In Silico Molecular Docking of Indonesian Medicinal Plants: Garcinia mangostana, Eurycoma longifolia, Brucea javanica and Momordica charantia as Neuraminidase Inhibitors Against H5N1

3.2.1 Docking coordinates

The center coordinates for constructing docking box parameters were determined using PyMOL, an open-source software tool for visualizing macromolecules [28]. As indicated in Table 1, the center coordinates were 3.605, 47.024, and 103.945. Utilizing these coordinates as the center, a docking box suitable for blind docking can be designed by setting its dimensions to encompass the entire receptor [20].

3.2.2 Positive control docking

To validate the docking setup and establish a benchmark for evaluating novel inhibitors, two clinically approved neuraminidase inhibitors, oseltamivir and peramivir, were selected as positive controls. These positive controls serve as a benchmark for evaluating the binding performance of the tested natural compound based on the predicted free binding energy [29].

The initial molecular docking experiment used oseltamivir carboxylate as a positive control. This is an active metabolite of oseltamivir, capable of binding and inhibiting influenza virus neuraminidases in vivo [30]. The results presented in Table 2 demonstrate that oseltamivir carboxylate exhibits a superior binding affinity (−6.45 kcal/mol) compared to the second positive control model, peramivir, which possesses a binding affinity of −6.093 kcal/mol. This is because a lower free energy indicates a more stable binding complex between the ligand and neuraminidase [Kurczab, 2017]. Thus, these findings indicate that oseltamivir carboxylate facilitates a more stable binding toward H5N1 influenza neuraminidase and functions as a more effective neuraminidase inhibitor compared to peramivir. Notably, this result aligns with the findings of Liu et al. [31], which reported a comparable binding energy result of approximately −6.91 kcal/mol for avian influenza neuraminidase.

Furthermore, Tao et al. [32] reported that oseltamivir carboxylate establishes a direct hydrogen bond with the amino acid side chain R118 (Arg118). However, MD simulations indicate that under physiological conditions, protein structure undergoes alterations, potentially disrupting the direct hydrogen bond and replacing it with a weaker water-mediated hydrogen bond. This suggests that while oseltamivir carboxylate exhibits the highest binding affinity compared to peramivir as a positive control, its binding strength may be reduced under physiological conditions. Additionally, it is crucial to consider natural mutations in the H5N1 neuraminidase gene that can further impact binding compatibility with oseltamivir carboxylate. A study by Yadav et al. [33] identified the I117V mutation in the H5N1 neuraminidase gene. This was followed up by Mhlongo and Soliman [34], who revealed that the I117V mutant complex formed fewer hydrogen bonds and did not interact with the Glu37 residue, resulting in structural instability, distortion of oseltamivir, and a diminished binding affinity.

3.2.3 Natural ligand docking

The results for the potential neuraminidase inhibitor are presented in Table 3. Gamma-mangostin from G. mangostana emerges as the most promising candidate due to its binding affinity of −9.713 kcal/mol, which is the lowest among all the results. However, the RMSD values of the two other selected models exceed 2 Å, indicating instability [35].

In comparison to the positive controls (oseltamivir and peramivir) presented in Table 2, Gamma-mangostin and Kuguacin J demonstrated greater predicted binding affinity and docking stability. These findings suggest that Kuguacin J and other plant-derived substances hold promise for further research and validation. Notably, the binding affinities of Kuguacin J surpass those of conventional antiviral controls such as oseltamivir and peramivir, which range from −5.953 to −6.45 kcal/mol. These results indicate that certain plant-derived substances may possess higher or equivalent inhibitory effects on neuraminidase compared to conventional medications.

In addition to Gamma-mangostin, several other compounds, as shown in Table 3, were identified as potential neuraminidase inhibitors. These include Alpha-mangostin and Garcinone C, both of which are xanthone derivatives found in the pericarp of G. mangostana. G. mangostana has been extensively studied for its pharmaceutical properties, including its antimicrobial activity [36].

