Introduction

DENV is a virus that belongs to the Flavivirus genus and family of flaviviruses. It has five serotypes: DEN-1, DEN-2, DEN-3, DEN-4 and DEN-5 share similar natural history1. A DEN-1, DEN-2, DEN-3, DEN-4 and DEN-5 share similar natural history1. It is estimated that around 3.6 billion people are at risk. 390 million infections have been reported, of which 96 million show symptoms. Of the total 96 million cases with symptoms, about 2 million lead to the severe diseases known as DHF and DSS2. DENV is responsible for a variety of clinical symptoms ranging from moderate fever, known as dengue fever, to severe and potentially life-threatening conditions leading to a disease named dengue hemorrhagic fever or dengue shock syndrome3. After the virus has entered the human body, the incubation phase lasts between 3 and 14 days.

At the onset of a DENV infection, symptoms such as headache, fever, muscle and joint pain, vomiting and skin rashes may occur. In more severe cases, life-threatening symptoms are often.

present, including thrombocytopenia (platelet count below 150,000 per microliter of blood), bleeding, low blood pressure, liver enlargement, high fever and edema or plasma leakage4. Despite the significant global health burden of DENV, specific antiviral therapies remain elusive and current treatments are primarily supportive. This underscores a critical research gap: the need for novel therapeutic agents capable of inhibiting DENV replication or mitigating its pathologic effects, such as thrombocytopenia. Natural products with their diverse chemical structures and biological activities offer a promising avenue for the discovery of such agents. Extracts from the papaya leaves (Carica papaya), for example, have been traditionally used and anecdotally reported to alleviate the symptoms associated with dengue, particularly thrombocytopenia.

The amino acid composition of the four DENV serotypes can vary by 25–40%. In addition, these serotypes can be further subdivided into genotypes, adding a reported further 3% variation to its molecular structure5. DEN-5, also known as sylvatic dengue virus, is a distinct serotype that belongs to various phylogenetic groupings. It is exclusively found in wildlife6. DENV is a diminutive, spherical virus measuring around 50 nanometers in diameter. The genome of the virus is composed of a positive-sense, single-stranded RNA with a length of around 10–11 kilobases. It also contains a single glycoprotein that can be further cleaved into three distinct structural proteins: capsid, membrane, and envelope. Furthermore, there are seven distinct non-structural (NS) proteins, namely NS1, NS2a, NS2b, NS3, NS4a, NS4b, and NS5. These NS proteins play a vital role in viral assembly and replication inside the host cell7.

DENV can infect many types of host cells, such as dendritic cells, macrophages, B cells and mast cells, with varying degrees of viral replication. In Chinese medicine, the term “bone-breaking fever” has been widely used to refer to a significant symptom observed in individuals infected with DENV. Recently, it has been discovered that DENV specifically targets osteoclasts, resulting in the production of cytokines at levels similar to DENV-infected macrophages8. Dendritic cells are the prime targets among the target cells and play a crucial role in initiating anti-DENV immune response, particularly in the early stages of infection9.

NS1 is crucial in viral replication, coagulopathy, vascular leakage, thrombocytopenia, and evasion of the immune system, as well as in the development of disease, and has been studied as a possible therapeutic target for the treatment of the disease10. NS1 is highly similar and conserved across different organisms and has a size of 45–56 kilodaltons and is made of 352 amino acids11. The DENV NS1 protein in the bloodstream binds directly to platelets through the Toll-like receptor 4 (TLR4) signaling pathway. This binding triggers the release of Adenosine diphosphate (ADP), which causes an increase in phosphatidylserine (PS) exposure and the expression of P-selectin on the platelet surface. These changes further activate the platelets and enhance their ability to aggregate. In addition, platelets that are triggered by NS1 are more prone to adhering to endothelium or being engulfed by macrophages. NS1 interacts with endothelial cells and macrophages, inducing their activation and stimulating the synthesis of cytokines. The thrombocytopenia and bleeding observed during a DENV infection can be related to the effects generated by NS112.

Medicinal plants, especially those possessing antiviral or other desirable medicinal properties, have been widely used through history for the treatment and prevention of many diseases13,14,15. The ethnomedicinal plants Carica papaya, Protium spruceanum, Mikania micrantha, Isatis tinctoria, and Clerodendrum viscosum are examples of plants commonly used for the treatment of DENV infections16. Carica papaya, popularly known as papaya or pawpaw, is a diploid species that belongs to the Caricaceae family. It is a polygamous and dicotyledonous plant and is widely recognized as one of the most economically important and well-known plants globally. Papaya contains a high concentration of phytochemicals, highly nutritional compounds that offer multiple health benefits and potential healing properties. The root, stem bark, latex, leaves, flowers, fruits and seeds from papaya have all been traditionally used for the therapeutic management of several ailments through history17,18.

Flavonoids have emerged as promising antiviral agents against dengue, as they are able to act directly on key viral components, particularly non-structural proteins essential for viral replication and assembly. Among these compounds, quercetin and kaempferol — abundant components of Carica papaya leaf extracts— - show significant inhibitory effects against NS2B-NS3 protease and NS5 polymerase, enzymes fundamental to the dengue virus replication process19,20,21. Consequently, the ability of these flavonoids to inhibit key enzymatic pathways underscores their central role as antiviral agents, disrupting essential stages of viral replication. In addition to their potent antiviral activity, these compounds also exhibit a favorable drug-likeness profile, highlighting their potential as candidates for further therapeutic development18. Similarly, apigenin, luteolin and olean-12-ene, also identified in papaya leaves, have garnered attention owing to their robust bioactivity and ability to modulate critical signaling pathways associated with viral infections. Remarkably, quercetin has the additional ability to reduce viral DENV RNA load in the host cell22,23, further highlighting its versatile antiviral potential. Thus, the collective antiviral mechanisms of these flavonoids contribute significantly to the documented therapeutic benefits of papaya leaf extracts, which are traditionally used to alleviate dengue-related symptoms, including thrombocytopenia and inflammation. Considering the established safety profile, wide availability and cost-effectiveness of papaya leaf extracts, it is scientifically compelling to investigate olean-12-ene, quercetin, apigenin, luteolin and kaempferol as promising candidates for dengue drug development24,25. Their high binding affinity, proven inhibitory activity against critical viral proteins and historical therapeutic use justify their selection for extensive in silico and in vitro studies to develop new therapeutics against dengue.

