In silico drug repurposing targeting fusion and nucleoprotein of human metapneumovirus: A step toward pandemic preparedness

Abstract

BACKGROUND:

Human metapneumovirus (HMPV) is an emerging respiratory pathogen affecting children, elderly individuals, and immunocompromised patients. Despite its disease burden, no antiviral treatment has been approved to date.

OBJECTIVE:

The present study aimed to identify the Food and Drug Administration-approved drugs with potential for repurposing against HMPV by targeting its key structural proteins-fusion (F) and nucleoprotein (N).

MATERIALS AND METHODS:

The crystallographic structures of HMPV fusion (Protein Data Bank [PDB] ID: 5WB0) and nucleoprotein (PDB ID: 5FVD) were retrieved, validated, and subjected to molecular docking. Ligands with favorable binding scores were further evaluated using molecular dynamics simulations and binding-free-energy calculations. Pharmacokinetic and toxicity profiles were predicted to assess their translational viability.

RESULTS:

For the fusion protein, rutin, carbetocin, and acarbose showed strong binding affinities and stable molecular interactions. For the nucleoprotein, mobocertinib, lapatinib, and levetiracetam emerged as top candidates, with mobocertinib showing the most favorable binding energy. Among all, levetiracetam displayed the most drug-like characteristics, including high gastrointestinal absorption, no major cytochrome P450 inhibition, and no violations of Lipinski’s rule.

CONCLUSION:

The study highlights mobocertinib, rutin, and levetiracetam as promising repurposed drugs against HMPV. While mobocertinib exhibited the strongest predicted binding affinity, levetiracetam demonstrated the best pharmacokinetic profile, making it a particularly viable candidate for further experimental validation. These results validate the usefulness of in silico drug repurposing in addressing unmet antiviral needs and warrant preclinical studies to evaluate therapeutic efficacy.

Introduction

Human metapneumovirus (HMPV), a member of the Pneumoviridae family, is an enveloped, negative-sense, single-stranded RNA virus that causes acute respiratory tract infections, particularly in young children, elderly individuals, and immunocompromised patients. Since its discovery in 2001, HMPV has been recognized as a leading cause of bronchiolitis and pneumonia globally, with disease severity comparable to that of respiratory syncytial virus (RSV). Despite its clinical significance, no vaccines or antiviral therapies are currently approved for HMPV.[1,2,3]

The HMPV genome encodes eight proteins, including three surface glycoproteins–fusion (F), glycoprotein (G), and small hydrophobic–and internal proteins such as the nucleoprotein (N), phosphoprotein (P), and large polymerase (L). Among these, the F protein is essential for viral entry and membrane fusion, whereas the N protein plays a critical role in RNA encapsulation and replication [Figure 1]. Both proteins are highly conserved and indispensable for the viral life cycle, making them attractive targets for antiviral drug development.[4,5,6,7,8]

Figure 1.

Schematic representation of the human metapneumovirus replication cycle. The virus attaches to the host cell through αvβ1 integrin (1), undergoes membrane fusion (2) and uncoating (3), followed by transcription (4), translation (5), replication of the viral genome (6), and assembly of progeny virions which bud from the host cell surface (7). Viral proteins fusion, glycoprotein, small hydrophobic, M, and the ribonucleoprotein complex (N, phosphoprotein, large polymerase) are involved in key stages of the viral lifecycle

F protein, synthesized from F₀ precursor protein, is required for the entry of the ribonucleoprotein (RNP) of virus into the host cell.[9,10] It contains a highly conserved Arg-Gly-Asp motif that binds to αvβ1 integrin protein present on the host cell surface as a cellular receptor.[11,12] Viral RNP containing negative-sense viral RNA (vRNA) genome is released into the host cell cytoplasm.[13] The HMPV RNA genome is encapsulated by the viral nucleoprotein, and it forms an RNA-N complex that serves as a template for genome replication and transcription by the viral polymerase complex, which is formed by L protein and its main cofactor, P protein, which is known as the RNA-dependent RNA polymerase complex. P protein, consisting of 30 amino acids (aa) in length, interacts with N protein (length 394 aa) at Glu50, Asp128, Arg132, Met135, Arg151, and Ser153.[14,15]

Conventional drug discovery for emerging viruses is time-intensive and cost-prohibitive. In contrast, drug repurposing offers a rapid and efficient strategy by identifying new therapeutic uses for approved drugs with known safety profiles. In this context, in silico screening combined with molecular dynamics (MD) simulations and pharmacokinetic predictions can serve as a powerful platform to prioritize lead candidates.

This study aimed to identify the Food and Drug Administration (FDA)-approved drugs with repurposing potential against HMPV by targeting the fusion and nucleoproteins through a comprehensive in silico approach involving docking, MM/GBSA binding energy calculations, MD simulations, and Absorption, Distribution, Metabolism, Excretion and Toxicity (ADMET) profiling.

