Molecular Dynamics Simulation of Ligands from Anredera cordifolia (Binahong) to the Main Protease (Mpro) of SARS-CoV-2

COVID-19 in Indonesia is considered to be entering the endemic phase, and the population is expected to live side by side with the SARS-CoV 2 viruses and their variants. In this study, procyanidin, oleic acid, methyl linoleic acid, and vitexin, four compounds from binahong leaves-tropical/subtropical plant, were examined for their interactions with the major protease (Mpro) of the SARS-CoV 2 virus. Molecular dynamics simulation shows that procyanidin and vitexin have the best docking scores of −9.132 and −8.433, respectively. Molecular dynamics simulation also shows that procyanidin and vitexin have the best Root Mean Square Displacement (RMSD) and Root Mean Square Fluctuation (RMSF) performance due to dominant hydrogen, hydrophobic, and water bridge interactions. However, further strain energy calculation obtained from ligand torsion analyses, procyanidin and vitexin do not conform as much as quercetin as ligand control even though these two ligands have good performance in terms of interaction with the target protein.


Introduction
Coronavirus disease  is an infectious disease caused by the SARS-CoV 2 virus. Tis disease was frst reported to have spread in Wuhan, China, on December 31st, 2019. Less than three months later, on March 9th, 2020, COVID-19 was declared a pandemic by the WHO (World Health Organization). Until now, in August 2021, the total number of cases of this disease reached 601 million around the world, resulting in the deaths of approximately 6.49 million victims. Indonesia is one of the countries most affected by COVID- 19. In Indonesia alone, 6.35 million cases have occurred since the pandemic started on February 2020, and it took 158 thousand lives in Indonesia [1]. Although in Jakarta, the Indonesian capital, COVID-19 was considered to be entering the endemic phase, where mortality is low [2], the countryside is still afected, and the population is expected to live side by side with the existence of the SARS-CoV 2 viruses and their variants [3].
Tere are two most common methods to explore COVID-19 antiviral medications, mainly through experiments and simulations. Experimental methods can be carried out in two ways: in vivo and in vitro. Meanwhile, the simulation method is called in silico. While experimental methods play a signifcant role in drug discovery, they can be time-consuming and costly [4]. Particularly when they are repeatedly employing the trial and error approach. To simulate the performance and efcacy of medications in treating COVID-19, the in silico method may be a preliminary step and can be very important to speed up drug discovery [5][6][7][8][9].
Tropical countries have abundant medicinal plant resources which have yet to be discovered [10][11][12]. Te binahong (Anredera cordifolia) is a plant native to tropical and subtropical South American countries and later brought to Southeast Asia. It is abundant in Indonesia, especially on the island of Java. Binahong is considered an herb and has long been used as medicine. Binahongs abundance of antioxidants makes it potentially useful as a wound healer [13] or possibly as a cure for cataracts [14]. Binahong has a strong antioxidant activity and is frequently utilized by locals as a treatment because it includes favonoid compounds in its leaves, stems, and fowers. In addition, this plant is frequently utilized to prevent the spread of related viruses like infuenza [15].
In this study, procyanidin, oleic acid, methyl linoleic acid, and vitexin were four compounds from binahong leaves [14] examined for their interactions with the major protease (Mpro) of the SARS-CoV 2 virus. Mpro is a key enzyme crucial for the processing of polyproteins translated from viral RNA, and it has a vital role in the reproduction of SARS-CoV 2 and the release of many of its proteins [16,17]. Te investigation of how each compound/ligand inhibits the replication of SARS-CoV 2 viruses in this study is facilitated with molecular docking simulations, followed by molecular dynamics, toxicity tests, and drug feasibility studies. We prescreened the docking score using an online molecular docking service. It turned out that procyanidin, oleic acid, methyl linoleic acid, and vitexin were the most viable choices to be studied. Tis study's fndings were compared with those of a control ligand, quercetin, which is efective at inhibiting Mpro SARS-CoV 2 [18]. In their publication, Agrawal et al. [19] stated that in vitro studies demonstrated that quercetin can interfere with multiple coronavirus entry and replication cycle steps.