The crystallographic structure from the dengue virus 1 NS2B/NS3 protease (PDB ID: 3l6p), the dengue type 2 NS1 protein (PDB ID: 4o6b) and the precursor membrane envelope protein (prM-E) (PDB ID: 3c5x) were extracted from Protein Data Bank (PDB), and selected for their central role in viral replication, assembly and pathogenesis. The NS2B/NS3 protease is essential for viral polyprotein processing, an essential step in viral maturation that makes it a prime drug target26,27. Meanwhile, NS1 is not only an essential part of the virus life cycle but also associated with immune evasion and endothelial barrier dysfunction27,28. Inhibition of NS1 could help to prevent serious complications such as vascular leakage. Finally, the heterodimer prM-E plays a central role in the entry of the virus into the bloodstream by mediating membrane fusion, which is another strategic target29,30. Targeting these three different proteins enables comprehensive research into several phases of the dengue virus life cycle.

The primary objective of this in silico study was to investigate the therapeutic potential of five natural compounds from Carica papaya — olean-12-ene, quercetin, apigenin, luteolin and kaempferol — as inhibitors of key DENV proteins, thus elucidating their possible role in attenuating DENV infection, including associated thrombocytopenia. This was achieved through a comprehensive computational approach that included density functional theory (DFT) for geometry optimization, molecular electrostatic potential (MEP) analysis and quantum chemical descriptor calculation, complemented by pharmacokinetic profiling of each compound’s absorption, distribution, metabolism, excretion and toxicity (ADMET) properties of the DENV NS2B-NS3 protease, NS1 protein and precursor membrane envelope protein (prM-E).

Computational method

Quantum calculations of the compounds

Optimization of the geometry

The structures of olean-12-ene, quercetin, apigenin, luteolin and kaempferol (Fig. 1) were first retrieved from the PubChem database31 (https://pubchem.ncbi.nlm.nih.gov/) and validated to ensure accurate representation. Subsequently, these ligands were modified with functional groups deemed relevant for improved bioactivity.

Fig. 1
figure 1

Selected ligands extracted from Carica papaya.

Full size image

The BIOVIA Discovery Studio 2023 v.22 software was deployed on a dedicated server at the Universidade Federal do Rio Grande do Norte (UFRN), configured with 24 processors and 64 GB RAM to optimize computing power. The environment was carefully prepared to comply with the latest BIOVIA standards and to avoid software version conflicts. The ligands were pre-processed using the ‘Clean Geometry’ tool and further refined using the ‘Minimize Ligands’ module under the CHARMm force field. The initial geometry optimization was performed with a combination of ‘Quick Minimization’ and comprehensive ‘Full Minimization’ procedures, using the ‘Smart Minimizer’ algorithm, a 1000 steps of steepest descent method. Up to 500 conformers per ligand with an RMSD threshold of 0.5 Å were generated, with the most energetically favorable conformations selected for advanced quantum chemical analysis.

The selected conformers underwent DFT optimization at the B3LYP/6-31G(d, p) level to obtain precise geometrical arrangements. These calculations were further supported by implicit solvation models simulating biological environments to approximate physiological conditions32,33. The main objective of this step was to acquire compounds with optimized geometric configurations, essential for the energetic and quantum chemical characterization of ligand-receptor complexes using molecular docking and molecular dynamics simulations34,35,36.

Molecular electrostatic potential (MEP)

To analyze the charge distribution, Molecular Electrostatic Potential (MEP) mapping was conducted, which represents the potentials of nuclei and electrons within the molecular space37,38. This process facilitated the visualization of the charge distribution, which made it possible to explore charge-dependent molecular properties. The electrostatic potential, V(r), generated by nuclei and electrons as a static charge distribution, was crucial for predicting the reactivity of olean-12-ene, quercetin, apigenin, luteolin and kaempferol and provided insights into their interactions and reactivity behavior39OBJ.

Quantum chemical descriptors

Using DFT as the basic computational tool, detailed quantum mechanical modeling was conducted to extract relevant data on electron behavior for the ligands olean-12-ene, quercetin, apigenin, luteolin and kaempferol. The analysis provided critical insights into quantum descriptors such as the energies of the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO), the HOMO-LUMO gap (GAP), the ionization potential (I), the electron affinity (A), the total hardness (ɳ), the softness (σ), the chemical potential (µ), the global electronegativity (χ) and the global electrophilicity index (ω)40,41.These parameters were mathematically defined and evaluated according to established formulas42:

$$\begin{gathered} {\text{GAP }} = {\text{ }}\varepsilon {\text{HOMO }} - {\text{ }}\varepsilon {\text{LUMO}} \hfill \\ {\text{I }} \approx {\text{ }} - {\text{ }}\varepsilon {\text{HOMO}} \hfill \\ {\text{A }} \approx {\text{ }} - {\text{ }}\varepsilon {\text{LUMO}} \hfill \\ \eta \approx {\text{ }}\raise.5ex\hbox{$\scriptstyle 1$}\kern-.1em/ \kern-.15em\lower.25ex\hbox{$\scriptstyle 2$} {\text{ }}(\varepsilon {\text{LUMO }} - \varepsilon {\text{HOMO}}){\text{ }} \approx {\text{ }}\raise.5ex\hbox{$\scriptstyle 1$}\kern-.1em/ \kern-.15em\lower.25ex\hbox{$\scriptstyle 2$} {\text{ }}\left( {{\text{I }} - {\text{ A}}} \right) \hfill \\ \sigma {\text{ }} = {\text{ 1}}/\eta \hfill \\ \mu {\text{ }} \approx {\text{ }} - {\text{ }}\raise.5ex\hbox{$\scriptstyle 1$}\kern-.1em/ \kern-.15em\lower.25ex\hbox{$\scriptstyle 2$} {\text{ }}(\varepsilon {\text{HOMO }} + \varepsilon {\text{LUMO}}){\text{ }} \approx {\text{ }} - {\text{ }}\raise.5ex\hbox{$\scriptstyle 1$}\kern-.1em/ \kern-.15em\lower.25ex\hbox{$\scriptstyle 2$} {\text{ }}\left( {{\text{I }} + {\text{ A}}} \right) \hfill \\ \chi {\text{ }} \approx {\text{ }} - {\text{ }}\mu {\text{ }} \approx {\text{ }}\left( {{\text{I }} + {\text{ A}}} \right)/{\text{2}} \hfill \\ \omega {\text{ }} \approx {\text{ }}\chi {\text{2}}/{\text{2}}\eta \hfill \\ \end{gathered}$$

These electronic parameters confirm the geometric stability of the poses, reveal electrophilic and nucleophilic hotspots that drive recognition by the DENV NS2B-NS3 catalytic protease dyad, and enable correlations between chemical reactivity, docking affinity, and ADMET predictions, strengthening the structure–activity relationships proposed in this study.

Drug-likeness and pharmacokinetics predictions of the compounds

Drug-likeness parameters were calculated using the webserver FAF-Drugs4.143. All five candidates extracted from Carica papaya — quercetin, kaempferol, apigenin, luteolin and olean-12-ene — were evaluated for properties such as Lipinski’s rule of five, Veber’s rule, Egan’s rule, solubility prediction index, hetero atoms/carbon (Het/carbon or H/C) ratio and total charge to better discern and identify the ligands with favorable oral bioavailability.