Materials and Methods

Protein validation and preparation

The HMPV proteins (fusion protein and nucleoprotein) were selected from the RCSB Protein Data Bank (PDB). The available three-dimensional (3D) X-ray crystallographic structures of protein on RCSB.org under PDB IDs: 5WB0 for F protein, PDB ID: 5FVD for N protein, were selected based on various parameters such as resolution, Ramachandran plot score, R free, and sidechain outliers.[16,17] Further, both the PDBs were validated using the PROCHECK tool saves v6.0.[18,19] The proteins for docking purpose were prepared using the protein preparation wizard module of Schrodinger Maestro version 2020.3 (Schrodinger Maestro is Global Company, Schrödinger, Inc). Water molecules were deleted beyond 5 Å from protein, het states were generated using the Epik module keeping pH at 7.0 ± 2.0, and preprocessing was initiated. The chain of interest was kept, the rest all were deleted in the review and modify tab. Finally, refinement was done, keeping PROPKA pH at 7.0 optimization was done, hetero atoms and water with H-bonds fewer than 3 Å to nonwater were removed, and keeping the optimized potential for liquid simulations 3e force field, minimization was performed.[20]

Ligand library preparation

The 3D structures of the FDA approved drug library were downloaded from drug databank. For virtual screening, the ligand library was prepared using the ligprep tool of Schrodinger Maestro software version 2020.03. Further, properties involving pharmacology and pharmacokinetic behavior based on the absorption, distribution, metabolism, and excretion were predicted using the Qikprop module and SwissADME webserver.[21,22]

Virtual screening workflow

The virtual screening workflow approach was used to identify the potential inhibitors. The ligands went under high-throughput virtual screening and standard precision (SP). Around 50% of the ligands having a binding affinity were considered for further screening, and the top 5% of the molecules were considered for further screening using good docking score, Molecular mechanics generalized boltzmann surface area (MMGBSA) score, and pharmacokinetic properties.[23]

Binding free energy estimation

Binding affinities of selected ligands were calculated using MMGBSA (molecular mechanics generalized born surface area) with the Variable-dielectric Generalized Born (VSGB) 2.0 solvation model in Prime. Docking poses from SP docking were used as input.

ADMET and drug-likeness prediction

Pharmacokinetic and physicochemical properties were predicted using QikProp (Schrödinger) and SwissADME (http://www.swissadme.ch). Toxicity prediction was done using the ProTox-II web server. Lipinski’s rule of five and cytochrome P450 (CYP) interactions were analyzed to assess drug-likeness and potential metabolism liabilities.[22]

Root mean square deviation

Root mean square deviation (RMSD) measures the average positional deviation of a group of atoms in each trajectory frame relative to a reference frame. It provides insights into the structural stability of a molecular system over the course of a simulation. For a given frame x, RMSD is calculated using the positions of N selected atoms after aligning them to the reference structure, typically the initial frame at time t = 0. This process is repeated for every frame in the trajectory.

The RMSD of the protein is calculated by aligning all trajectory frames to the backbone of the reference frame and then computing the RMSD based on the selected atoms. Monitoring protein RMSD helps assess the structural stability and conformational changes during the simulation. A stable simulation typically shows RMSD values that plateau near a thermal average toward the end of the run.[24] For small, globular proteins, fluctuations around 1.3 Å are generally acceptable. Significantly higher deviations may indicate major structural rearrangements. If the protein RMSD continues to trend upward or downward near the simulation’s end, it may suggest that the system has not yet reached equilibrium, implying that the simulation length is insufficient for conclusive analysis.

Root mean square fluctuation

To examine atomic-level fluctuations within a ligand, ligand root mean square fluctuation (L-RMSF) is an effective metric. It quantifies the positional variability of individual ligand atoms throughout the simulation, offering a breakdown based on atom-specific fluctuations, often visualized in a two-dimensional structural representation. These atom-wise fluctuations provide valuable insights into the flexibility of ligand fragments, their interaction dynamics with the protein, and the potential entropic contributions to binding. For analysis, the protein-ligand complex is first aligned to the reference protein backbone, and root mean square fluctuation (RMSF) is then computed using the heavy atoms of the ligand.

Radius of gyration

The radius of gyration (ROG) quantifies the spatial distribution of a molecule’s atoms relative to its center of mass, effectively describing how extended or compact the structure is. It is calculated as the root mean square distance of each atom from the molecule’s center of mass. A lower ROG value indicates that the atoms are more tightly clustered around the center, reflecting a compact or globular structure. In contrast, a higher ROG value suggests that the atoms are more dispersed, pointing to an extended or flexible conformation.

Principal component analysis and dynamic cross-correlation matrix

To investigate the large-scale conformational motions and inter-residue dynamic behavior of the protein systems, principal component analysis (PCA) and dynamic cross-correlation matrix (DCCM) analyses were performed using trajectories obtained from MD simulations. These analyses were carried out using getting VMD software and the Bio3D library in R.

For PCA, the positional covariance matrix was constructed from the backbone atoms of the protein after fitting all trajectory frames to a reference structure to remove translational and rotational motions. Eigenvectors and eigenvalues were computed from this matrix to identify the dominant collective motions. The first few principal components (typically PC1 and PC2 or PC3) that contribute most to the overall motion were analyzed to visualize the conformational sampling and flexibility of each system.