The study found that Alpha-mangostin exhibited a binding affinity of −7.94 kcal/mol, suggesting a favorable interaction with the neuraminidase active site. However, its RMSD values exceeded 2 Å in models 2 and 3, indicating a lack of consistency in the predicted binding conformations. Garcinone C, on the other hand, demonstrated a slightly stronger binding affinity of −8.039 kcal/mol but exhibited less stable docking, as evidenced by the RMSD values of 3.711 Å for model 2 and 3.763 Å [35]. Thus, the presence of Alpha-mangostin, Garcinone C, and Gamma-mangostin within G. mangostana underscores its potential as a natural source of neuraminidase inhibitors.

Several other compounds sourced from B. javanica, as listed in Table 3, demonstrated binding compatibility with neuraminidase, including Bruceanol G, Bruceoside C, and Bruceantinol. These quassinoids are well-known constituents of B. javanica, a medicinal plant traditionally utilized in Southeast Asian herbal medicine. It is recognized for its pharmacological properties, particularly its anticancer, antimalarial, and anti-inflammatory effects [37–39]. The molecular docking analysis conducted in this study showed the highest binding affinities, ranging from −7.247 to −7.401 kcal/mol, indicating a favorable interaction with the neuraminidase active site. However, the RMSD values of all compounds originating from this plant, excluding model 2 of Bruceoside C, exceeded 2 Å, suggesting a lower consistency in the predicted binding conformations across docking models [35]. Bruceoside C in model 2 exhibited an RMSD value of 1.972 Å, indicating a more stable pose in that docking instance. Nevertheless, the overall docking behavior of Bruceanol G, Bruceoside C, and Bruceantinol underscores their potential as neuraminidase inhibitors.

Furthermore, Quercetin, a flavonoid also identified in B. javanica, exhibited a remarkable binding affinity of −8.557 kcal/mol, ranking among the more effective compounds in this study. In contrast to quassinoids, Quercetin is a flavonoid belonging to a class of polyphenolic compounds known for their antioxidant, anti-inflammatory, and antiviral properties, particularly against influenza viruses, as shown in numerous studies [40–42]. For instance, Mehrbod et al. [43] and Wu et al. [44] reported that Quercetin and its derivatives, such as isoQuercetin, inhibit influenza A virus replication by modulating viral entry and suppressing neuraminidase activity. This aligns with the docking results of the present study, wherein Quercetin exhibited a strong predicted interaction with the neuraminidase active site. However, its RMSD values across binding modes were significantly above 2 Å, indicating substantial variability in pose stability. Nevertheless, the docking findings, in conjunction with existing literature, support the therapeutic potential of Quercetin as a neuraminidase inhibitor. Kuguacin J and Momordicin I, both derived from M. charantia, also demonstrate notable binding affinity toward neuraminidase, with values of −7.35 kcal/mol and −7.661 kcal/mol, respectively.

M. charantia, commonly known as bitter melon, is a widely utilized medicinal plant in Asia for its antidiabetic, antiviral, and anti-inflammatory properties [45]. While both compounds exhibited relatively strong binding potential, their docking results indicated RMSD values exceeding 2 Å, suggesting variability and diminished stability in their predicted binding conformations. The compound from M. charantia, Momordicin I, showed RMSD values of 4.538 Å in model 2 and 21.135 Å in model 6, respectively, indicating a significant deviation from the reference pose. These findings indicate that while the compounds may possess affinity toward the target, the absence of binding mode consistency limits their reliability without modifications.

Eurycomalactone, a quassinoid compound isolated from E. longifolia, demonstrated a binding affinity of −7.088 kcal/mol in the current molecular docking analysis, suggesting a potentially favorable interaction with the neuraminidase active site. Eurycomalactone is widely recognized for its diverse pharmacological activities, including anti-inflammatory and anticancer effects [46, 47]. Although its direct antiviral action has received less extensive research, the findings indicate that eurycomalactone may possess neuraminidase-inhibitory potential. However, all RMSD values recorded for its docking models were above 2 Å, indicating reduced conformational stability in this study. These findings suggest that further testing is necessary to ascertain its efficacy against the influenza virus.

