrophobic interactions, \nelectrostatic contacts, and van der Waals interactions using PyMOL and Discovery Studio Visualizer (18). \nResidues involved in ligand stabilization were identified, and their interactions within active or functionally \nrelevant regions were further examined to clarify the possible inhibitory mechanism.\n2.6 Molecular Dynamics Simulation Studies\nThe stability, flexibility, and dynamic properties of the protein and protein-ligand complexes were \nexamined through molecular dynamics (MD) simulations. All simulations were conducted using \nGROMACS 2025.1 (19). The protein was parameterized with the CHARMM36 force field, while ligand \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 7 \ntopologies were generated using SwissParam (20). Prior to simulation, the systems were energy-minimized \nusing a gradient-based algorithm for 2,500 steps to eliminate steric clashes and optimize structural \ngeometry. The SPC water model was used to solvate each system in a cubic box with periodic boundaries. \nNa+ and Cl- counter-ions were added to Gmx Genion to achieve overall charge neutrality.  \nAfter minimization, a two-stage equilibration was carried out. First, a 100 ps NVT equilibration was carried \nout using the v-rescale thermostat to stabilize the temperature at 310 K. This was followed by a 100 ps NPT \nequilibration using the Parrinello-Rahman barostat to maintain pressure at 1 bar and achieve density \nequilibration. Position restraints were applied to protein heavy atoms during equilibration. Bond constraints \nwere imposed using the LINCS algorithm, and long-range electrostatic interactions were treated with the \nParticle Mesh Ewald (PME) method. After equilibration, three unrestrained 100 ns MD simulations were \nperformed for all systems. Trajectories were saved at appropriate intervals for subsequent analyses.\n2.6.1 Root Mean Square Deviation (RMSD) \nRMSD is a metric that quantifies the distance between frames. It is determined for every profile frame. The \nroot mean square deviation of frame x is calculated.\n𝑅𝑀𝑆𝐷𝑋 = 1\n𝑁\n𝑁\n𝑖=1\n𝑟′𝑖(𝑡𝑥) ― 𝑟𝑖 𝑡𝑟𝑒𝑓\n2\nWhere N is the number of selected atoms, ri (t ref) represents the position of atom I in the reference structures \ndenotes the position of the same atom in frame x after optimal structural superposition onto the reference. \nAll RMSD calculations were performed using the gmx function in GROMACS2025.1, and the resulting \ntrajectories were visualized using XmGrace.\n2.6.2 Root Mean Square Fluctuation (RMSF)\nRMSF values can be used to determine the local flexibility based on residue displacements during the MD \nsimulation. The Root Mean Square Fluctuation (RMSF) is useful for characterizing local changes along the \nprotein chain. The RMSF for residue i is calculated using:\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 8 \n𝑅𝑀𝑆𝐹𝐼 = 1\n𝑇\n𝑇\n𝐼=1\n(𝑟𝑖(𝑡) ― (𝑟𝑖))2\nWhere r i (t) represents the position I at time t and r i is its time averaged position. Unlike RMSD, which \ncompares each frame to a reference structure, RMSF characterizes local fluctuations around the mean \ntrajectory. Peaks in the RMSF profile correspond to flexible regions-typically loops and the N or C terminal \nsegments, while lower values indicate more rigid, structurally stable regions of the protein.\n2.6.3 Radius of Gyration (Rg)\nThe radius of gyration (Rg) is the mass-weighted root mean square distance of a group of atoms around a \ncommon center of mass. Rg is widely used to evaluate the global compactness and folding stability of \nbiomolecular systems during MD simulations. A stable Rg profile indicates maintenance of tertiary \nstructure, whereas significant deviations imply expansion, unfolding, or structural reorganization. In this \nstudy, Rg was calculated for each trajectory frame using GROMACS analysis tools to monitor the \ncompactness of the protein and protein-ligand complexes over the 100 ns simulation.\n2.6.4 Solvent Accessible Surface Area (SASA)\nThe Solvent Accessible Surface Area (SASA) was computed using the gmx sasa module in GROMACS \n2025.1 along the 100 ns MD trajectory. SASA provides insights into the extent of solvent exposure and \npotential structural rearrangements at the protein surface. A default probe radius of 0.14 nm was applied, \ncorresponding to the radius of a water molecule. SASA values were extracted at regular intervals for the \nentire protein-ligand complex to evaluate dynamic changes in surface exposure, folding stability, and \nsolvent interactions throughout the simulation.