DCCM analysis was conducted by calculating the time-correlated fluctuations between Cα atoms of residue pairs across the simulation trajectory. The correlation coefficient values range from −1 to +1, where positive values indicate correlated motions and negative values reflect anticorrelated motions. The DCCM plots were generated using Bio3D in R and custom Python scripts to visualize differences in internal dynamic coordination between the apo and ligand-bound complexes. These techniques allowed a deeper understanding of how ligand binding alters the dynamic behavior and communication networks within the HMPV nucleoprotein (5FVD) and fusion protein (5WB0).

Ethics statement

As this study is entirely computational and does not involve any animal or human subjects, ethical approval was not required.

Results

Protein validation

To ensure the accuracy and suitability of the selected protein structures for virtual screening and binding affinity prediction, both the HMPV fusion protein (PDB ID: 5WB0, Chain F) and nucleoprotein (PDB ID: 5FVD, Chain A) were subjected to structural validation using PROCHECK and ERRAT2.

Validation of fusion protein (protein data bank ID: 5WB0)

The PROCHECK analysis [Figure 2b] revealed that 91.9% of residues were located in the most favoured regions of the Ramachandran plot, 7.6% in additional allowed regions, and only 0.3% each in generously allowed and disallowed regions. The G-factor values were within acceptable limits (dihedrals: −0.07; covalent: 0.56; overall: 0.18), indicating good stereochemical quality. All planar groups were within the acceptable range (100%), and a total of 13 bad contacts were noted. The ERRAT2 analysis [Figure 2a] provided an overall quality factor of 97.368%, suggesting a high-quality model. Most residues had error values below the 95% rejection threshold, with only minor deviations observed between residues 75–90 and 270–310. These small variations were not considered structurally significant [Figure 2a and b].

Figure 2.

Protein validation results: (a and b) Illustrate the ERRAT2 and PROCHECK validation results, respectively, for the fusion protein (PDB ID: 5WB0). (c and d) Present the corresponding ERRAT2 and PROCHECK validation outcomes for the nucleoprotein (PDB ID: 5FVD)

Validation of nucleoprotein (protein data bank ID: 5FVD)

The PROCHECK results for the nucleoprotein [Figure 2d] indicated excellent geometry, with 97.2% of residues in the most favored regions, 2.8% in allowed regions, and 0% in generously allowed or disallowed regions. No residues were flagged as outliers. The overall G-factor was 0.36 (dihedrals: 0.21; covalent: 0.59), and all planar groups remained within the ideal limit. While a total of 93 bad contacts were identified, the overall geometry remained robust and acceptable. The ERRAT2 quality assessment [Figure 2c] produced an exceptionally high score of 99.280%, confirming that the structure is reliable for structure-based drug design. Only a minor region around residue 40 exceeded the 95% error threshold, which did not significantly affect the global model quality [Figure 2c and d].

Molecular docking and MMGBSA binding energy analysis

Fusion protein (protein data bank ID: 5WB0)

The docking analysis against the fusion protein identified several FDA-approved drugs with notable binding affinities. Nicotinamide adenine dinucleotide (NAD) (NADH) exhibited the best docking score of −6.796 kcal/mol, followed by carbetocin (−6.698 kcal/mol) and rutin (−6.516 kcal/mol). Although other ligands such as ceftolozane, dipyrithione, and acarbose demonstrated moderate docking scores (ranging from −6.192 to −5.804 kcal/mol), their binding profiles remained within acceptable thresholds [Figure 3a]. To complement docking, MMGBSA binding free energy calculations were performed. Among all screened compounds, rutin showed the most favourable binding energy (−54.6 kcal/mol), followed closely by acarbose (−53.09 kcal/mol) and carbetocin (−51.54 kcal/mol) [Figure 3b]. These results suggest thermodynamically stable interactions of these ligands with the fusion protein. 3D binding interaction visualizations further revealed detailed hydrogen bonding and hydrophobic interactions stabilizing the ligand-receptor complexes for carbetocin, rutin, and acarbose [Figure 3c-e]

Figure 3.

Docking and MMGBSA interaction analysis of ligands targeting HMPV fusion protein (PDB ID: 5WB0). (a) Docking scores (kcal/mol) of selected ligands, with NADH, Carbetocin, and Rutin showing the best affinities. (b) MMGBSA binding free energy analysis revealed Rutin, Acarbose, and Carbetocin as the most stable binders. 3D interaction diagrams of Carbetocin (c), Rutin (d) and Acarbose (e) showing their binding within the fusion protein binding site

Nucleoprotein (protein data bank ID: 5FVD)

Docking of selected compounds to the nucleoprotein revealed levetiracetam as the top-ranked molecule (−5.603 kcal/mol), followed by mobocertinib (−5.652 kcal/mol) and lapatinib (−5.751 kcal/mol). Other ligands such as isavuconazonium and cangrelor showed slightly lower docking scores but remained within the active range [Figure 4a]. The MMGBSA energy analysis confirmed mobocertinib as the most stable binder (−42.55 kcal/mol), followed by lapatinib (−40.75 kcal/mol) and levetiracetam (−20.57 kcal/mol) [Figure 4b]. Mobocertinib and lapatinib, both kinase inhibitors, formed stable complexes, as evidenced by multiple polar and nonpolar interactions in the binding pocket. 3D interaction mapping highlighted the binding poses of levetiracetam, lapatinib, and mobocertinib within the nucleoprotein active site [Figure 4c-e], supporting their potential repurposing as HMPV inhibitors.