The in silico molecular docking approach is widely recognized for its efficiency in conducting preliminary screening of binding affinity and pose stability of natural ligands against the neuraminidase active site [48]. These findings hold biological significance, as neuraminidase is a pivotal enzyme in the influenza virus life cycle [49]. By inhibiting this enzyme, viral release and transmission can be stopped. Furthermore, natural ligands derived from plants exhibiting high binding affinities and docking patterns demonstrate greater efficacy as antiviral agents against H5N1. Therefore, a collaboration between natural product research and computer sciences will help identify more treatment options that are effective and sustainable.

3.2.4 RMSF results

RMSF measures the movement of a single amino acid during a MD simulation, thus quantifying flexibility at the residue level [50]. In order to be considered stable, ligand-protein complexes must have protein backbone RMSF values below 2.5 Å. According to this metric, the average RMSF values of all H5N1-ligand complexes shown in Table 4 (ranging from 0.713 to 0.830 Å) indicate that protein structures remained stable upon binding with selected natural ligands. Among these, the Bruceantinol-bound complex exhibited the lowest mean RMSF, suggesting relatively reduced overall flexibility. Meanwhile, Garconone C showed the highest mean RMSF, though still well within the acceptable stability threshold.

Despite the overall stability indicated by the average RMSF values, the maximum RMSF values for several complexes exceeded 5 Å, with the Alpha-mangostin – and Eurycomalactone-bound complexes displaying the highest peak fluctuations. These elevated maximum RMSF values suggest that certain regions within the protein undergo localized instability. However, such high RMSF values are more commonly associated with solvent-exposed loop regions and terminal residues that are more flexible. This behavior has been observed in multiple MD studies, where fluctuating loop regions correspond to higher RMSF peaks [51]. Therefore, the presence of higher-than-acceptable maximum RMSF values does not necessarily reflect the instability of the ligand-binding cores, but rather normal dynamic behavior of flexible protein regions during the simulation.

3.2.5 Adme-tox results

Water solubility is an important property for drugs because it allows them to spread evenly in the body and work effectively. It is especially critical for effective oral delivery. Good water solubility also provides drugs with higher bioavailability and better chemical stability [52]. Compounds with solubility values ≥0 are classified as highly soluble; those with values between 0 and −2 are considered soluble; values ranging from −2 to −4 indicate slight solubility; and compounds with solubility values below −4 are regarded as insoluble [53]. Therefore, according to the defined standard, most of the compounds are not soluble, as most of their values were below −4. Caco-2 permeability is commonly regarded as the “gold standard” in vitro test for determining how much of a medicine would actually enter the human body after swallowing [54]. High-permeability drugs have Papp values between 13 and 76.7 × 10⁻⁶ cm/s, low-permeability drugs have Papp values below 0.74 × 10⁻⁶ cm/s, and moderate-permeability drugs fall between 1.29 and 16.0 × 10⁻⁶ cm/s. Based on Table 5, all ligands had low permeability values, as all the values stood below 1.29. Ligands with low Caco2 permeability (log Papp in 10⁻⁶ cm/s) may suggest poor absorption in the gastrointestinal system, which is potentially caused by restricted passive diffusion or other absorption processes [55]. Even so, low Caco-2 permeability might also imply a paracellular diffusion mechanism rather than transcellular diffusion or carrier-mediated transport [56]. As transdermal drug delivery systems rely on a medication’s capacity to cross the skin barrier, the low values contraindicate the use of these compounds as skin-applied products. Compounds having a LogKp value greater than −2.5 cm/h have limited skin permeability [57]. Therefore, according to Table 5, all compounds had low skin permeability, as the values stood above the threshold. This suggests that the compounds had limited transdermal applications. As for the P-glycoprotein substrate, P-glycoprotein I inhibitor, and P-glycoprotein II inhibitor values, all ligands can be actively transported out of cells by P-gp. Meanwhile, most of the ligands can inhibit P-glycoprotein I, except for Quercetin and Eurycomalactone. On the other hand, half of the ligands, which include Gamma-mangostin, Garcinone C, Alpha-mangostin, Momordicin I, and Kuguacin J, can also inhibit P-glycoprotein II. As for HIA, it is a crucial pharmacokinetic mechanism that determines the bioavailability of a medication or molecule after absorption in the small intestine [58]. A high value of HIA, HIA ≥30%, is considered to be well absorbed in the human intestine [59]. Therefore, it can be concluded that all of the proposed ligands are highly absorbable by the human intestine, as all the HIA values were above 60%, with the majority of them having HIA values above 70%, while the highest HIA percentage was achieved by Momordicin I.