\n2.6.5 Hydrogen Bond Analysis\nThe stability and formation of hydrogen bonds were measured on the 100 ns simulation in the gmx hbond \nmodule of GROMACS version 2025.1. Hydrogen bonds were identified using standard geometric criteria: \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 9 \na donor-acceptor distance ≤ 0.35 nm and a donor-hydrogen-acceptor angle ≥ 120°. The number of hydrogen \nbonds for each frame of the trajectory was also calculated to examine the intramolecular stability as well as \nto determine the effect of the ligand binding to the stability and distribution of the hydrogen bonds within \nthe protein structure.\n2.6.6 MMGBSA Binding Free Energy\nMolecular Mechanics/Generalized Born Surface Area (MM/GBSA) calculations were used to determine \nthe free binding energy between the ligand and the protein and were done with gmx_MMPBSA tool \n(v1.6.3), which connects GROMACS trajectories to the AMBER free-energy calculations (21). This \nmethod breaks down the binding free energy (ΔG_bind) into molecular mechanical energy terms (van der \nWaals and electrostatic interactions) and solvation free-energy contributions (polar and nonpolar), \naccording to the following equation: \n∆𝐺𝑏𝑖𝑛𝑑 =  ∆𝐸𝑣𝑑𝑊 +  ∆𝐸𝑒𝑙𝑒 +  ∆𝐺𝑝𝑜𝑙𝑎𝑟 +  ∆𝐺𝑛𝑜𝑛𝑝𝑜𝑙𝑎𝑟 \nThe Generalized Born (GB) implicit solvent model was used to compute the polar solvation energy, with \nthe igb = 5 parameter corresponding to the GB-Neck2 model. The nonpolar solvation term (ΔG_nonpolar) \nwas estimated from the solvent-accessible surface area using the LCPO method. Entropic contributions had \nbeen ignored, as is usual with comparative binding affinity studies, and has no effect on the relative ranking \nof ligands.\nThe 1001 snapshots were obtained in an equal manner in the 100 ns production trajectory (MD_center.xtc). \nMM/GBSA calculations were performed on the complex, the receptor (protein only) and the ligand, \nseparately. The binding free energy was calculated using the standard thermodynamic cycle:\n∆𝐺𝑏𝑖𝑛𝑑 =  𝐺𝑐𝑜𝑚𝑝𝑙𝑒𝑥 ― (𝐺𝑟𝑒𝑐𝑒𝑝𝑡𝑜𝑟 +  𝐺𝑙𝑖𝑔𝑎𝑛𝑑)\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 10 \nThis protocol facilitated fine breakdown of contributions to binding including van der Waals, electrostatic, \npolar and nonpolar solvation contributions which offers a sound energetic assessment of the interactions \nbetween ligands and proteins.\n2.7 Pharmacoinformatics Studies\nSwissADME was used to predict the drug-likeness and pharmacokinetic properties of Withaferin A (22). \nParameters analyzed included Lipinski’s Rule of Five compliance, molecular mass, hydrogen bond donors \nand acceptors, lipophilicity, and gastrointestinal absorption.\nAdditional pharmacokinetic predictions, including cytochrome P450 interactions and ADME parameters, \nwere evaluated using pkCSM (23). \n2.8 Toxicity Analysis\nThe pkCSM web server was used to determine the toxicity prediction of Withaferin A. Parameters evaluated \nincluded hepatotoxicity, AMES mutagenicity, cardiotoxicity (hERG inhibition), and acute toxicity \nprediction. These computational toxicity assessments provide preliminary safety insights before \nexperimental validation.  \n3. Results\n3.1 Retrieval of Targeted Proteins\nThe three-dimensional crystal structures of the Marburg virus proteins were successfully retrieved from the \nRCSB Protein Data Bank. The selected targets included viral protein 35 (VP35; PDB ID: 4GH9), and two \nnucleoproteins (NP; PDB ID: 4W2Q and 4W2O). All structures were downloaded in PDB format and \nprepared for subsequent computational analyses. Structural preprocessing involved the removal of \nheteroatoms and crystallographic water molecules, followed by the addition of hydrogen atoms and the \nassignment of appropriate charges. The prepared protein structures were then successfully utilized for \ndownstream molecular docking and molecular dynamics simulations.\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 11 \n3.2 Molecular Modelling and Quality Assessment\nThe retrieved protein structures were evaluated for stereochemical quality using Ramachandran plot \nanalysis. Then their 3D structures, Ramachandran plot and Errat quality values of the proteins are given in \nFigure 2, Figure 3 and Figure 4.\nTable 1. Refined protein model with their ERRAT values, Ramachandran plot results, and binding site \nresidues.\nProtein name ERRAT value Favored region (%) Disallowed region (%)\n4GH9 100 97.1% 0.0%\n4W2O 94.0476 86.9% 0.0%\n4W2Q 92.5816 93.2% 0.0%\nThe Ramachandran plots illustrated that most amino acid residues were found in favored and additionally \nallowed regions, with a minimal percentage of the residues found in disallowed regions. These findings \nverify the structural reliability and geometric stability of the chosen proteins for structure-based \ncomputational studies.  \n3.3 Molecular Docking and Binding Site Analysis\nMolecular docking of Withaferin A (CID: 265237) with VP35 and NP of the Marburg virus, conducted \nusing AutoDock Vina, is given in Figure 5. The predicted binding affinities demonstrated favorable \ninteractions with all three targets. The docking scores were -8.2 kcal/mol for VP35 (PDB ID: 4GH9), -8.3 \nkcal/mol for NP (PDB ID: 4W2O), and -9.5 kcal/mol for NP (PDB ID: 4W2Q). Among the selected \nproteins, nucleoprotein (NP) exhibited the strongest binding affinity with Withaferin A, indicating a \ncomparatively more stable ligand-protein interaction.\nTable 2. Binding sites of the best metabolites with target protein. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 12 \nHDock CB-Dock\nProtein Ligand Docking \nScore\nConfidence \nScore\nVina \nScore\nHydrogen \nBond \nInteractions\nHydrophobic \nInteractions (Alkyl \n/ Pi-Alkyl)\n4W2O\nWithaferin \nA\n-201.33 0.7363\n-8.3 \nkcal/mol\nLYS-145 (B), \nASP-53 (A)\nALA-52 (A), LYS-\n475 (F)\n4GH9\nWithaferin \nA\n-128.07 0.3921\n-8.2 \nkcal/mol\nGLN-233, \nLYS-211\nLEU-215, PHE-\n218, ALA-214, \nLYS-237, ALA-\n210, TYR-240, \nLYS-241\n4W2Q\nWithaferin \nA\n-184.40 0.6655\n-9.5 \nkcal/mol\nGLU-649 \n(H), THR-\n551 (G)\nLYS-314 (D), \nALA-318 (D)\nAmong the three targets, the NP-Withaferin A complex exhibited the lowest binding energy.\nInteraction analysis revealed that the formation of hydrogen bonds, hydrophobic interactions, and van der \nWaals contacts have been formed in the binding pockets of the proteins. \nThe 2D interaction diagrams of the proteins with Withaferin A are illustrated in Figure 6. \nThe functional domains of each protein were found to have key interacting residues, which showed stable \naccommodation of their ligands in the predicted binding sites. \n3.4 Analysis of Molecular Dynamics Simulation and MM/GBSA\n3.4.1 Root Mean Square Deviation (RMSD) \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 13 \nThe RMSD profiles of the three complexes 4GH9-CID 265237 (WITHAFERIN A, black), 4W2Q-CID \n265237 (WITHAFERIN A, red), and 4W2O-CID 265237 (WITHAFERIN A, green) demonstrated overall \nstructural stability throughout the 100 ns simulation are given in Figure 7. The 4GH9 complex maintained \nRMSD values predominantly between 0.04-0.06 nm, with occasional spikes reaching approximately 0.18-\n0.19 nm around mid-simulation. The 4W2Q complex showed slightly lower deviation, fluctuating within \n0.04-0.07 nm, with rare peaks near 0.16 nm. In contrast, 4W2O exhibited comparatively higher fluctuations, \ngenerally ranging between 0.06-0.10 nm, but without prolonged instability. The consistently low RMSD \nvalues (<0.2 nm) for all systems indicate stable protein-ligand conformations, with 4GH9 and 4W2Q \nshowing comparatively better structural rigidity than 4W2O. \n3.4.2 Root Mean Square Fluctuation (RMSF)\nThe RMSF analysis revealed residue-level flexibility differences among the complexes. The Residue-wise \nroot means square fluctuation (RMSF) profiles of the three complexes during the molecular dynamic \nsimulation are presented in Figure 8. For 4GH9, fluctuations were mostly confined to 0.05-0.10 nm, with \na notable peak reaching approximately 0.20 nm near residue ~310, indicating localized loop flexibility. The \n4W2Q complex showed fluctuations primarily within 0.04-0.15 nm, with occasional peaks around 0.18-\n0.19 nm in the N-terminal region. Similarly, 4W2O exhibited RMSF values largely between 0.04-0.17 nm, \nwith one prominent peak near 0.20 nm around residue ~100. Overall, most residues in all systems remained \nbelow 0.10 nm, suggesting stable backbone dynamics, while higher peaks correspond to flexible loop \nregions rather than structural destabilization.\n3.4.3 Radius of Gyration (Rg)\nThe radius of gyration analysis indicated compact structural behavior for all complexes across 100 ns (0-\n100,000 ps) is given in Figure 9. The 4GH9 complex maintained an average Rg of approximately 1.45-\n1.48 nm, showing minor fluctuations (~±0.02 nm), suggesting a stable and compact structure. Both 4W2Q \nand 4W2O complexes exhibited slightly lower Rg values, averaging around 1.38-1.40 nm, with minimal \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 14 \ndeviation during the simulation. No significant expansion or unfolding events were observed in any system. \nThe relatively constant Rg values confirm that ligand binding did not induce major conformational \nexpansion, and the complexes remained structurally compact throughout the trajectory.