Subsequently, the study was extended to a detailed assessment using ADMETlab 3.0, including total polar surface area (tPSA), formal charge and quantitative estimation of drug similarity (QED), among other related metrics. In addition, the ALARM NMR (A La Assay to detect Reactive Molecules by Nuclear Magnetic Resonance) property was analyzed to predict potential metabolic liabilities. These analyzes provided a nuanced understanding of the behavior of the ligands in biological systems and supported the prediction of their efficacy and safety profiles44.

By combining these tools, a synergistic evaluation of the ligands was achieved in the study. FAF-Drugs4.1 provided a robust screening platform, while ADMETlab 3.0 extended the depth of analysis and enabled the exploration of more sophisticated interactions and properties. This integrated approach highlighted the multidimensional nature of ligand profiles and provided methodological rigor and comprehensive insights into their potential applications.

Finally, all results were critically examined for consistency and correlations between physicochemical and pharmacokinetic parameters were explored to strengthen the reliability of the conclusions. The use of validated tools previously used in other studies40,45,46 emphasizes the robustness of the methodology and makes this study a rigorous and reliable exploration of these biologically important compounds.

Regarding the pharmacokinetic (PK) properties of potential therapeutic agents, particularly Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET), play an essential role in the preliminary stages of drug discovery and development47. For this study, pkCSM48 (ADMETlab 3.049, vNN-ADMET50, admetSAR 3.051, PRED-hERG37 and Adverpred52 web servers were used to perform comprehensive drug likeness screening and ADMET assessments for the selected ligands: Olean-12-ene, quercetin, apigenin, luteolin and kaempferol. These assessments were based on a series of calculated physicochemical parameters and ADMET-related descriptors.

The input data for each ligand was represented by the corresponding SMILES (Simplified Molecular Input Line Entry System) strings submitted to the prediction models on each server. This input method enables accurate calculation of pharmacokinetic profiles and ensures a comprehensive assessment of properties such as bioavailability, blood-brain barrier permeability, cytochrome P450 interactions and potential toxicities. The results of these predictions are very important as poor pharmacokinetic profiles remain a major cause of clinical trial failure53. Consequently, a comprehensive preclinical analysis of these PK characteristics and drug-likeness suitability is indispensable to bring candidates through the early phase of drug development.

Molecular docking

Molecular docking is an important step in analysing the binding properties of biological compounds, especially those of medical importance54,55. To prepare each receptor in an optimal conformation for docking analyzes and to minimize structural discrepancies resulting from experimental determination methods, three dengue virus protein targets — dengue virus 1 NS2B-NS3 protease (Fig. 2, PDB ID: 3L6P), Denv2-NS1 protease (Fig. 3, PDB ID: 4O6B) and the envelope protein (Fig. 4, PDB ID: 3C5X) — were obtained from the Protein Data Bank (PDB) (https://www.rcsb.org/). Prior to docking, these structures were pre-processed using AutoDockTools (ADT), a component of the MGLTools package (https://ccsb.scripps.edu/mgltools/downloads/). This step included the removal of water molecules, the assignment of Kollman charges and the addition of missing hydrogen atoms to polar residues. In parallel, the ligand structures — olean-12-ene, quercetin, apigenin, luteolin and kaempferol — were retrieved from PubChem (https://pubchem.ncbi.nlm.nih.gov/) and converted to PDB format to ensure compatibility with the subsequent docking steps. The putative binding pockets in each receptor were identified using the DockThor 2.0 web server (https://dockthor.lncc.br/v2/), which provided several potential cavities. For each protein, the pocket with the highest rank was designated as the predicted binding site and its coordinates were cross-validated with ADT. A cubic grid box with dimensions of 40 × 40 × 40 Å was centered on this site to define the search space for all subsequent docking calculations. Molecular docking was performed using HADDOCK2.4 (https://rascar.science.uu.nl/haddock2.4/). The protocol consisted of an initial rigid-body docking phase, followed by a semiflexible refinement phase and a final water refinement step. For each receptor–ligand system, 50 docking poses were generated. The resulting complexes were clustered based on the HADDOCK scoring function, which integrates van der Waals, electrostatic, and desolvation energy components in conjunction with root-mean-square deviation (RMSD) metrics. Upon completion of the docking simulations, the most representative clusters were selected for in-depth evaluation. In addition to extracting the HADDOCK scores and associated energetic parameters, the predicted free energies of binding (ΔG_prediction) and alternative binding energy measures (ΔG_score) were calculated. Detailed atom–atom contact analyzes were performed to elucidate the hydrophobic and polar interactions that stabilize each receptor–ligand complex. Finally, the representative complexes from the highest scoring clusters were further refined in PRODIGY (Protein Binding Energy Prediction), which provided additional insight into the Gibbs free energy (ΔG) and quantified the extent of interfacial atomic contacts.

Fig. 2
figure 2

Schematic representation of the DENV-1 NS2B-NS3 protease (left), with one of its binding sites highlighted as a red sphere and delimited for future docking analysis. The van der Waals interaction regions are shown in transparent yellow (left), together with the motif topology of the same protein (PDB ID: 3L6P) (right), with its distinct long loop regions with a single short alpha helix secondary structure.

Full size image
Fig. 3
figure 3

The structural model of the Denv2-NS1 protease (left) shows a red sphere marking an important binding site, and one of the binding sites selected for protein-ligand docking. The transparent yellow shading shows the van der Waals interaction regions, as well as the motif topology of the protein (PDB ID: 4O6B) (right). More helicoidal structures are present in this protease model, although the binding site was still found in a beta-sheet rich region.

Full size image
Fig. 4
figure 4

Illustration of the DENV envelope protein (left) highlighting one of the binding sites as a bright red sphere for docking purposes, the van der Waals interaction regions, and, on the right, the protein motifs in detail (PDB ID: 3C5X).

Full size image

Results

Quantum analysis

Geometry optimization

The geometries of five ligands - olean-12-ene, quercetin, apigenin, luteolin and kaempferol — were optimized and characterized. First, each ligand was optimized using a Dreiding-based forcefield, followed by minimization within a dielectric solvent environment using the CHARMm and SmartMinimizer methods.

For olean-12-ene, we identified 254 unique pharmacophore fingerprints with a single stable conformation, averaging 1.00 conformation per ligand. The radius of gyration was determined to be 4.61 Å, while the solvent-accessible surface area averaged 646.64 Ų. The RMSD values remained uniformly at 0 Å.

In comparison, quercetin exhibited 2170 unique pharmacophore fingerprints supported by 8 conformations, with an average of 8.00 conformations per ligand. This ligand exhibited an average radius of gyration of 4.11 Å and a solvent accessible surface area of 489.77 Ų. The pairwise RMSDs ranged from 0.31 to 1.48 Å, with a mean value of 1.03 Å.