VDss (human) refers to the human steady-state volume of distribution, where low VDss implies that the medicine spends the majority of its time in the bloodstream, which is important for treatments that target blood vessels, and high VDss suggests extensive tissue binding/partitioning is frequently observed with lipophilic substances or medicines that bind to intracellular proteins or fat. Generally, various thresholds are used to characterize distribution quantities as low, moderate, or large. Values can range from 0.04 L/kg (all compounds in plasma) to hundreds of L/kg (most compounds in tissue). Thresholds for designating compounds as “moderate” or “high” volume vary in the literature [60]. Based on Table 6, all values of VDss were below 1, and most were below 0. The ligand with the highest VDss value was Kuguacin J (0.782). A Fu value below 0.05 was considered low, while a value above 0.2 was considered high [61]. Therefore, it can be concluded that the majority of the ligands had moderate to high Fu levels, while two ligands, Alpha-mangostin and Kuguacin J, had a value of 0. According to previous findings, compounds with blood–brain barrier permeability (BBB +) generally have log BB values exceeding 0, whereas BBB-impermeable compounds display log BB values less than −0.3 [62]. Therefore, according to Table 6, all ligands except Kuguacin J are not able to pass the BBB (blood-brain barrier). For CNS permeability, compounds with log PS values greater than −2 are generally considered capable of penetrating the central nervous system, indicating higher blood–brain barrier permeability, whereas compounds with log PS values below −3 show poor CNS penetration and low BBB permeability [63]. Some of the ligands (e.g., Gamma-mangostin, Alpha-mangostin, Momordicin I, and Kuguacin J) have log PS values over −2, suggesting greater potential for central nervous system penetration. Several substances have log PS values below −3 (e.g., Quercetin, Bruceanol G, Bruceoside C, Eurycomalactone, and Bruceantinol), indicating minimal CNS permeability and blood-brain barrier penetration. The log BB measures the brain vs. blood concentrations at equilibrium, mixing the effects of distribution and binding, while log PS reflects the intrinsic permeability rate across the BBB before equilibrium [64]. For the metabolism precision result (refer to Table 6), none of the ligands are substrates of CYP2D6; most of them are substrates of CYP3A4, except for Gamma-mangostin and Quercetin; and the majority of them are inhibitors of CYP1A2, except for Gamma-mangostin and Quercetin. Additionally, most of them are not inhibitors of CYP2C19 and CYP2C9 except for Gamma-mangostin and Alpha-mangostin. All the ligands are inhibitors of CYP2D6, and only some of them (Gamma-mangostin, Alpha-mangostin, and Momordicin I) are inhibitors of CYP3A4.

As for the excretion prediction results, Table 8 shows that none of the ligands acted as substrates to the renal OCT2. OCT2, or organic cation transporter 2, is an enzyme that functions in the initial step of cationic drug excretion [65]. Thus, drugs that are substrates to this enzyme are likely to be excreted through this process. Since none of the drugs act as an OCT2 substrate, they are not likely to be excreted through this process. As for how well the drugs will be cleared from the body, this property can be quantified with total clearance. Drugs with a total clearance over 5 mL/min/kg, or around 0.699 log mL/min/kg, are considered to have a high clearance [66]. Table 8 shows that none of the ligands reach this threshold, so they do not have high clearance. Additionally, drugs that do not act as OCT2 substrates but have high clearance are considered to have good bioavailability due to efficient metabolism and absorption [67]. All of the drugs pass the first property needed to be considered as having potentially good bioavailability, but none of them have what is considered high clearance.