\n3.4.4 Solvent Accessible Surface Area (SASA)\nThe SASA profiles demonstrated moderate surface exposure variations among the complexes. Time-\ndependent solvent-accessible surface area (SASA) of the three protein-ligand complexes over the 100 ns \nmolecular dynamics simulation, which are given in Figure 10. The 4GH9-CID 265237 (WITHAFERIN A) \nsystem showed SASA values fluctuating between 71-76 nm², indicating comparatively higher solvent \nexposure. The 4W2Q complex ranged between 68-74 nm², whereas 4W2O displayed slightly lower values, \nmostly between 64-70 nm². Although minor oscillations were observed throughout the simulation, no \nabrupt or sustained increases were detected, suggesting the absence of unfolding events. The relatively \nlower SASA in 4W2O corresponds with its slightly lower compactness variation observed in Rg analysis.\n3.4.5 Hydrogen Bonds\nThe hydrogen bond profiles for the three protein-ligand complexes during the 100 ns molecular dynamics \nsimulation are shown in Figure 11. The 4GH9 complex formed between 0-3 hydrogen bonds, frequently \nmaintaining 1-2 bonds during most of the trajectory. The 4W2Q complex also exhibited 0-3 hydrogen \nbonds, with sustained interactions particularly between 50-80 ns. Similarly, 4W2O maintained 0-3 \nhydrogen bonds, predominantly stabilizing at 1-2 bonds in the latter half of the simulation. Although \ntransient bond breaking and reforming were observed, the consistent presence of hydrogen bonds across all \nsystems supports stable ligand binding. Among them, 4W2Q showed comparatively more persistent \nhydrogen bonding during mid-to-late simulation phases.\n3.4.6 MMGBSA\nMM-GBSA binding free energy calculations for Withaferin A with VP35 (4GH9), NP (4W2O), and NP \n(4W2Q) of Marburg virus are given in Figure 12. Among the complexes, 4GH9-CID 265237 \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 15 \n(WITHAFERIN A) exhibited the most favorable total binding free energy (ΔG = -6.99 kcal/mol), followed \nby 4W2O-CID 265237 (WITHAFERIN A) (ΔG = -5.68 kcal/mol), whereas 4W2Q-CID 265237 \n(WITHAFERIN A) showed the weakest binding affinity (ΔG = -1.16 kcal/mol). Negative ΔG values \nindicate favorable binding, with van der Waals interactions being the dominant contributor. The gas-phase \ninteraction energy (GGAS) was strongly favorable for 4GH9 (-13.98 kcal/mol) and 4W2O (-11.95 \nkcal/mol), largely driven by significant van der Waals contributions (VDWAALS: -9.71 and -8.38 kcal/mol, \nrespectively) and electrostatic interactions (EEL: -4.26 and -3.57 kcal/mol). In contrast, 4W2Q displayed \nsubstantially weaker van der Waals (-2.3 kcal/mol) and electrostatic (-1.03 kcal/mol) contributions, \nexplaining its reduced binding affinity. Although the polar solvation energy (EGB) opposed binding in all \nsystems (8.2, 7.35, and 2.47 kcal/mol for 4GH9, 4W2O, and 4W2Q, respectively), this unfavorable \ncontribution was partially offset by favorable non-polar solvation energy (ESURF: -1.22, -1.09, and -0.3 \nkcal/mol). Overall, the stronger van der Waals and electrostatic interactions in 4GH9 and 4W2O \ncompensated for solvation penalties, confirming more stable ligand binding compared to 4W2Q.             \nTable 3: MMGBSA Binding Energy Components\nEnergy \nComponent\n4GH9-CID 265237 \n(WITHAFERIN A)\n4W2Q-CID 265237 \n(WITHAFERIN A)\n4W2O-CID 265237 \n(WITHAFERIN A)\nVDWAALS -9.71 -2.3 -8.38\nEEL -4.26 -1.03 -3.57\nEGB 8.2 2.47 7.35\nESURF -1.22 -0.3 -1.09\nGGAS -13.98 -3.33 -11.95\nGSOLV 6.98 2.17 6.27\nTotal ΔG -6.99 -1.16 -5.68\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 16 \n3.5 Pharmacoinformatics Studies\nDrug-likeness and pharmacokinetic properties of Withaferin A were evaluated using SwissADME and \npkCSM represented in Figure 13. The major drug-likeness parameters that Withaferin A met as expected \nwere acceptable molecular weight, number of hydrogen bond donors and acceptors, and lipophilicity. \nPredicted gastrointestinal absorption was favorable, and cytochrome P450 interaction profiles were within \nacceptable ranges.\nTable 4. ADME and drug-likeness properties of Withaferin A.\nParameters Withaferin A\nFormula C15H12O5\nMolecular weight (g/mol) 272.25\nNum. heavy atoms 20\nNum. H-bond acceptors 5\nNum. H-bond donors 3\nMolar Refractivity 71.57\nPhysico-chemical parameters\nTPSA (Ų) 86.99\nLog P (iLOGP) 1.75\nLog P (XLOGP3) 2.52\nLog P (WLOGP) 2.19\nLog P (MLOGP) 0.71\nLog P (SILICOS-IT) 2.05\nConsensus Log P 1.84\nLipophilicity\nLog P (iLOGP) 1.75\nGI absorption High\nPharmacokinetics\nBBB permeant No\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 17 \nP-gp substrate Yes\nCYP1A2 inhibitor Yes\nCYP2C19 inhibitor No\nCYP2C9 inhibitor No\nCYP2D6 inhibitor No\nLog S (ESOL) -3.49\nSolubility (mg/ml; mol/l) 8.74×10⁻² ; 3.21×10⁻⁴\nClass Soluble\nLog S (Ali) -3.99\nSolubility (mg/ml; mol/l) 2.77×10⁻² ; 1.02×10⁻⁴\nClass Soluble\nLog S (SILICOS-IT) -3.42\nSolubility (mg/ml; mol/l) 1.04×10⁻¹ ; 3.82×10⁻⁴\nWater Solubility\nClass Soluble\nPAINS 0\nBrenk 0\nLeadlikeness Yes\nMedicinal Chemistry\nSynthetic accessibility 3.01\n3.6 Toxicity Analysis of the Selected Metabolite\nToxicity prediction analysis was conducted using pkCSM. Withaferin A had been estimated to be non-\nmutagenic in AMES test and showed acceptable hepatotoxicity and cardiotoxicity profiles. Acute toxicity \npredictions were within permissible computational thresholds.  \nTable 5. Absorption, Distribution, Metabolism, Excretion and Toxicity parameter of selected metabolite.\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 18 \nCategory Model Name Predicted \nValue\nUnit\nAbsorption Water solubility -3.224 Numeric (log mol/L)\nAbsorption Caco2 permeability 1.029 Numeric (log Papp in 10⁻⁶ \ncm/s)\nAbsorption Intestinal absorption (human) 91.31 Numeric (% Absorbed)\nAbsorption Skin permeability -2.742 Numeric (log Kp)\nAbsorption P-glycoprotein substrate Yes Categorical (Yes/No)\nAbsorption P-glycoprotein I inhibitor No Categorical (Yes/No)\nAbsorption P-glycoprotein II inhibitor No Categorical (Yes/No)\nDistribution VDss (human) -0.015 Numeric (log L/kg)\nDistribution Fraction unbound (human) 0.064 Numeric (Fu)\nDistribution BBB permeability -0.578 Numeric (log BB)\nDistribution CNS permeability -2.215 Numeric (log PS)\nMetabolism CYP2D6 substrate No Categorical (Yes/No)\nMetabolism CYP3A4 substrate No Categorical (Yes/No)\nMetabolism CYP1A2 inhibitor Yes Categorical (Yes/No)\nMetabolism CYP2C19 inhibitor No Categorical (Yes/No)\nMetabolism CYP2C9 inhibitor No Categorical (Yes/No)\nMetabolism CYP2D6 inhibitor No Categorical (Yes/No)\nMetabolism CYP3A4 inhibitor No Categorical (Yes/No)\nExcretion Total clearance 0.06 Numeric (log ml/min/kg)\nExcretion Renal OCT2 substrate No Categorical (Yes/No)\nToxicity AMES toxicity No Categorical (Yes/No)\nToxicity Max. tolerated dose (human) -0.176 Numeric (log mg/kg/day)\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 19 \nToxicity hERG I inhibitor No Categorical (Yes/No)\nToxicity hERG II inhibitor No Categorical (Yes/No)\nToxicity Oral rat acute toxicity (LD₅₀) 1.791 Numeric (mol/kg)\nToxicity Oral rat chronic toxicity \n(LOAEL)\n1.944 Numeric (log mg/kg_bw/day)\nToxicity Hepatotoxicity No Categorical (Yes/No)\nToxicity Skin sensitisation No Categorical (Yes/No)\nToxicity T. pyriformis toxicity 0.369 Numeric (log µg/L)\nToxicity Minnow toxicity 2.136 Numeric (log mM)\n4. Discussion  \nPrevious computational studies that focused on Marburg virus (MARV) proteins have mostly used \nmolecular docking approaches and virtual screening methods in order to discover potential inhibitory \ncompounds. For example, several studies targeting VP35 inhibitors have been screened using natural \nproduct libraries, and reported acceptable docking scores for selected hits. However, subsequent analyses \nwere often limited to interaction profiling and ADMET predictions, without performing extensive \nmolecular dynamics validation (24). In a similar approach, repurposing studies on other MARV proteins \nhave relied primarily on docking and pharmacokinetic evaluations to identify potential inhibitors. But, \ncomprehensive long-timescale MD simulations were generally not employed to verify the stability of these \ninteractions under dynamic conditions (25). These approaches are important for early-stage lead discovery. \nHowever, static docking snapshots overlook important factors such as protein flexibility, ligand-induced \nconformational changes, and solvent-mediated interactions, all of which are critical for molecular \nrecognition and binding energetics (7).  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 20 \nA major shortcoming of many previous studies in which only docking scores were used for ranking \ncandidate compounds. Docking-derived poses may represent energetically favorable conformations within \na rigid environment. Yet, these conformations may prove unstable or transient when subjected to solvated, \ndynamic conditions that more accurately mimic the intracellular environment (26). In addition, several \nvirtual screenings have focused on individual viral targets, which can limit the translational relevance of \nthe identified compounds. Because the intricate and multi-protein nature of viral replication and immune \nevasion are complex (27). Although several amino acid-based interaction studies targeting MARV VP35 \nor the viral methyltransferase domain have reported promising binding affinities, these interactions were \nnot consistently validated through extended MD simulations or time-resolved stability analyses (25).