Figure 4.

Docking and MMGBSA interaction analysis of ligands targeting HMPV nucleoprotein (PDB ID: 5FVD). (a) Docking scores (kcal/mol) for top-ranked ligands, with levetiracetam, mobocertinib, and lapatinib performing well. (b) MMGBSA binding energy analysis showing mobocertinib and lapatinib as the most thermodynamically favorable candidates. (c-e) 3D interaction views of levetiracetam (c), lapatinib (d), and mobocertinib (e) docked into the nucleoprotein pocket

ADMET and drug-likeness evaluation of fusion protein ligands

To evaluate the pharmacokinetic suitability and oral bioavailability of the lead compounds identified through docking and MMGBSA analysis, six FDA-approved ligands were subjected to ADMET profiling. These include rutin, carbetocin, and acarbose (targeting the fusion protein, PDB: 5WB0), and mobocertinib, levetiracetam, and lapatinib (targeting the nucleoprotein, PDB: 5FVD) [Table 1].

Table 1.

Absorption, distribution, metabolism, excretion, and toxicity and drug-likeness properties of the top six repurposed ligands targeting human metapneumovirus fusion and nucleoprotein

Properties/ligands	Rutin	Carbetocin	Acarbose	Mobocertinib	Levetiracetam	Lapatinib
Physicochemical properties
Formula	C27H30O16	C45H69N11O12S	C25H43NO18	C32H39N7O4	C8H14N2O2	C29H26CIFN4O4S
Molecular weight (g/moL)	610.52	988.16	645.40	585.70	170.21	581.06
Lipophilicity
Log Po/w	−1.51	−1.48	−6.06	3.91	0.10	5.19
Water solubility
Solubility	Soluble	Poorly soluble	Soluble	Poorly soluble	Soluble	Insoluble
Pharmacokinetics
GI absorption	Low	Low	Low	High	High	Low
BBB permeant	No	No	No	No	No	No
P-gp substrate	Yes	Yes	Yes	Yes	No	No
CYP1A2 inhibitor	No	No	No	No	No	No
CYP2C19 inhibitor	No	No	No	Yes	No	Yes
CYP2C9 inhibitor	No	No	No	Yes	No	Yes
CYP2D6 inhibitor	No	No	No	Yes	No	Yes
CYP3A4 inhibitor	No	No	No	Yes	No	Yes
Druglikeness
Lipinski	No; 3 violations	No; 3 violations	No; 3 violations	No; 2 violations	Yes; 0 violations	Yes; 1 violation

Physicochemical and lipophilicity profile

The molecular weights of most ligands exceeded the ideal threshold of 500 g/mol, except levetiracetam (170.21 g/mol), which aligns well with drug-likeness parameters. Mobocertinib (585.70 g/mol), carbetocin (988.16 g/mol), lapatinib (581.06 g/mol), rutin (610.52 g/mol), and acarbose (645.40 g/mol) all exceeded this limit, indicating potential issues with permeability and bioavailability. Lipophilicity varied significantly as acarbose had a very low log P (−6.06), suggesting high hydrophilicity. Rutin and carbetocin were also hydrophilic (Log P < −1.5). In contrast, lapatinib (5.19) and mobocertinib (3.91) were highly lipophilic. Levetiracetam showed an optimal Log P of 0.10. Water solubility was acceptable only for levetiracetam and rutin, whereas carbetocin and mobocertinib were poorly soluble, and lapatinib was insoluble.

Pharmacokinetics

Gastrointestinal (GI) absorption: high GI absorption was observed for mobocertinib and levetiracetam, whereas rutin, carbetocin, acarbose, and lapatinib exhibited poor intestinal absorption. Mobocertinib, rutin, carbetocin, and acarbose were identified as P-glycoprotein (P-gp) substrates, indicating possible efflux issues. Mobocertinib and lapatinib inhibited multiple CYP enzymes (CYP2C19, CYP2C9, CYP2D6, CYP3A4), posing risks of drug–drug interactions. In contrast, rutin, carbetocin, acarbose, and levetiracetam showed no significant CYP inhibition, making them metabolically safer candidates.

Drug-likeness assessment

Levetiracetam fully satisfied Lipinski’s criteria (0 violations), supporting its favorable oral bioavailability. Rutin, carbetocin, and acarbose violated three rules (e.g., NorO >10, NHorOH >5), suggesting poor absorption and permeability. Mobocertinib and lapatinib had 2 and 1 violations, respectively, indicating marginal deviations.

Molecular dynamics simulation analysis of human metapneumovirus fusion protein

To investigate the structural stability and binding dynamics of potential inhibitors targeting the HMPV fusion protein, MD simulations were performed for the Apo form and three selected ligands, acarbose, carbetocin, and rutin, over a 100 ns time scale. Key structural descriptors, including RMSD, RMSF, radius of gyration (ROG), and binding free energy (ΔG), were evaluated to assess the conformational behavior and ligand interaction strength [Table 2 and Figure 5].