Apigenin showed a total of 688 unique pharmacophore fingerprints across 3 conformations, averaging 3.00 conformations per ligand. It displayed an average radius of gyration of 4.05 Å, and the solvent accessible surface area was determined to be 467.68 Ų. The RMSD values ranged from 0.09 to 0.30 Å, with an average RMSD of 0.20 Å.

For luteolin, 1153 unique pharmacophore fingerprints were observed, distributed across 4 conformations, with an average of 4.00 conformations per ligand. The radius of gyration was calculated to be 4.13 Å, and the solvent accessible surface area averaged 481.62 Ų. The RMSD values ranged from 0.32 to 1.43 Å, with a mean RMSD of 1.05 Å.

Kaempferol, the final ligand, had 1616 unique pharmacophore fingerprints and 7 conformations, with an average of 7.00 conformations per ligand. The radius of gyration was 4.04 Å, and the solvent-accessible surface area measured 479.56 Ų. The pairwise RMSDs varied from 0.09 to 0.77 Å, with an average RMSD of 0.39 Å.

The most stable conformer of each ligand was further refined using quantum mechanical methods to support subsequent electronic characterization and docking studies.

Quantum energies

Quantum mechanical analyzes were performed on the ligands olean-12-ene, quercetin, apigenin, luteolin and kaempferol to characterize their energetic properties and electrostatic profiles. The total energy values determined for the same molecules were − 1245.92 Ha, −1169.08 Ha, −1010.55 Ha, −1089.81 Ha and − 1089.81 Ha, respectively, indicating their relative stability in the ground state. The binding energies refer to the amount of energy released when the molecule is formed from its individual atoms and the values of −90.11 Ha for olean-12-ene, −74.22 Ha for quercetin, −65.48 Ha for apigenin, −69.85 Ha for luteolin and − 69.85 Ha for kaempferol, with olean-12-ene having a significantly lower binding energy, indicating a particularly stable configuration.

The band gap energies, reflecting the electronic properties and potential chemical reactivity, were measured at 0.25 Ha, 0.13 Ha, 0.15 Ha, 0.15 Ha and 0.14 Ha for the ligands, respectively. The dipole moments, representing the charge distribution within each molecule, were 0.13 for olean-12-ene, 2.81 for quercetin, 2.81 for apigenin, 2.48 for luteolin and 2.44 for kaempferol. The higher dipole moments observed for quercetin and apigenin indicate significant charge separation, possibly increasing solubility and interaction capacity.

Cavity volumes, which provide information on the structural space available for intermolecular interactions, were 3366.90 ų for olean-12-ene, 1977.63 ų for quercetin, 1864.34 ų for apigenin, 1927.64 ų for luteolin and 1921.58 ų for kaempferol. The larger cavity volume of olean-12-ene could enable more extensive interactions with the target receptors. The dielectric and solvation energies, which represent energy differences in the transition from vacuum to solvent, were calculated to be −0.0023 Ha for olean-12-ene, −0.0663 Ha for quercetin, −0.0565 Ha for apigenin, −0.0694 Ha for luteolin and − 0.0516 Ha for kaempferol. These values show the stability of these ligands in different environments, which are important for solubility and biological functionality. The surface areas calculated for olean-12-ene, quercetin, apigenin, luteolin and kaempferol were 1426.69, 1016.32, 962.97, 1000.37 and 987.83, respectively, providing information about their interaction potential with the biological environment. The molecular electrostatic potential (MEP) maps (Fig. 5) illustrate the spatial distribution of charge, highlighting the electron donor sites, particularly at the hydroxyl groups, and the electron acceptor regions around the nitrogen atoms. The MEP surfaces provide a visual representation of the electrophilic (red/yellow) and nucleophilic (blue) regions and aid in the prediction of ligand interactions with protein targets.

Fig. 5
figure 5

3D representation of lead compound’s isosurfaces and respective quantum energy parameters (left) – total energy, binding energy, highest occupied molecular orbital (HOMO) energy, lowest unoccupied molecular orbital (LUMO) energy, band gap energy, dipole mag, cavity volume, dielectric energy, solvation energy, surface area — and their HOMO and LUMO surfaces (right).

Full size image

Quantum chemical descriptors

Olean-12-Ene exhibited the largest HOMO-LUMO gap of 6.912 eV, indicating that it has the highest stability and the lowest reactivity among the compounds. This large gap implies a significant energy requirement for electronic excitation and reduces the likelihood that it will participate in chemical reactions. The ionization potential (I) of olean-12-ene was calculated to be 5.865 eV, while the electron affinity (A) was − 1.047 eV, resulting in a chemical hardness (η) of 3.456 eV and a softness (σ) of 0.289 eV⁻¹. Its chemical potential (µ) was 2.409 eV, the electronegativity (χ) was determined to be −2.409 eV and the electrophilicity index (ω) was measured to be 10.030 eV, reflecting a moderate potential for stable interactions.

Quercetin exhibited a moderate balance between reactivity and stability with a HOMO-LUMO gap of 3.622 eV. Its ionization potential was 5.728 eV, and the electron affinity was 2.106 eV. The chemical hardness was determined to be 1.811 eV, while the softness was 0.552 eV⁻¹. Quercetin’s chemical potential reached 3.917 eV, its electronegativity (χ) was measured to be −3.917 eV, and it exhibited an electrophilicity index of 13.895 eV, suggesting that it can form relatively stable yet reactive interactions.

Apigenin presented a HOMO-LUMO gap of 4.218 eV, indicating slightly greater stability than quercetin. The ionization potential and electron affinity were measured to be 6.105 eV and 1.887 eV, respectively. The chemical hardness of apigenin was 2.109 eV and the softness was 0.474 eV⁻¹. The chemical potential was 3.996 eV, its electronegativity (χ) was calculated to be −3.996 eV, and the electrophilicity index reached 16.838 eV, the highest value among the compounds, indicating that apigenin has the greatest ability to interact with electrophilic sites.

Luteolin exhibited a HOMO-LUMO gap of 4.034 eV, suggesting a balance between reactivity and stability, although it is slightly less reactive than apigenin. The ionization potential was 5.956 eV, and the electron affinity was 1.922 eV, resulting in a chemical hardness of 2.017 eV and a softness of 0.496 eV⁻¹. The chemical potential was 3.939 eV, the electronegativity (χ) was determined to be −3.939 eV and the electrophilicity index was 15.647 eV, supporting its potential to form stable complexes.