Based on the data in Table 9, Quercetin and Eurycomalactone were found to exhibit AMES toxicity. Regarding organ safety, most compounds were predicted to be safe for the liver; however, Kuguacin J was identified as a potential hepatotoxin. Furthermore, none of the evaluated ligands or compounds were found to cause skin sensitization. The hERG gene encodes a potassium channel in the heart that regulates the heartbeat. Inhibition of these channels can cause long QT syndrome and potentially deadly cardiac arrhythmias [68]. From the results, almost all compounds were safe for the heart, as they did not inhibit both hERG I and hERG II, except for Gamma-mangostin, Garcinone C, and Eurycomalactone. CYP2C19 is a key enzyme in the liver that breaks down medicines. Blocking CYP2C19 inhibits the metabolism of other pharmaceuticals, which can lead to hazardous drug-drug interactions [69]. According to the results, only Gamma-mangostin and Garcinone C were known to inhibit CYP2C19.

3.2.6 Strengths and limitations of the study

The experimental findings demonstrated promising results in identifying ligands derived from indigenous plants in Indonesia as potent neuraminidase inhibitors. All proposed ligands exhibited greater binding affinity compared to positive controls, suggesting enhanced inhibition of H5N1 neuraminidase. These discoveries may advance the development of treatments for H5N1 infection, should it emerge as a global concern or pandemic in the future. The availability of suitable ligands sourced from locally occurring plants in Indonesia facilitates their extraction and manufacturing as drugs without concerns regarding resource scarcity.

The utilization of plants as drugs presents a renewable source of medicine with diverse therapeutic activities and structural characteristics. Furthermore, plant-derived compounds often exhibit enhanced stability, reduced side effects, and pleiotropic druggability compared to synthetic drugs [70].

Despite the high binding affinity of the ligands, several limitations arise from the application of blind docking. Blind docking encounters challenges in accuracy and efficiency due to the extensive search space involved. It may be less precise than targeted docking if the active site is well-defined and specific interactions are desired [38]. Furthermore, critical aspects affecting binding affinity predictions, including solvent effects and hydrogen bonding descriptions, were not incorporated in the experiment [48]. Lastly, the provided binding affinity and RMSD did not elucidate which molecules interacted with the receptor, and there is little knowledge regarding the precise molecular interactions between ligands and receptors.

3.2.7 Future suggestions

Given that all the ligands have been confirmed to bind with the protein through molecular docking and dynamics, density functional theory analyses can be conducted for a more precise prediction of their binding affinities. Density functional theory employs quantum mechanical calculations to model the electronic structures of compounds and provide realistic interactions between them [71]. However, in-silico methods solely provide predictions and, consequently, must undergo in vitro validation [29]. Therefore, with the knowledge that all of the tested natural compounds possess the capability of binding with H5N1 based on their binding affinities, future studies should validate these findings by analyzing their interactions with in vitro methods through the utilization of techniques such as isothermal titration calorimetry (ITC), high-throughput stability, mobility, or spectroscopic shift assays [17]. Conversely, the interaction of ligands with proteins can be experimentally determined by visualizing the structures of the components at an atomic level using methods like X-ray crystallography, nuclear magnetic resonance (NMR), Laue X-ray diffraction, small-angle X-ray scattering, and cryo-electron microscopy [28].

3.2.8 Future applications

The study revealed that herbal medicines, such as Gamma-mangostin and Kuguacin J, exhibit greater binding affinities and more stable docking postures compared to conventional antivirals like oseltamivir and peramivir. These molecules present novel avenues for drug development and discovery, with the potential for further validation through in vitro and in vivo research. Should these molecules prove effective, structural modifications can be undertaken to enhance their pharmacokinetic properties while mitigating potential toxicity. Furthermore, these compounds may be formulated into nasal sprays or inhalable devices for early-stage influenza intervention. As these compounds are derived from easily-accessible medicinal plants in Indonesia, they hold the potential to foster locally generated, cost-effective antiviral therapies, thereby improving medication accessibility in Indonesia. The integration of computational drug development with traditional ethnobotanical knowledge may lead to the identification of multi-target antiviral medications capable of effectively combating emerging influenza strains and bolstering global pandemic preparedness.