Table 2.

Molecular dynamics simulation metrics for human metapneumovirus fusion protein (protein data bank ID: 5WB0)

Complex	RMSD (nm)	RMSF (nm)	ROG (nm)	Binding free energy (kJ/moL)
Apo	0.298±0.047	0.199±0.084	3.009±0.033	-
Acarbose	0.266±0.040	0.132±0.059	2.998±0.032	−3.936±5.026
Carbetocin	0.342±0.086	0.166±0.071	3.022±0.037	−22.889±5.507
Rutin	0.310±0.064	0.175±0.076	3.005±0.051	−22.770±4.455

RMSD=Root mean square deviation, RMSF= Root mean square fluctuation, ROG=Radius of gyration

Figure 5.

Molecular dynamics simulation analysis of human metapneumovirus (HMPV) fusion protein (Protein Data Bank ID: 5WB0). (a) Root mean square deviation (RMSD) plots showing backbone stability of the Apo form and ligand-bound complexes (acarbose, carbetocin, rutin) across 100 ns. Carbetocin showed the highest RMSD (0.342 ± 0.086 nm), suggesting increased flexibility, while Acarbose remained most stable (0.266 ± 0.040 nm). (b) Root mean square fluctuation representing residue-level flexibility. The apo form displayed the highest fluctuation (0.199 ± 0.0837 nm), while the acarbose-bound complex showed reduced atomic movement (0.132 ± 0.059 nm), indicating local stabilization. (c) Radius of gyration (Rg) indicating compactness of the protein-ligand systems. All complexes retained structural compactness, with values ranging from 2.998 ± 0.032 nm (acarbose) to 3.022 ± 0.037 nm (carbetocin). (d) Binding free energy (ΔG) profiles computed via MM/PBSA. carbetocin (−22.889 ± 5.507 kJ/mol) and rutin (−22.770 ± 4.455 kJ/mol) exhibited stronger binding compared to acarbose (−3.936 ± 5.026 kJ/mol), suggesting higher affinity toward the HMPV fusion protein

The RMSD results revealed moderate structural stability across all systems. Acarbose showed the lowest RMSD value (0.266 ± 0.040 nm), indicating minimal deviation from the initial conformation. In contrast, carbetocin exhibited the highest RMSD (0.342 ± 0.086 nm), suggesting greater conformational flexibility. The apo protein and rutin-bound complex displayed intermediate values of 0.298 ± 0.047 nm and 0.310 ± 0.064 nm, respectively [Figure 5a].

RMSF values reflected residue-level flexibility, with the apo structure showing the highest average fluctuation (0.199 ± 0.084 nm). Ligand-bound complexes demonstrated reduced flexibility, particularly acarbose (0.132 ± 0.059 nm), indicating improved stabilization upon binding [Figure 5b].

The radius of gyration (ROG) values remained relatively constant across all systems, reflecting consistent structural compactness. Acarbose again showed the most compact form (2.998 ± 0.032 nm), followed closely by rutin and the apo structure, whereas carbetocin presented a slightly more extended profile (3.022 ± 0.037 nm) [Figure 5c].

Regarding binding free energy, both carbetocin and rutin exhibited strong binding affinities, with ΔG values of –22.889 ± 5.507 kJ/mol and –22.770 ± 4.455 kJ/mol, respectively. In contrast, acarbose showed a significantly weaker binding interaction (–3.936 ± 5.026 kJ/mol), suggesting lower affinity towards the fusion protein [Figure 5d].

Summary of structural and energetic parameters obtained from 100 ns MD simulations of the apo form and ligand-bound complexes with acarbose, carbetocin, and rutin. RMSD and RMSF represent global and residue-level atomic deviations, respectively. Radius of gyration (ROG) reflects the overall compactness of the protein, and binding free energy was calculated using the MM/PBSA method. Values are reported as mean ± standard deviation deviation.

Molecular dynamics simulation analysis of human metapneumovirus nucleoprotein

To evaluate the dynamic stability and interaction profiles of ligand-bound and unbound forms of the HMPV nucleoprotein, classical MD simulations were conducted for 100 ns. The apo form and three top-performing ligands–lapatinib (DB01224), levetiracetam (DB01202), and mobocertinib (DB15465)–were included for comparative assessment. Key structural and energetic descriptors, including RMSD, RMSF, radius of gyration (ROG), and binding free energy (ΔG), were analyzed [Table 3 and Figure 6]

Table 3.

Molecular dynamics simulation metrics for human metapneumovirus nucleoprotein (protein data bank ID: 5FVD)

Complex	RMSD (nm)	RMSF (nm)	ROG (nm)	Binding free energy (kJ/moL)
Apo	0.271±0.054	0.136±0.058	2.333±0.014
Lapatinib	0.246±0.037	0.103±0.038	2.326±0.011	−12.575±5.236
Levetiracetam	0.283±0.044	0.125±0.040	2.292±0.011	−12.575±5.236
Mobocertinib	0.254±0.060	0.131±0.047	2.326±0.016	−52.384±4.310

RMSD=Root mean square deviation, RMSF= Root mean square fluctuation, ROG=Radius of gyration

Figure 6.