\nIn the current study, these shortcomings have been mitigated with a multi-target in silico evaluation of \nWithaferin A against three structurally and functionally diverse MARV proteins: the VP35, and \nnucleoproteins (NP). By integrating molecular docking with extended (100 ns) MD simulations and MM-\nGBSA binding free energy calculations, this current collaborative work offers a dynamic evaluation of the \nligand-protein interactions beyond static docking predictions. \nThe MD analyses revealed stable backbone RMSD profiles, limited residue fluctuations in key binding \nregions, preserved structural compactness, and sustained intermolecular hydrogen bonding across the \nsimulation trajectories. These results confirm that the docked complexes maintained dynamic stability \nunder simulated physiological conditions. Together, these observations suggest that the ligand remained \nstably bound over time, rather than representing transient or non-physiological docking conformations.\nThe MM-GBSA binding free energy calculations offered a quantitative insight into the energetic \ncontributions governing complex formation. Several complexes showed moderate binding free energies, \nlargely driven by van der Waals and electrostatic interactions, whereas solvation penalties reduced the \noverall free energy gain. Notably, one protein-ligand complex exhibited substantially weaker binding \nenergetics, indicating that Withaferin A does not interact uniformly across all MARV targets. This \nvariability highlights the value of a multi-target evaluation and illustrates that structural stability observed \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 21 \nin MD simulations does not always translate to strong binding affinity. Overall, these results suggest that \nWithaferin A exhibits measurable but moderate interaction strength, consistent with lead compounds rather \nthan an inhibitor with relatively high affinity. \nIn addition to interaction analyses, many previous MARV-focused computational studies have lacked \ncomprehensive pharmacoinformatics and toxicity assessments, which are crucial for prioritizing candidates \nin early-stage. To address this, the present study incorporated in silico ADME and toxicity evaluations \nusing SwissADME and pkCSM to evaluate drug-likeness, pharmacokinetics, and safety parameters. The \nresults indicated acceptable drug-like properties and a favorable predicted toxicity profile, supporting the \nsuitability of Withaferin A as an early-stage antiviral lead from a pharmacological perspective. \nNevertheless, these predictions are still by nature computational and must be experimentally confirmed. \nThe results showed good drug-like characteristics and predicted toxicity profile, demonstrating the \nsuitability of Withaferin A as an early-stage antiviral lead from a pharmacological point of view. \nNevertheless, these predictions are still by nature computational and must be experimentally confirmed.  \nTo the best of our knowledge, no previous study has comprehensively evaluated the inhibitory potential of \nWithaferin A against multiple Marburg virus protein targets using an integrated docking, molecular \ndynamics simulation, and MM-GBSA approach. By overcoming key limitations of the previous studies, \ni.e., lack of dynamic validation, single-target focus, and incomplete pharmacoinformatics assessment, this \nstudy provides a more thorough and realistic evaluation of ligand-protein interactions. Although, the \nbinding energies observed were moderate and suggest that Withaferin A may not act as a potent standalone \ninhibitor, it represents a promising scaffold for further structural optimization and experimental validation \nin the context of multi-target antiviral drug discovery against Marburg virus.  \n5. Conclusion\nThe present study used an integrated computational approach to assess the antiviral potential of Withaferin \nA (CID 265237) against three vital proteins of Marburg virus, including VP35 (4GH9), NP (4W2Q and \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 22 \n4W2O). Molecular docking studies demonstrated favorable binding affinities of Withaferin A toward all \ntarget proteins. These interactions illustrate its potential to engage functionally relevant domains associated \nwith viral replication, immune evasion, and virion assembly.\nTo validate the docking predictions, 100 ns molecular dynamics simulations were conducted using \nGROMACS. The simulations exhibited stable RMSD convergence, controlled residue flexibility, sustained \nstructural compactness, and persistent hydrogen bonding over the simulation timeframe. MM-GBSA \nbinding free energy analysis further supported thermodynamically favorable interactions. The binding was \npredominantly driven by van der Waals forces, with electrostatic interactions contributing to overall \ncomplex stability. Such dynamic analyses offer greater assurance regarding the stability and persistence of \nthe protein-ligand complexes beyond static docking predictions.  \nImportantly, the multi-target evaluation approach used in this study distinguishes it from previous single-\ntarget investigations. By exhibiting stable interactions with VP35, and NP; Withaferin A might have the \npotential to interfere with multiple stages of the viral life cycle, which could minimize the chances of \ndeveloping resistance. \nAlthough the results are based on computational predictions and require experimental validation. \nNevertheless, the findings collectively highlight Withaferin A as a promising multi-target novel compound \nagainst Marburg virus. This study represents mechanistic insights and establishes a foundation for future \nresearch conducted both in vitro and in vivo, to develop effective antiviral therapeutics against MARV.\n6. Figure Legends\nFigure 1: Schematic flowchart summarizing the methodological framework throughout the study.\nFigure 2: Structure prediction and validation of 4W2Q protein (i) 3D model, (ii) Ramachandran plot and \n(iii) Errat quality value.\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 23 \nFigure 3: Structure prediction and validation of 4W2O protein (i) 3D model, (ii) Ramachandran plot and \n(iii) Errat quality value.\nFigure 4: Structure prediction and validation of 4GH9 protein (i) 3D model, (ii) Ramachandran plot and \n(iii) Errat quality value.\nFigure 5: Docking conformation of Withaferin A with i. VP35 (4GH9), ii. NP (4W2O), iii. NP (4W2Q).\nFigure 6: 2D interaction diagram of i. VP35 (4GH9), ii. NP (4W2O), iii. NP (4W2Q) with Withaferin A \ncomplex.\nFigure 7: Time-dependent backbone root mean square deviation (RMSD) of the 4GH9-CID 265237, \n4W2Q-CID 265237, and 4W2O-CID 265237 complexes during 100 ns of MD simulation.\nFigure 8: Residue-wise root means square fluctuation (RMSF) profiles of the three complexes during the \nmolecular dynamic simulation.\nFigure 9: Radius of gyration (Rg) profiles of the three protein-ligand complexes over 100 ns of MD \nsimulation.\nFigure 10: Time-dependent solvent-accessible surface area (SASA) of the three protein-ligand complexes \nover the 100 ns molecular dynamics simulation.\nFigure 11: Hydrogen bond profiles for the three protein-ligand complexes during the 100 ns molecular \ndynamics simulation.\nFigure 12: MM-GBSA binding free energy calculations for Withaferin A with VP35 (4GH9), NP (4W2O), \nand NP (4W2Q) of Marburg virus.\nFigure 13: Pharmacokinetic properties of Withaferin A. \n7. Acknowledgements\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 24 \nThe authors sincerely acknowledge the contributions of all co-authors for their involvement in the \nconceptualization, design, analysis, and preparation of this manuscript. We also thank the Department of \nAnimal and Fish Biotechnology, Faculty of Biotechnology and Genetic Engineering, Sylhet Agricultural \nUniversity, for providing the necessary facilities and support to conduct this study. \n8. References\n1. Feldmann H, Geisbert TW. Ebola haemorrhagic fever. Lancet. 2011;377(9768):849-862. \ndoi.org/10.1016/S0140-6736(10)60667-8\n2. World Health Organization. Marburg virus disease. 2023. Available from: \nhttps://www.who.int/news-room/fact-sheets/detail/marburg-virus-disease\n3. De Clercq E, Li G. Approved antiviral drugs over the past 50 years. Clin Microbiol Rev. \n2016;29(3):695-747. https://doi.org/10.1128/cmr.00102-15\n4. Bharat TA, Noda T, Riches JD, Kraehling V, Kolesnikova L, Becker S, et al. Structural dissection \nof Ebola virus and its assembly determinants using cryo-electron tomography. Proc Natl Acad Sci \nU S A. 2012;109(11):4275-4280. https://doi.org/10.1073/pnas.1120453109\n5. Edwards MR, Liu G, Mire CE, Sureshchandra S, Luthra P, Yen B, et al. Differential regulation of \ninterferon responses by Ebola and Marburg virus VP35 proteins. Cell Rep. 2016;14(7):1632-1640. \nhttps://doi.org/10.1016/j.celrep.2016.01.049\n6. Bhunia SS, Saxena M, Saxena AK. Ligand- and structure-based virtual screening in drug discovery. \nIn: Biophysical and computational tools in drug discovery. Springer; 2021. p.281-339. \nDOI:10.1007/7355_2021_130\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 25 \n7. Trott O, Olson AJ. AutoDock Vina: improving the speed and accuracy of docking with a new \nscoring function, efficient optimization, and multithreading. J Comput Chem. 2010;31(2):455-461. \nhttps://doi.org/10.1002/jcc.21334\n8. Van Der Spoel D, Lindahl E, Hess B, Groenhof G, Mark AE, Berendsen HJ. GROMACS: fast, \nflexible, and free. J Comput Chem. 2005;26(16):1701-1718. https://doi.org/10.1002/jcc.20291\n9. Genheden S, Ryde U. The MM/PBSA and MM/GBSA methods to estimate ligand-binding \naffinities. Expert Opin Drug Discov. 