Computational Methods
Molecular docking was carried out using the Maestro software's glide docking feature [20]. Te Desmond function of the Maestro software was then used to run a molecular dynamics simulation for 20 ns with a trajectory record of 20 ps using an npt ensemble.
Initially, the LigPrep function of the Maestro application was used to prepare the three-dimensional structures of each ligand. During this preparation, hydrogen atoms are removed. Epik v2.9 was used to check for potential ionization states of ligands and tautomers in the pH range of 7 ± 2. Using the Protein Preparation Wizard function of the Maestro program, the three-dimensional structure of proteins is prepared by attaching hydrogen atoms, eliminating water molecules that do not interact with proteins, enhancing protein structure, and optimizing hydrogen bonds in proteins.

Molecular Docking.
Te grid area is determined using the receptor grid generation feature to identify the region in the system that serves as a receptor. Te grid area is set using the receptor grid generation feature in the Maestro. Te grid is set with the inhibitory center of XF-1, the native ligand of PDB ID 7L11, at the protein's active site. We locked a grid area with the center coordinates of X � 150.68, Y � 125.21, and Z � 233.34 with a box size of 20. In the Ligand section, the Van der Waals radius is scaled to 1, with a partial charge cutof of 0.25. Molecular docking uses an extra-high level of accuracy (XP) option.

Molecular Dynamics Simulation.
We used the feature XP Glide docking for molecular dynamics simulation, where the input is prepared and docked protein and ligands. Te system builder feature is set by setting the solvent with the SPC (water molecule) model in the form of a cubic simulation box. Te simulation box is calculated using the bufer method with a distance of a � 10Å, b � 10Å, and c � 10Å. Te next step is to minimize the solvent volume in which the ions section must be set to exclude the placement of ions and salts, which are 20Å apart. Ten 4Na + ions and 0.15 M salt were added for neutralization. Te system that has been set is then ready for molecular dynamics simulation with the Desmond Molecular Dynamics feature. Molecular dynamics simulation was carried out for 20 ns, with a trajectory record of 20 ps, in an npt ensemble [7]. In addition, we conducted ligand torsion analysis. Desmond feature provides internal torsional energy, which can be used to calculate strain energy using the following equation: , Es � <E * > − <E>. (1) where E(τ) is torsion energy, k is Boltzaman constant, T is temperature, 〈E * 〉 is energy for corresponding temperature, and <E> is the ftted energy. Detailed derivation and information can be obtained in [23].

Lipinski's Rule of Five, Druglikeness Score, and ADMET Prediction.
Te results of Lipinski's rule and Druglikeness scores are shown in Table 3 and ADMET predictions are shown in Table 4. Except for procyanidin, every ligand in Table 3 complies with Lipinski's rule because this ligand has a high violation score. Procyanidin has been proven to show benefcial pharmacological properties, including antioxidant, antibacterial, and anti-infammatory efects [24]. Tis molecule has a molecular weight greater than 500 and hydrogen interactions greater than 5. Other ligands, however, are typically safe to use as drugs. Lipinski's rule is not the only factor in assessing whether a molecule is suitable for use as a drug. Even though many compounds fail to follow the Lipinski rules and are predicted to have low bioavailability, they are frequently utilized as medications on the market [25]. Te limits of Lipinski's rules demand a more thorough drug feasibility investigation, specifcally the prediction of ADMET properties. Table 4 shows the results of the ADMET prediction. Te information in the table is then analyzed using the theory detailed on the pkSCM website https://structure.bioc.cam. ac.uk/pkcsm. All ligands are fairly water-soluble, which qualifes them for drug usage and allows for simple dissolution in the body. Te very modest maximal dosage further supports this for these two ligands. Oleic acid and methyl linoleic acid ligands have the lowest volume of distribution (VDss) (measured in log L/kg) of all the ligands when it comes to distribution. Oleic acid and methyl linoleic acid ligands are CYP2D6 substrates and inhibitors based on their metabolic characteristics, which can afect how the body processes drugs because CYP2D6 and CYP3A4 cytochromes are crucial for this process. In excretion properties, the total clearance value of each ligand was in the medium category (0.3-0.7) except for methyl linoleic acid and oleic acid ligands with low total clearance values, which suggests that the kidneys have trouble excreting methyl linoleic acid and oleic acid ligands. Te last analysis is related to toxicity; all ligands passed the ADMET toxicology test. However, methyl linoleic acid and oleic acid ligands have skin sensitization properties that can harm the skin and several layers of tissue when these drugs are in the body.
Te ADMET prediction ofers more precise forecasts for each ligand's toxicity and pharmacology. Te reasoning, as mentioned earlier, leads to the conclusion that, compared to other ligands, methyl linoleic acid and oleic acid ligands tend to be inappropriate for drug use. However, we acknowledge that the drug-likeness score is only partial. Even with negative prediction, we still consider those compound to be docked.