Kaempferol exhibited a HOMO-LUMO gap of 3.754 eV, indicating a similar reactivity profile to that of quercetin but slightly higher stability. The ionization potential and electron affinity were calculated to be 5.799 eV and 2.045 eV, respectively, resulting in a chemical hardness of 1.877 eV and a softness of 0.533 eV⁻¹. Its chemical potential was 3.922 eV, its electronegativity (χ) was measured to be −3.922 eV and the electrophilicity index was 14.433 eV, indicating a moderate balance between reactivity and stability.

The results for these properties are summarized in Table 1. The quantum chemical (QC) descriptors analyzed — HOMO-LUMO gaps, ionization potentials, electron affinities, chemical hardness (η), softness (σ), chemical potentials (µ), electronegativities (χ), and electrophilicity indices (ω) — provide a comprehensive framework for understanding the stability and reactivity of the compounds olean-12-ene, quercetin, apigenin, luteolin, and kaempferol. Of these, olean-12-ene exhibits the highest stability, as evidenced by its large HOMO-LUMO gap (6.912 eV) and chemical hardness (3.456 eV). This reflects a low propensity for electronic transitions and high reactivity, consistent with the principle of maximum hardness56,57. This stability combined with its negative electron affinity (−1.047 eV) suggests minimal non-specific interactions, which could increase selectivity in therapeutic applications58. In contrast, quercetin with its smaller HOMO-LUMO gap (3.622 eV) exhibits intermediate reactivity, as indicated by its chemical hardness (1.811 eV), softness (0.552 eV⁻¹) and higher electron affinity (2.106 eV). These properties suggest that it can form reactive yet stable complexes with biological targets, making it well suited for non-covalent interactions such as hydrogen bonding in therapeutic contexts59.

Apigenin and luteolin show reactivity profiles intermediate between those of quercetin and olean-12-ene, with HOMO-LUMO gaps of 4.218 eV and 4.034 eV, respectively. Apigenin presented the highest electrophilicity index (16.838 eV), highlighting its ability to interact with nucleophilic sites, while the balanced profile of luteolin favors stable complex formation60,61. Kaempferol, with a HOMO-LUMO gap of 3.754 eV, mirrors the moderate reactivity of quercetin but is slightly more stable, as confirmed by its chemical hardness (1.877 eV) and electronegativity (−3.922 eV), which point towards interactions with electrophilic targets62. Overall, the stability of olean-12-ene could reduce off-target effects, while the higher reactivity of the other flavonoids could facilitate critical biochemical interactions, making them promising candidates for therapeutic applications against dengue63.

Table 1 Quantum chemical descriptors.
Full size table

Drug-likeness and pharmacokinetics

The evaluation of drug-likeness and pharmacokinetics at an early stage of drug development is crucial to ensure that preparations meet therapeutic requirements beyond purely structural criteria. This includes assessing absorption, distribution and excretion, ensuring sufficient concentrations at the target site and matching pharmacodynamics with pharmacokinetic profiles64. Such an integrated approach facilitates decision making and enables proactive strategies, including the development of prodrugs, to improve the suitability of candidates64.The results for all endpoints are shown in Fig. 6.

Fig. 6
figure 6

Physicochemical (left) and ADMET (right) properties for each compound. Blue denotes values that fit within acceptable ranges for the compounds’ medical applications, while red are values that fell short of their desirable cut-offs. For the CYP enzyme families, blue indicates interaction (inhibits or is a substrate) whereas red indicates non-interacting compound.

Full size image

Physicochemical

The physicochemical analysis of apigenin, kaempferol, luteolin, quercetin and olean-12-ene revealed different profiles. The flavonoids exhibited higher density and moderate hydrogen bonding capabilities, reflecting their balanced polar and non-polar properties. Olean-12-ene, on the other hand, exhibited low density, no hydrogen bonding ability and high lipophilicity (LogP ~ 7.7), which limited its water solubility and possibly reduced its oral bioavailability. Its rigid structure offered advantages in specificity but compromised adaptability.

Quercetin was characterized by its extensive hydrogen bonding (12 in total), high solubility (15,228.150 mg/L), favorable hydrophilicity (tPSA 131.360 Ų) and optimal balance between lipophilicity and water solubility. Apigenin, kaempferol and luteolin exhibited moderate physicochemical properties — reasonable hydrogen bonding, balanced LogP values (~ 2) and good solubility — suggesting suitability for oral administration and potential therapeutic efficacy65,66,67,68.

The pronounced hydrophobicity of olean-12-ene (LogD 5.319) and the low polarity (tPSA = 0) require formulation strategies such as SEDDS, liposomes or surfactants69 ALT®70:. In contrast, flavonoids better meet the drug-likeness criteria as they have moderate molecular weights (~ 270–302 Da) and ring systems that promote predictable pharmacokinetic behavior.

In summary, apigenin, kaempferol, luteolin and quercetin have a favorable profile for oral formulations and therapeutic development. Quercetin in particular is characterized by high solubility and strong hydrogen bonding potential. Although the lipophilicity of olean-12-ene could open up niche applications, it requires advanced delivery systems to overcome its low water solubility and achieve adequate bioavailability.

Medicinal chemistry

Medicinal chemistry evaluation shows that apigenin, kaempferol, luteolin, quercetin and olean-12-ene have varying degrees of reactivity, regulatory compliance and drug-likeness. ALARM NMR alerts — indicating potentially harmful thiol reactivity71 — were not issued for olean-12-ene, but were present for the flavonoids, with luteolin and quercetin being the most affected (3 alerts each). Only quercetin violated the BMS rule, suggesting possible structural liabilities, while increasing alerts under the chelation rule from apigenin (0) to quercetin (2) indicated increased metal binding potential relevant to antioxidant activity and metal ion regulation72,73.

All compounds met the Egan rule for oral bioavailability, but olean-12-ene did not meet the Golden Triangle, GSK and Lipinski rules due to high MW and logD values, which reduces its absorption prospects74. The flavonoids largely met these criteria and had minimal Lipinski violations and favorable NPscore values, reflecting their natural origin. Luteolin and quercetin elicited PAINS alerts, indicating possible nonspecific interactions, while apigenin, kaempferol, and olean-12-ene did not. With the exception of olean-12-ene, all met the Pfizer rule.

QED values ranged from 0.349 (olean-12-ene) to 0.632 (apigenin), indicating moderate drug-likeness and room for optimization75. The synthetic accessibility (SAscore) remained manageable despite the structural complexity. The TrafficLights PhysChem score showed risks mainly for olean-12-ene and to a lesser extent for apigenin and quercetin. All compounds fulfilled the Veber rule, underlining their potential for oral bioavailability.

Overall, apigenin, kaempferol and luteolin proved to be promising candidates with minimal violations. Quercetin, while offering remarkable properties of a natural product, requires careful optimization due to higher reactivity and PAINS warnings. Olean-12-ene may need structural modifications or advanced delivery strategies to overcome its poor drug-likeness profile.