Molecular dynamics simulation analysis of human metapneumovirus nucleoprotein (Protein Data Bank ID: 5FVD). (a) Root mean square deviation (RMSD) plots showing structural deviations of the apo form and ligand-bound complexes (lapatinib, levetiracetam, mobocertinib) over 100 ns. Lapatinib exhibited the lowest RMSD (0.246 ± 0.037 nm), indicating superior conformational stability. (b) Root mean square fluctuation analysis reflecting per-residue flexibility. The apo form showed the highest atomic fluctuations (0.136 ± 0.058 nm), while lapatinib again displayed the least fluctuation (0.103 ± 0.038 nm), suggesting tighter binding-induced rigidity. (c) Radius of gyration (Rg) plots representing structural compactness. All systems remained stably folded, with levetiracetam showing the most compact profile (2.292 ± 0.011 nm). (d) Binding free energy (ΔG) profiles computed using MM/PBSA. Mobocertinib exhibited the strongest interaction (−52.384 ± 4.310 kJ/mol), followed by lapatinib and levetiracetam (both −12.575 ± 5.236 kJ/mol), indicating favorable binding energetics

The RMSD profiles indicated that all protein-ligand complexes maintained structural stability throughout the simulation. Lapatinib exhibited the lowest average RMSD value (0.246 ± 0.037 nm), indicating high conformational stability, whereas levetiracetam showed a slightly elevated RMSD (0.283 ± 0.044 nm), suggesting moderate deviation. The apo protein had an RMSD of 0.271 ± 0.054 nm, whereas the mobocertinib-bound complex displayed a comparable value (0.254 ± 0.060 nm).

RMSF analysis revealed that ligand binding reduced atomic fluctuations in comparison to the apo form. The lapatinib complex had the lowest RMSF (0.103 ± 0.038 nm), indicating minimal residue-level fluctuations. RMSF values for levetiracetam and mobocertinib were 0.125 ± 0.040 nm and 0.131 ± 0.047 nm, respectively, whereas the apo form showed the highest fluctuation (0.136 ± 0.058 nm).

The radius of gyration (ROG) values remained stable across all systems, reflecting consistent compactness. Levetiracetam showed the most compact structure (2.292 ± 0.011 nm), whereas apo, lapatinib, and mobocertinib had slightly higher but comparable ROG values (ranging from 2.326 to 2.333 nm).

Finally, the binding free energy (ΔG) calculated using MM/PBSA demonstrated that mobocertinib exhibited the strongest binding affinity (−52.384 ± 4.310 kJ/mol), followed by lapatinib and levetiracetam (both −12.575 ± 5.236 kJ/mol). These results underscore Mobocertinib’s potential as a high-affinity binder to the HMPV nucleoprotein.

This table presents key structural and binding parameters derived from 100 ns MD simulations of the apo form and ligand-bound complexes (lapatinib, levetiracetam, and mobocertinib). RMSD and RMSF quantify global and residue-specific structural deviations, respectively. Radius of gyration (ROG) reflects protein compactness, whereas binding free energy was calculated using the MM/PBSA method. Data are presented as mean ± standard deviation.

Binding free energy decomposition (MM/PBSA)

To evaluate the binding energetics of selected ligands against the HMPV fusion protein (5WB0) and nucleoprotein (5FVD), MM/PBSA energy decomposition was performed. The results are summarized in Figure 7.

Figure 7.

MM/PBSA binding free energy decomposition analysis of human metapneumovirus (HMPV) proteins. (a) Energy decomposition profiles of ligand-bound complexes with the HMPV fusion protein (Protein Data Bank ID: 5WB0). Contributions from van der Waals interactions, electrostatic energy, polar solvation energy, and Solvent accessible surface area (SASA) energy are shown for acarbose, carbetocin, and rutin. Among these, carbetocin and rutin exhibited significantly favorable total binding energies (−22.889 and −22.770 kJ/mol, respectively), primarily driven by strong van der Waals and SASA interactions. (b) Energy decomposition of ligand interactions with the HMPV nucleoprotein (PDB ID: 5FVD). Lapatinib and mobocertinib displayed the most favorable binding energies (−61.374 and −52.384 kJ/mol, respectively), attributed to strong van der Waals forces and substantial SASA contributions. Levetiracetam showed minimal interaction energy, indicating a weaker binding profile. Energy values were computed using the MM/PBSA approach over 100 ns MD simulation trajectories

For 5WB0, carbetocin and rutin exhibited favorable total binding free energies (−22.889 kJ/mol and −22.770 kJ/mol, respectively), primarily driven by van der Waals and Solvent accessible surface area (SASA) contributions. Acarbose showed a much weaker interaction (−3.936 kJ/mol), suggesting poor binding affinity [Figure 7a].

In the 5FVD complexes, lapatinib demonstrated the strongest binding affinity (−61.374 kJ/mol), followed by mobocertinib (−52.384 kJ/mol), both of which had substantial van der Waals and SASA contributions. Levetiracetam showed minimal binding (−12.575 kJ/mol), consistent with its modest interaction energies [Figure 7b].

These results further support the stabilizing effect of lapatinib and mobocertinib on the nucleoprotein, and carbetocin/rutin on the fusion protein.