Molecular Docking
Results. Molecular docking simulations were performed for each test ligand (an herbal compound from the binahong plant) and quercetin as the control ligand. Te data obtained are the molecular docking score      Journal of Tropical Medicine (docking score) and binding energy [26]. Te data obtained as a result of the molecular docking simulation is shown in Table 5. Te ligands are sorted by molecular docking score. A relatively small molecular binding score indicates a stable bond between the ligand and the Mpro virus target protein.
A lower molecular binding score indicates an optimal binding site and binding energy. In addition to the molecular docking score (or docking score), the binding energy (bind energy) score is also obtained using the MMGBSA feature in the Maestro program. MMGBSA is one of the most popular methods for calculating the free energy involved in ligand binding to target proteins [27,28]. Te energy scores obtained by the MMGBSA indicate that the smaller the energy score, the stronger the ligand and protein binding [29]. Based on the docking score, procyanidin has the best molecular docking scores, followed by vitexin, quercetin, methyl linoleic acid, and oleic acid. However, based on the binding energy scores, methyl Linoleic acid has the lowest binding energy score, followed by procyanidin, oleic acid, vitexin, and quercetin. Tus, procyanidin demonstrates better bonding performance with target proteins than other ligands. Tese results are not conclusive yet since both docking and MMGBASA scores can be used independently as scoring functions to predict binding afnity and ligand performance. Te extent of our calculation deviation for the docking score is 0.001.
Te 2D simulation of each ligand with the protein target is shown in Figure 3

Root Mean Square Deviation (RMSD).
RMSD is an atom's average movement or displacement at a specifc time interval. In other words, RMSD is data that shows the average distance between atoms (ligands) and proteins. Te RMSD plot for each ligand is shown in Figure 4.
Te +20 ps (0.02 ns) timestep, or the start of the simulation, saw a signifcant increase in the concentration of each ligand. Te red plot for each ligand displays the RMSD value for the alpha carbon (C-α), which is observed. Te interaction between ligand and protease is considered stable when the RMSD value of the backbone is under 2.5Å [30]. Te average C-α RMSD value for the quercetin control ligand is below 2.5Å and is stable throughout the simulation, as shown in Figure 4(a). Overall, the quercetin control ligand's average RMSD value was 2.15Å. Procyanidin and vitexin ligands both have an average RMSD value for alpha carbon <2.5Å, for procyanidin, it is 2.08Å, while for vitexin, it is 2.05. Meanwhile, the RMSD values for the ligands methyl linoleic acid and oleic acid tend to fuctuate throughout the simulation, with an average RMSD value of 2.5Å for methyl linoleic acid and 2.51Å for oleic acid.
According to the decreasing value of RMSD, the procyanidin ligand has the lowest average RMSD value. Ten, followed by quercetin (control ligand), vitexin, methyl linoleic acid, and oleic acid. Tus, procyanidin, quercetin, and vitexin have a better binding interaction with the target protein compared to the other ligands since the lower RMSD value denotes that the binding of the ligand to the target protein is relatively stable. Procyanidin is an ACE (angiotensin I converting enzyme) activity inhibitor due to tetramer, and its inhibition activity depends on its structure [31]. Although ACE and ACE-2 (which are important in SARC-COV 2) are not related directly, inhibiting ACE reduces the production of Angiotensin-2, which is important to the ACE-2 mechanism [32].
To make sure that ligands are stable, we conducted ftting, which assumes that RMSD behaves like a power function. Te ftting equation is given by the following equation: where K 1 and K 2 are arbitrary constants. Te lower value of K 2 demonstrates more stable ligands. Te K 2 coefcient value of ftted RMSD is given in Figure 5. Te plot shows that Procyanidin, Vitexin, and Methyl linoleic acid have a relatively lower value of K 2 coefcient which demonstrates higher stability than other ligands.