ADMET

The ADMET profiles of apigenin, kaempferol, luteolin, quercetin and olean-12-ene generally indicate favorable oral absorption, although the main parameters vary. In Caco-2 assays, olean-12-ene showed remarkably high permeability (1.261 > 0.9), followed by apigenin (1.007), indicating a high potential for intestinal absorption76. In contrast, the permeability of quercetin was lower (−0.299). The PAMPA results showed high passive permeability for kaempferol (++), moderate for apigenin and quercetin (+) and low for luteolin and olean-12-ene (-). None of the compounds were identified as P-gp substrates, and only olean-12-ene inhibited P-gp (+++), possibly affecting concomitantly administered drugs.

All compounds achieved an absorption of ≥ 50% (+++), although the early thresholds (F20%, F30%) varied. Despite the low MDCK permeability, the predicted human intestinal absorption (HIA) remained high: olean-12-ene (97.56%), apigenin (93.25%), luteolin (81.13%), quercetin (77.21%) and kaempferol (74.29%)77. The good aqueous solubility of quercetin (logS − 2.93) supports oral bioavailability78 while the poor solubility of olean-12-ene (logS − 5.870) may require special formulations. All showed stability in human liver microsomes (t½ ≥30 min), suggesting a favorable metabolic profile79. In summary, these compounds generally show promising absorption and metabolic properties. The strong permeability and absorption of olean-12-ene contrasts with its low solubility and P-gp II inhibition, indicating the need for formulation strategies. The high solubility and balanced profile of quercetin make it an attractive candidate for oral administration.

Distribution properties have a major impact on drug efficacy and safety as they influence tissue penetration and plasma concentrations80,81,82. All five compounds — apigenin, kaempferol, luteolin, quercetin and olean-12-ene — exhibited high plasma protein binding (> 90%), limiting their free active fractions. Olean-12-ene had a higher volume of distribution and was permeable to the BBB (logBB = 0.87), indicating greater CNS exposure, while the flavonoids had more limited tissue distribution and negligible BBB penetration. Although the extensive tissue distribution of olean-12-ene may improve therapeutic targeting to tissues, it raises concerns about tissue-specific toxicity.

Transporter-mediated effects were also significant. All agents inhibited key hepatic uptake (OATP1B1/1B3) and efflux transporters (BCRP, MRP1) to varying degrees, which could alter the pharmacokinetics of concomitantly administered drugs and affect therapeutic outcomes. The strong BSEP inhibition of olean-12-ene (+++) could increase the risk of cholestasis, while kaempferol and apigenin moderately inhibited BSEP (++), and all compounds showed some ability to interfere with renal and hepatic transporters (OCT1/2), emphasizing the complexity of their distribution profiles.

Metabolically, the flavonoids strongly inhibited CYP3A4, CYP2C8 and CYP1A2, increasing the potential for drug-drug interactions83,84,85. Although olean-12-ene is not a CYP3A4 inhibitor, it served as a substrate for several CYP isoforms and exhibited distinct metabolic behavior. The UGT substrate activity of flavonoids suggests efficient glucuronidation pathways, whereas olean-12-ene has no UGT metabolism. Despite these interactions, kaempferol, quercetin and olean-12-ene exhibited high metabolic stability in human liver microsomes, suggesting longer half-lives and potentially more constant plasma levels.

In terms of excretion, all compounds exhibited moderate plasma clearance (~ 5.6–9.4 ml/min/kg) and short half-lives (~ 1–1.6 h), with the exception of olean-12-ene, which had an extremely short half-life (~ 0.09 h). Apigenin, kaempferol, luteolin and quercetin benefited from high renal and plasma clearance, which enabled efficient excretion. In contrast, the low total clearance of olean-12-ene with rapid elimination suggests that it relies on alternative routes of excretion.

Overall, these data point to complex distribution, metabolic and excretion profiles that may influence dosing strategies, combination therapies and safety considerations. The distinct tissue distribution and metabolic uniqueness of olean-12-ene, as well as the potent CYP inhibition and efficient excretion of flavonoids, underscore the importance of early, integrated ADMET assessment to optimize clinical outcomes.

Toxicity often proves to be a major obstacle in drug development and frequently stops candidates in preclinical phases86.Computational predictions for apigenin, kaempferol, luteolin, quercetin and olean-12-ene revealed a number of potential adverse effects. All were classified as non-blockers in binary hERG assessments, but apigenin and luteolin showed moderate cardiotoxicity in multiclass models, and olean-12-ene was classified as a potential hERG-II inhibitor87. The probability of drug-induced liver injury (DILI) was moderate for the flavonoids (0.703–0.796) and lower for olean-12-ene (0.362)88. The flavonoids also had a higher genotoxicity risk (0.975–0.987) compared to olean-12-ene (0.321), while olean-12-ene had a higher likelihood of carcinogenicity (0.899) and nephrotoxicity (0.72)89. All compounds had a low probability of neurotoxicity and no predicted reproductive toxicity.

In terms of the environment, olean-12-ene posed a greater risk for aquatic toxicity and bioaccumulation, but was classified as biodegradable, unlike the non-biodegradable flavonoids90.Regulatory warnings were issued for skin sensitization and genotoxicity, particularly for apigenin. While the flavonoids overall tended towards hepatotoxicity and genotoxicity, olean-12-ene raised concerns for cardiotoxicity, nephrotoxicity and environmental hazards. These results emphasize the need for early toxicity screening and possible structural changes or formulation strategies in drug development.

Molecular docking

Molecular docking assays were performed to evaluate the interactions between five papaya-derived compounds — olean-12-ene, quercetin, apigenin, luteolin and kaempferol — and three major dengue virus protein targets: the NS2B-NS3 protease (PDB: 3L6P), the NS1 protein (PDB: 4O6B) and the envelope protein (PDB: 3C5X) (Fig. 10 and Supplementary Tables 1–6). The analysis revealed that the binding affinities and interaction patterns depend on both the structural features of the ligand and the physicochemical properties of the target protein, with hydrophobic and polar interactions synergistically stabilizing the complexes91.