Principal component analysis and dynamic cross-correlation matrix analysis

To investigate large-scale motions and residue interaction dynamics of the HMPV fusion protein (PDB ID: 5WB0), PCA and DCCM analyses were conducted for the apo form and its ligand-bound complexes with acarbose, carbetocin, and rutin.

In the PCA results, the apo protein explored a broader conformational space, with the first and third principal components accounting for 27.12% and 11.32% of the total variance, respectively. Ligand-bound systems exhibited more localized and compact motion patterns. Specifically, acarbose-bound complex showed PC1 = 23.78% and PC3 = 7.48%, carbetocin displayed PC1 = 24.12%, PC3 = 7.21%, and rutin accounted for PC1 = 23.40%, PC3 = 7.35%. This reduction in principal component variance in ligand-bound systems suggests restricted motion and enhanced conformational stability upon ligand binding [Figure 8a].

Figure 8.

Principal component analysis (PCA) and dynamic cross-correlation matrix (DCCM) of human metapneumovirus (HMPV) fusion protein (Protein Data Bank ID: 5WB0). (a) PCA plots showing the conformational distribution of the apo form and ligand-bound complexes (acarbose, carbetocin, rutin) along the top three principal components (PC1, PC2, PC3). The percentage of variance explained by each principal component is labeled on the axes. Scree plots (rightmost column) display the eigenvalue ranking and proportion of variance captured by dominant motions. Ligand-bound systems exhibit more restricted sampling compared to the apo form, indicating enhanced conformational stability. (b) DCCM maps representing correlated (blue) and anticorrelated (light yellow) motions between residue pairs for each system. Compared to the apo structure, ligand-bound complexes show increased positive correlation regions, suggesting ligand-induced stabilization and modified internal dynamic communication across the HMPV fusion protein

DCCM heatmaps revealed that the apo form exhibited extensive anticorrelated motions between several distant residues, reflecting higher intrinsic flexibility. In contrast, ligand-bound complexes–especially carbetocin and rutin–demonstrated increased positive correlation among adjacent and functionally important residues. This indicates ligand-induced stabilization and more coherent internal dynamics, suggesting enhanced allosteric communication within the protein upon binding [Figure 8b].

Principal component analysis and dynamic cross-correlation matrix analysis –5FVD (human metapneumovirus nucleoprotein)

To further understand the conformational flexibility and internal dynamics of the HMPV nucleoprotein (PDB ID: 5FVD), PCA and DCCM analyses were performed for the apo protein and its ligand-bound complexes with levetiracetam, lapatinib, and mobocertinib.

PCA revealed that the apo form exhibited broader motion with PC1 = 27.42%, PC2 = 18.84%, and PC3 = 8.54% of the total variance. Among ligand-bound systems, mobocertinib showed the most pronounced restriction along secondary components (PC2 = 9.13%, PC3 = 7.99%) while capturing the largest primary movement (PC1 = 38.56%), indicating a dominant constrained motion along a single direction. Levetiracetam retained a similar primary motion to apo (PC1 = 27.42%) but had slightly higher secondary component variability (PC2 = 20.93%, PC3 = 9.37%). Lapatinib exhibited reduced variability overall, with PC1 = 25.98%, PC2 = 14.88%, and PC3 = 7.84% [Figure 9a].

Figure 9.

Principal component analysis (PCA) and dynamic cross-correlation matrix (DCCM) of human metapneumovirus nucleoprotein (Protein Data Bank ID: 5FVD). (a) PCA plots representing the conformational sampling along principal components PC1, PC2, and PC3 for the apo form and ligand-bound complexes (lapatinib, levetiracetam, and mobocertinib). The percentage variance explained by each component is indicated. The scree plots (rightmost column) show the eigenvalue distribution and the proportion of variance accounted for by the dominant modes of motion. (b) DCCM maps depicting the dynamic correlation of Cα atomic fluctuations between residue pairs. Positive correlations (blue) indicate residues moving in the same direction, whereas negative correlations (yellow) represent anticorrelated motions. Ligand-bound systems show enhanced correlated motions compared to the apo form, indicating stabilization and altered dynamic communication across the protein

DCCM analysis showed that ligand binding significantly altered internal residue correlation patterns. The apo protein displayed wider regions of anticorrelated motions, whereas all ligand-bound complexes–particularly mobocertinib–exhibited increased positive correlations, indicating more synchronized residue interactions and improved structural coherence upon ligand association [Figure 9b].