Root Mean Square Fluctuation (RMSF)
. RMSF measures how far atomic locations have deviated from the initial point. In other words, RMSF illustrates how dynamic the protein-ligand interaction is. Similar characteristics can be seen in the RMSF plot of ligand-protein for each ligand, where the RMSF value at the N-terminal and C-terminal residues (start and end) is very high because these residues represent the tails or ends of the protein structure, which are free to move and highly reactive. Te RMSF plots are displayed in Figure 6 for each ligand. Te RMSF plot produces a plot with peaks. Te fuctuation value is high throughout the simulation since each       peak denotes a high RMSF value. In this analysis, ligand and protein interactions with RMSF values above 2.5Å are considered to have bonds that are less stable. Figure 6(a) shows the RMSF analysis for the quercetin control ligand. Te total number of residues with a high RMSF value (RMSF value >2.5Å) was 9, three of which were protein binding sites. Tere are ten residues for procyanidin ligands with high RMSF, and fve of these are protein binding sites. Te oleic and methyl linoleic acid ligands have seven residues with high RMSF values, of which three are protein binding sites. Four of the seven residues in the vitexin ligand, which has a relatively high RMSF value, are binding sites. According to the data above-given, the procyanidin ligand contains ten peaks, ten of which have an RMSF value of >2.5Å. Quercetin had nine peaks, followed by vitexin, methyl linoleic acid, and oleic acid, each of which had seven peaks. Five out of ten residues are binding sites for the procyanidin ligands. Four of seven binding site residues for vitexin ligands had RMSF values of more than 2.5Å. Oleic acid had 4 out of 7 residues, while methyl Linoleic acid had just 3 out of 7 residues. Te control ligand quercetin had only 3 out of 9 residues. Te target protein binds to the control ligands quercetin > methyl linoleic acid > vitexin and oleic acid > procyanidin, according to the results of the RMSF analysis.

Protein-Ligand Contact.
Te interaction simulation diagram is a bar chart representing the bond of each residue with the ligand and the type of bond formed between the residue and the ligand. Te plot's x-axis is the residues sorted by an index number, and the y-axis is the interaction fraction of each bond between the protein and ligands. Te bonds in the target ligand-protein are hydrogen interactions, hydrophobic contacts, ionic interactions, and water bridges.
Te hydrogen interaction is the most important bond in the simulation because this type of bond is very strong and plays a very important role in the determination and specifcation of ligands as drugs. Another dominant interaction is the hydrophobic interaction that occurs when both the ligand and the residues repel water (hydrophobic) so that they bind to each other. Hydrophobic interaction is indicated by a green bar plot, while a purple bar plot indicates hydrophobic interactions. Ionic interactions occur in   polar ligands and polar residues. A red bar plot indicates these interactions. Water bridge interactions occur at residues and ligands that interact with hydrogen and are mediated or mediated by water. A blue bar plot indicates this interaction. Te interaction between each test ligand and the control quercetin ligand is shown in Figure 7. Te x-axis on the plot shows the residues interacting with the ligand, and the y-axis on the plot shows the bond fraction. For example, the 0.3 fraction shows that +30% of these interactions occur during the simulation time. However, in the plot of some ligands, there are several residues whose fraction value is greater than 1, which indicates that these residues have several contacts at once.
Overall, procyanidin ligands had the most interactions with residues among other ligands, with a reasonably high interaction fraction, reaching number 3, and it was followed by control ligands quercetin, vitexin, oleic acid, and methyl linoleic acid. Procyanidin ligands, control quercetin, and vitexin have a dominant hydrogen interaction, making the interaction between these ligands and the target protein very strong. In addition, interactions with water molecules (water bridges) are also quite dominant.
Meanwhile, oleic acid and methyl linoleic acid ligands have low interaction fractions. Te interaction of these ligands is also not as signifcant as the other three, which in turn, makes the interaction of these two ligands with the target protein poor. Overall, the most dominant interactions are hydrogen, hydrophobic, and water bridge interactions.