Quercetin showed a favorable predicted ΔG score on HADDOCK (− 7.77 kcal/mol) against the NS2B-NS3 protease, driven by van der Waals contributions (− 34.45 kcal/mol) and negative electrostatic energy (− 59.46 kcal/mol), consistent with its capacity for hydrogen bonding and charge interactions. In contrast, olean-12-ene, which is characterized by the largest HOMO-LUMO gap (6.91 eV) and high chemical hardness (3.46 eV), formed complexes based on steric complementarity and hydrophobic interactions, as shown by its binding to NS1 (desolvation energy: −13.08 kcal/mol). While the NS2B-NS3 protease favored polar ligands such as quercetin and apigenin, the envelope protein showed exceptionally strong binding across all compounds (ΔG prediction scores: −13.50 to − 15.59 kcal/mol), likely due to charged residues in its binding pocket that facilitate electrostatic stabilization92. These results can be seen in Supplementary Tables 1–3 and Fig. 10. Quantum chemical descriptors further elucidated these trends. Flavonoids with smaller HOMO-LUMO gaps (3.62–4.22 eV) and low chemical hardness showed greater electronic reactivity, favoring reversible interactions via hydrogen bonding and dipole-dipole attractions93. Apigenin, with the highest electrophilicity index (16.84 eV), showed enhanced interactions with electrophilic regions, consistent with its moderate docking results (e.g. −7.69 kcal/mol for NS2B-NS3 protease). Conversely, the electronic stability of olean-12-ene correlated with its preference for hydrophobic environments, as shown by its van der Waals dominance in NS1 (− 29.64 kcal/mol) and envelope protein complexes (− 71.03 kcal/mol). These observed binding modes and energetic contributions are consistent with the current understanding of flavonoid-protein interactions, where a combination of hydrogen bonds, often involving hydroxyl groups of the flavonoids and polar residues in the protein’s active site, and hydrophobic interactions with aromatic or aliphatic residues, collectively contribute to binding affinity and specificity. The ability of in silico docking studies and other computational methods to identify these key interactions and predict favorable binding poses for natural compounds, such as the flavonoids investigated here, is a widely recognized approach in the early stages of drug discovery, guiding further experimental validation14,15,94. Our detailed analysis of specific interactions with DENV NS2B-NS3 protease, NS1, and envelope protein aligns with this paradigm, pinpointing potential pharmacophoric features crucial for antiviral activity.

Despite the favorable docking values of olean-12-ene, its high lipophilicity (LogP ~ 7.7), poor solubility (logS ≈ − 5.87) and P-glycoprotein inhibition pose challenges for in vivo applicability95. In contrast, flavonoids such as quercetin and apigenin showed balanced ADMET profiles, including moderate permeability, higher solubility (logS − 2.93 for quercetin) and compliance with drug-likeness criteria96. Although some flavonoids triggered PAINS alerts (e.g., luteolin and quercetin), their molecular reactivity was consistent with the target requirements because the hydrogen bond donors/acceptors of quercetin stabilized the interactions in the active sites of protease and envelope protein (e.g., 1094 CO contacts for NS2B-NS3 protease).

A closer examination of the binding modes provided additional insight into how each compound interacts with the NS2B-NS3 protease, NS1 and the envelope protein. In the NS2B-NS3 protease, olean-12-ene binds mainly to the hydrophobic residues TRP17, ILE30, LEU31, ILE42 and ARG192, highlighting its affinity for non-polar microenvironments (Fig. 7A and B). Rasool et al.97 reported comparable docking results using different ligands targeting the same receptor. Quercetin made contacts with VAL204, GLY203, TYR200, PRO182, SER185, THR184 and GLY183, reflecting its broader ability to form both van der Waals and hydrogen bonds (Fig. 7C and D). Notably, one of the three contacts of quercetin with SER185 was an unfavorable interaction, indicating a local electrostatic or steric conflict, although the overall docking result was favorable. Apigenin, another polar flavonoid, interacted with TYR211, ASP179, GLY203, ASN202, HIS101, SER185, THR184 and GLY183, again indicating polar contacts with several amino acids (Fig. 7E and F). Similar to quercetin, the interaction of apigenin with SER185 also involved unfavorable binding among three contacts, demonstrating that even ligands with high affinity can have localized destabilizing interactions. Luteolin bound ALA218, LYS124, VAL173, ILE215, ALA216, ALA214, and ASN202 (Fig. 7G and H), whereas kaempferol bound to LEU178, TYR200, ASN202, HIS101, SER185, GLY183, and THR184 (Fig. 7I and J). These binding profiles highlight how different chemical scaffolds utilize different residue combinations within the NS2B-NS3 protease, either through hydrophobic channels, hydrogen bond donors/acceptors, or electrostatic complementarity.

Fig. 7
figure 7

Molecular interactions of the ligands olean-12-ene (orange), quercetin (yellow), apigenin (cyan), luteolin (purple) and kaempferol (magenta) with the NS2B-NS3 protease (red). Hydrophobic π-alkyl interactions are shown in light purple, T-shaped π–π interactions in medium purple and π–σ interactions in dark purple. Conventional hydrogen bonds are shown in green, while carbon-hydrogen bonds are shown in white. Unfavorable interactions are indicated in red.

Full size image

In the case of Denv2-NS1 protease, olean-12-ene again showed a strong propensity for hydrophobic or partially polar residues — ARG192, VAL194, LYS189, ALA187, ILE21, PHE20, VAL5 and LYS14 — confirming previous findings of its preference for such pockets (Fig. 8A and B). Quercetin formed important contacts with ARG294, ARG314, SER315, GLY259 and HIS261, indicating potential charge interactions in addition to the typical van der Waals forces (Fig. 8C and D). Apigenin targeted GLU12, LYS14, VAL5 and GLY3, although the only contact with GLU12 proved to be an exclusively unfavorable interaction (Fig. 8E and F). This is particularly relevant as it could reduce the overall stability of the complex despite its otherwise moderate docking performance. Luteolin interacted with GLU12, LYS14, VAL5 and ILE19, suggesting a similar combination of polar and non-polar sites (Fig. 8G and H). Kaempferol showed contacts with HIS261, GLY259, SER315, ARG314, ARG294 and GLU334 (Fig. 8I and J), highlighting its ability to interact with residues that can offer hydrogen bonding or ionic interactions. These patterns indicate that NS1, which is known to be strongly involved in immune evasion, has a mixture of hydrophobic and polar interactions, with the hydrophobic region particularly favoring non-polar ligands such as olean-12-ene. The selection of this therapeutic target is of considerable clinical relevance as it has been the focus of others in silico and in vitro studies98,99.

Fig. 8
figure 8

Interaction profile of the ligands olean-12-ene (orange), quercetin (yellow), apigenin (cyan), luteolin (purple) and kaempferol (magenta) with the DENV2-NS1 protease (blue). Hydrophobic π-alkyl interactions are represented in light purple, T-shaped π–π interactions in medium purple and π–σ interactions in dark purple. Conventional hydrogen bonds are shown in green, while carbon-hydrogen bonds appear in white. Unfavorable interactions are marked in red, while electrostatic interactions are highlighted in orange.