2015;10(5):449-461. \nhttps://doi.org/10.1517/17460441.2015.1032936\n10. Navabhatra A, Yangchaiya C. Antiviral drug discovery in emerging economics: challenges and \nopportunities. In: Global Perspectives on Antiviral Drug Development. 2025. p.139-174. \nhttps://doi.10.4018/979-8-3693-9276-8.ch004\n11. Löscher W. Single-target versus multi-target drugs versus combinations of drugs with multiple \ntargets: preclinical and clinical evidence for the treatment or prevention of epilepsy. Front \nPharmacol. 2021;12:730257. https://doi.org/10.3389/fphar.2021.730257\n12. Denaro M, Smeriglio A, Barreca D, De Francesco C, Occhiuto C, Milano G, et al. Antiviral activity \nof plants and their isolated bioactive compounds: an update. Phytother Res. 2020;34(4):742-768. \nhttps://doi.org/10.1002/ptr.6575\n13. Alanazi HH, Elfaki E. The immunomodulatory role of Withania somnifera (L.) Dunal in \ninflammatory diseases. Front Pharmacol . 2023;14:1084757. \nhttps://doi.org/10.3389/fphar.2023.1084757\n14. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, et al. The Protein Data Bank. \nNucleic Acids Res. 2000;28(1):235-242. https://doi.org/10.1093/nar/28.1.235\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 26 \n15. Rose PW, Prlić A, Altunkaya A, Bi C, Bradley AR, Christie CH, et al. The RCSB Protein Data \nBank: integrative view of protein, gene and 3D structural information. Nucleic Acids Res . \n2016;gkw1000. https://doi.org/10.1093/nar/gkw1000\n16. Kim S, Chen J, Cheng T, Gindulyte A, He J, He S, et al. PubChem in 2021: new data content and \nimproved web interfaces. Nucleic Acids Res. 2021;49(D1):D1388-D1395. \nhttps://doi.org/10.1093/nar/gkaa971\n17. O’Boyle NM, Banck M, James CA, Morley C, Vandermeersch T, Hutchison GR. Open Babel: an \nopen chemical toolbox. J Cheminform. 2011;3(1):33. https://doi.org/10.1186/1758-2946-3-33\n18. Patil R, Das S, Stanley A, Yadav L, Sudhakar A, Varma AK. Optimized hydrophobic interactions \nand hydrogen bonding at the target-ligand interface leads the pathways of drug-designing. PLoS \nOne. 2010;5(8):e12029. https://doi.org/10.1371/journal.pone.0012029\n19. Abraham MJ, Murtola T, Schulz R, Páll S, Smith JC, Hess B, et al. GROMACS: high performance \nmolecular simulations through multi-level parallelism from laptops to supercomputers. SoftwareX. \n2015;1:19-25. https://doi.org/10.1016/j.softx.2015.06.001\n20. Zoete V, Cuendet MA, Grosdidier A, Michielin O. SwissParam: a fast force field generation tool \nfor small organic molecules. J Comput Chem. 2011;32(11):2359-2368. \nhttps://doi.org/10.1002/jcc.21816\n21. Wang L, O’Mara ML. Effect of the force field on molecular dynamics simulations of the multidrug \nefflux protein P-glycoprotein. J Chem Theory Comput. 2021;17(10):6491-6508. \nhttps://doi.org/10.1021/acs.jctc.1c00414\n22. Daina A, Michielin O, Zoete V. SwissADME: a free web tool to evaluate pharmacokinetics, drug-\nlikeness and medicinal chemistry friendliness of small molecules. Sci Rep . 2017;7:42717. \nhttps://doi.org/10.1038/srep42717\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\nPage | 27 \n23. Pires DE, Blundell TL, Ascher DB. pkCSM: predicting small-molecule pharmacokinetic and \ntoxicity properties using graph-based signatures. J Med Chem. 2015;58(9):4066-4072. \nhttps://doi.org/10.1021/acs.jmedchem.5b00104\n24. Alsaady IM, Bajrai LH, Alandijany TA, Gattan HS, El-Daly MM, Altwaim SA, et al. \nCheminformatics strategies unlock Marburg virus VP35 inhibitors from natural compound library. \nViruses. 2023;15(8):1739. https://doi.org/10.3390/v15081739\n25. De A, Subramanian S, Nayak P, Pal K. In silico drug repurposing of potential antiviral inhibitors \ntargeting methyltransferase (2′-O-MTase) domain of Marburg virus. In Silico Pharmacol. \n2025;13(2):70. DOI:10.1007/s40203-025-00355-z\n26. Hasan GM, Mohammad T, Zaidi S, Shamsi A, Hassan MI. Molecular docking and dynamics in \nprotein serine/threonine kinase drug discovery: advances, challenges, and future perspectives. \nFront Pharmacol. 2025;16:1696204. https://doi.org/10.3389/fphar.2025.1696204\n27. Kirchmair J, Distinto S, Roman Liedl K, Markt P, Rollinger JM, Schuster D, et al. Development of \nanti-viral agents using molecular modeling and virtual screening techniques. Infect Disord Drug \nTargets. 2011;11(1):64-93. http://dx.doi.org/10.2174/187152611794407782 \n  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted March 7, 2026. ; https://doi.org/10.64898/2026.03.06.710011doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}