Discussion

HMPV remains a major cause of pediatric and geriatric lower-respiratory disease, yet no licensed antivirals or vaccines have reached the clinic, underscoring the need for rapid therapeutic discovery. Recent reviews highlight the continuing global burden of HMPV and the paucity of drug-development pipelines, despite encouraging proof-of-concept screens for small-molecule entry blockers and replication inhibitors.[25]

Fusion protein (5WB0) as an entry-blockade target

The prefusion F glycoprotein drives viral–cell membrane fusion and, like its RSV counterpart, exposes a druggable cavity adjacent to the internal fusion peptide.[26] Our PROCHECK/ERRAT2 validation confirmed that the available crystal structure (97.4% ERRAT2 score; 91.9% favoured Ramachandran residues) is of sufficient stereochemical quality for atomic-level modeling. This is critical because subtle backbone errors can exaggerate docking pose artefacts. Docking and MM/GBSA ranked Rutin, Carbetocin, and Acarbose as the most favorable binders, with ΔG values of −54.6, −51.5 and −53.1 kcal mol⁻¹, respectively. Rutin, a polyphenolic flavanol with documented broad-spectrum antiviral activity against influenza, enteroviruses, and SARS-CoV-2, has not been explored against HMPV, but its high van der Waals/SASA contribution suggests it can snugly tether the fusion peptide in a prefusion-locked state.[27] Carbetocin, an eight-residue oxytocin analog approved for postpartum atony, engages the same pocket predominantly through hydrophobic contacts but, unlike rutin, maintains an extended backbone, which may explain the higher RMSD (0.342 nm) observed during MD simulations. Acarbose, despite its impressive enthalpic score, showed weak MM/PBSA ΔG (−3.9 kJ mol⁻¹) and large polar-solvation penalties, indicating that its apparent affinity is likely an artefact of docking in a solvent-stripped environment.

One-hundred-nanosecond MD trajectories confirmed that rutin and carbetocin stabilize the fusion protein–reflected by reduced RMSF and narrowed principal-component sampling–whereas the acarbose complex undergoes rapid breathing motions that erode binding free energy. PCA and DCCM analyses further demonstrated that ligand binding dampened the long-range anticorrelated motions characteristic of the apo protein, replacing them with positive correlations that signal allosteric rigidification. Such ligand-induced “stiffening” has been proposed as a hallmark of productive entry inhibitors in related viral fusion systems.[28]

ADMET profiling reveals a trade-off between potency and developability. Rutin and carbetocin violate three of five Lipinski rules and are predicted to have low intestinal absorption; however, both are amenable to nasal or parenteral administration, a route already exploited for peptide-based RSV F inhibitors. Levetiracetam–although not among the most potent F binders–emerged as the only compound that fully satisfies Lipinski criteria, highlighting the importance of parallel optimisation for permeability once in vitro efficacy is confirmed.

Nucleoprotein (5FVD) as a replication target

The nucleoprotein (N) encapsidates viral RNA and nucleates inclusion-body assembly through interactions with the phosphoprotein (P), making it indispensable for genome replication.[29,30] Structural validation (99.3% ERRAT2; 97.2% favored residues) allowed confident assessment of ligand binding. Kinase inhibitors lapatinib and mobocertinib achieved markedly favorable MM/PBSA energies (−61.4 kJ mol⁻¹ and −52.4 kJ mol⁻¹, respectively) and exhibited the lowest RMSD/RMSF values across all nucleoprotein simulations, indicating tight, deformation-resistant complexes.

Lapatinib has already been reported to suppress replication of coronaviruses and orthomyxoviruses by modulating host-cell EGFR/HER2 signaling pathways, making it an attractive repurposing candidate.[31,32] Although antiviral data for mobocertinib are currently lacking, its superior binding energetics and cell-permeable physicochemical profile justify experimental testing. The fact that both molecules are approved oral drugs expedites pharmacokinetic translation, but caution is warranted owing to their broad CYP-inhibition spectra and potential cardiotoxicity.[33]

MM/PBSA per-residue decomposition showed that van der Waals and nonpolar SASA terms dominate the favorable energy landscape for both proteins, suggesting that future lead optimization efforts should prioritise hydrophobic scaffold elaboration and polar-surface minimization to enhance receptor desolvation while preserving key hydrogen bond anchors.

Study limitations and future directions

Our pipeline provides multi-layered in silico evidence but is not a substitute for bench validation. The large, polar nature of rutin, carbetocin, and acarbose predicts limited cellular uptake, and peptide ligands may succumb to proteolysis. Conversely, kinase inhibitors carry an inherent risk of host-signaling interference. Subsequent work should therefore confirm antiviral potency in primary human airway cultures, employ alanine-scanning mutagenesis to verify binding hotspots; screen rutin and peptide analogues with enhanced permeability (e.g., prodrug or stapled peptide formats); and evaluate kinase inhibitor combinations that retain antiviral potency while minimizing off-target effects.

Conclusion

This comprehensive structure-guided repurposing study identifies two distinct chemotypes with therapeutic promise against HMPV. Polyphenolic rutin and the uterotonic peptide carbetocin stabilise the fusion glycoprotein in a prefusion-locked conformation, whereas EGFR/HER2-targeted small-molecule inhibitors lapatinib and mobocertinib show nanosecond-scale stability and high binding affinity for the viral nucleoprotein. While ADMET liabilities remain for the macromolecular ligands and safety monitoring is necessary for kinase inhibitors, the FDA approval status of all four compounds offers an expedited route to preclinical antiviral testing. The combined docking, MM/GBSA, MD, PCA, and DCCM workflow presented here provides a robust framework for prioritizing candidates and rationally directing medicinal-chemistry optimization toward first-in-class HMPV therapeutics.

Conflicts of interest

There are no conflicts of interest.

Funding Statement

Nil.