Full size image

As for the envelope protein, which is critical for viral entry, the docking data showed robust complex formation with all five ligands, consistent with a strong predicted ΔG of approximately − 13.50 to − 15.59 kcal/mol. Olean-12-ene bound to LYS334, PRO356, LEU351, ARG350, and VAL354, indicating a predominantly nonpolar and cationic environment that stabilizes hydrophobic interactions (Fig. 9A and B). Quercetin interacted with GLN248, ASP249, THR68, ASP47, LEU65, LYS64 and VAL252, consistent with its known ability to form hydrogen bonds and electrostatic complementarity (Fig. 9C and D). Apigenin showed interactions with VAL251, ASP47, LEU65, THR79 and VAL250, again illustrating its mixture of polar and hydrophobic contacts (Fig. 9E and F). Luteolin formed two interactions with THR50, one of which was unfavorable, in addition to binding with THR48 and THR69. This single repulsive contact emphasizes the dynamic nature of ligand-receptor complementarity, even with ligands that otherwise exhibit good affinity (Fig. 9G and H). Finally, kaempferol associated with LEU351, ARG350, LYS36 and ASN37 due to a combination of hydrophobic pockets (LEU351) and electrostatic interactions (ARG350, LYS36) (Fig. 9I and J). These extensive binding traces confirm the idea that the dengue envelope protein contains charged residues that favor both electrostatic stabilization and van der Waals contacts. Other studies are investigating different strategies to inhibit the dengue virus envelope protein (DENV) and highlight its importance as a target for further research100.

Fig. 9
figure 9

Molecular interaction analysis of the ligands olean-12-ene (orange), quercetin (yellow), apigenin (cyan), luteolin (purple) and kaempferol (magenta) with the envelope protein (green). Hydrophobic π-alkyl interactions are illustrated in light purple, T-shaped π–π interactions in medium purple and π–σ interactions in dark purple. Conventional hydrogen bonds are indicated in green, while carbon-hydrogen bonds are shown in white. Unfavorable interactions are shown in red, and electrostatic interactions are highlighted in orange.

Full size image

The NS2B-NS3 protease, which is crucial for the processing of the viral polyprotein, therefore showed a particularly strong affinity for flavonoids, suggesting a possible inhibition of replication101.In this context, the unfavorable contacts observed for quercetin and apigenin on SER185 do not negate their overall potential, but highlight the need for structural optimization if these molecules are to be further developed. NS1, which is involved in immune evasion, was strongly modulated by hydrophobic ligands such as olean-12-ene, although the single repulsive interaction of apigenin with GLU12 suggests that despite moderate affinity, precise spatial or electronic compatibility may need to be fine-tuned. The envelope protein, which is essential for viral entry, showed broad ligand affinity, highlighting its susceptibility to a range of interactions, from purely hydrophobic to highly polar. The partial repulsion of luteolin at THR50 suggests that this protein may also be sterically or electrostatically constrained. Overall, these results argue for a multitarget inhibition strategy using papaya-derived compounds — particularly flavonoids — to potentially disrupt multiple stages of the dengue virus life cycle, including replication, immune evasion and cell entry102.

While this in silico integration of docking, quantum chemistry and ADMET provides mechanistic insights, experimental validation is essential to account for constraints such as protein flexibility, solvent effects and ligand restructuring. Future studies should employ techniques such as surface plasmon resonance or X-ray crystallography to confirm binding affinities and substantiate interactions between key residues, especially when unfavorable contacts have been identified. In addition, optimizations of the structure–activity relationship (SAR) could improve pharmacokinetics, particularly for olean-12-ene, by increasing solubility and mitigating potential toxicity risks103. Molecular dynamic simulations could also prove useful to better illustrate the binding energies and find the most ideal conformation for each protein-ligand complex, and could prove helpful for future in silico studies by providing a more energetically favorable complex. Likewise, other experimental methods such as enzymatic inhibition experiments using recombinant NS2B-NS3 protease to confirm binding and quantify IC50 levels, cell-based antiviral experiments in DENV-infected Huh-7 or Vero cells to assess reductions in viral replication and markers associated with thrombocytopenia, and an assessment in AG129 mouse models of dengue to monitor platelet counts, viremia, and toxicity profiles could be employed to further validate our in silico findings with experimental data.

For flavonoids, adjusting the positioning or number of hydroxyl groups could reduce the occurrence of repulsive interactions (e.g. with SER185 or GLU12) while maintaining the favorable hydrogen bonds required for stable complex formation. (Fig. 10).

Fig. 10
figure 10

HADDOCK molecular docking analysis and the calculated ΔG predicted for each protein-ligand complex, highlighting the greater binding energies with the envelope protein (green). The respective values for each predicted ΔG and other docking results can be seen in Supplementary Tables 1–3.

Full size image

Quercetin stands out as the main candidate as it combines one of the most favorable binding to the catalytic protease NS2B-NS3 (−7.77 kcal/mol) and envelope protein (−14.1 kcal/mol) with good water solubility (logS − 2.93) and favorable ADMET indices. Kaempferol also showed a good pharmacokinetic profile with equally strong binding energies to the target proteins, with a good coupled binding energy of −7.89 kcal/mol when complexed with the protease NS2B-NS3, making it another potential antiviral compound.

Conclusion

This in silico study showed the therapeutic potential of five compounds extracted from papaya against DENV. The flavonoids, particularly quercetin, apigenin, kaempferol and luteolin, exhibited favorable ADMET profiles, suggesting better drug-likeness and lower toxicity risk compared to the more lipophilic olean-12-ene. Quantum chemical analysis showed that olean-12-ene is the most stable compound (HOMO-LUMO gap: 6.912 eV), while flavonoids such as quercetin (HOMO-LUMO gap: 3.622 eV) and apigenin (electrophilicity index: 16.838 eV) have a balanced reactivity suitable for biochemical interactions. Molecular docking simulations revealed significant binding affinities of these compounds, especially quercetin and olean-12-ene, to key DENV targets such as NS2B-NS3 protease, NS1 protein, and the envelope protein, with quercetin showing particularly strong interactions with NS2B-NS3 protease (−7.77 kcal/mol) and envelope protein (−14.10 kcal/mol).

Collectively, these computational results highlight quercetin, kaempferol, apigenin and luteolin as promising candidates for further investigation. Olean-12-ene showed stability and some strong binding, but its ADMET profile suggests challenges. However, the identified flavonoids show a good balance between predicted efficacy and safety. To further substantiate these in silico results and understand the dynamic behavior of these ligand-protein interactions, future work should include molecular dynamics simulations, especially for the most promising flavonoid-target complexes. These MD studies would provide deeper insights into the stability of the complex, conformational changes and free energies of binding. Subsequently, it is essential to validate these candidates in vitro, including enzymatic inhibition assays (e.g. against recombinant DENV NS2B-NS3 protease to determine IC50 values) and cell-based antiviral assays using DENV-infected cell lines (e.g. Huh-7 or Vero cells) to assess the reductions of viral replication. Such experimental data are essential to validate these in silico predictions and to advance these natural products towards potential therapeutic anti-DENV development, particularly with regard to their role in mitigating DENV-associated thrombocytopenia.