Ligand Torsion Analysis.
Each ligand has a single, rotatable bond. In molecular dynamics simulation, ligands as chemical compounds move without breaking the chemical bonds. One of the movements that occur is rotamer, where this movement is in the form of rotation that occurs in single bonds in the ligand [33]. Tis ligand torsion conformation  SER_46  MET_49  PHE_140  LEU_141  ASN_142  GLY_143  SER_144  CYS_145  HIS_163  HIS_164  HIS_172  MET_165  GLU_166  LEU_167  PRO_184  PRO_168  VAL_186  ASP_187  ARG_188  THR_190  ALA_191  GLN_192  GLN_189  ALA_193  ALA_194 THR_24  THR_25  THR_26  MET_49  PHE_140  LEU_141  LEU_27  ASN_142  ASN_119  GLY_170  GLY_143  SER_144  SER_46  CYS_145  CYS_44  HIS_163  HIS_41  HIS_172  HIS_164  MET_165  GLU_166  GLU_47  LEU_167  PRO_168  ASP_187  ARG_188  THR_304  THR_190 GLN_192  can provide information on how well the ligand torsion is based on theoretical calculations and the position the ligand tends to desire throughout the simulation [23]. Te value of internal torsional energy is needed to see the strain energy present in the ligand. Strain energy is the total energy of the molecule. Tis strain energy can explain how easily or not the ligand conforms. Te strain energy value is obtained by equation (2).
From the calculation results of each rotatable bond on each ligand, the internal torsional energy values are shown in Table 6. Table 6 indicates the smallest strain energy values indicated by the methyl Linoleic acid ligand and the control ligand quercetin, followed by oleic acid, vitexin, and procyanidin, which had the highest strain energy value. When the ligand binds to a protein, the conformation of the ligand will experience various disturbances (rotational movement) that come from the ligand's binding to the protein itself. Protein and ligand interactions can be either attractive or repulsive. Due to this ligand-protein interaction, the conformation of the ligand will be seen to fuctuate at higher temperatures [23], and this phenomenon is explained by parameter b in equation 4.4 [34].
Strain energy indicates whether a ligand has Procyanidin and vitexin ligands have a higher strain energy value than the control ligand quercetin, which indicates that these two ligands do not conform as much as the quercetin and the interaction between the ligand and target protein is relatively stable. Based on the results of molecular docking and molecular dynamics analysis, both procyanidin and vitexin have good performance in terms of interaction with the target protein. Tis proves that procyanidin and vitexin ligands do not form as much conformation as quercetin control ligands when interacting with target proteins, as seen from the E s values of these ligands, which are more favorable (positive) than control quercetin ligands and it was found that many conformations to form stable interactions with the target protein [23].

Conclusions
Overall, all ligands complied with Lipinski Rule of Five, except for procyanidins. From the score, methyl linoleic acid and oleic acid are not eligible because they have negative scores. Tus, two compounds are not eligible to be used as a drug. ADMET predictions also support that these ligands are not suitable for drug use. Based on molecular docking scores and binding energy, procyanidin, vitexin, and quercetin control ligands demonstrated stronger interactions than methyl linoleic acid and oleic acid ligands. Molecular dynamics simulation showed that procyanidin had the best RMSD and RMSF performance of all ligands, followed by vitexin and quercetin control ligands. From this simulation, it is known that the dominant interactions in each ligand and target protein are hydrogen, hydrophobic, and water bridges interactions. In addition, the strain energy calculation shows that even though procyanidin and vitexin have good performance in terms of interaction with the target protein, they have a higher strain energy value than the control ligand quercetin, which indicates that these two ligands do not conform as much as the quercetin. However, overall ligands extracted from binahong leaves have good performance in terms of molecular dynamics and docking score results comparable to quercetin inhibiting Mpro SARS-CoV 2.

Data Availability
Te data that support the fndings of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.