Effectiveness of Natural Antioxidants against SARS-CoV-2? Insights from the In-Silico World

The SARS CoV-2 pandemic has affected millions of people around the globe. Despite many efforts to find some effective medicines against SARS CoV-2, no established therapeutics are available yet. The use of phytochemicals as antiviral agents provides hope against the proliferation of SARS-CoV-2. Several natural compounds were analyzed by virtual screening against six SARS CoV-2 protein targets using molecular docking simulations in the present study. More than a hundred plant-derived secondary metabolites have been docked, including alkaloids, flavonoids, coumarins, and steroids. SARS CoV-2 protein targets include Main protease (MPro), Papain-like protease (PLpro), RNA-dependent RNA polymerase (RdRp), Spike glycoprotein (S), Helicase (Nsp13), and E-Channel protein. Phytochemicals were evaluated by molecular docking, and MD simulations were performed using the YASARA structure using a modified genetic algorithm and AMBER03 force field. Binding energies and dissociation constants allowed the identification of potentially active compounds. Ligand-protein interactions provide an insight into the mechanism and potential of identified compounds. Glycyrrhizin and its metabolite 18-β-glycyrrhetinic acid have shown a strong binding affinity for MPro, helicase, RdRp, spike, and E-channel proteins, while a flavonoid Baicalin also strongly binds against PLpro and RdRp. The use of identified phytochemicals may help to speed up the drug development and provide natural protection against SARS-CoV-2.


Introduction
SARS-CoV-2 (Severe Acute Respiratory Syndrome Coronavirus 2) originated in the Wuhan province of Central China in December 2019 [1] and its disease, COVID-19, was declared as a pandemic on 11 March 2020, after the infection spread globally [2]. This

Main Protease (M Pro )
The phytochemical ligands in this investigation were screened against M Pro . The ligands with the best docking against the Main Protease are listed in Table 1. Glycyrrhizin, 18,β-Glycyrrhetinic acid, Rhodiolin, Baicalin, and Silymarin were the best five ligands with high binding energies and dissociation constants ( Figure 2). Glycyrrhizin showed binding to M Pro at two different binding sites with the highest binding scores of −9.57 and −9.46 kcal·mol −1 and a dissociation constant of 0.11 and 0.76 µM. The best Glycyrrhizin binding site is the conventional M Pro active site located between domains I and II. This site mainly consists of Thr 24 , Thr 25 , Thr 26 Table 1). His 41 and Cys 145 are the key residues involved in the enzyme active site and have been previously defined in M Pro active site by an X-ray crystallographic structure (PDB ID 6WQF) obtained at room temperature [38]. The His 41 and Cys 145 form a catalytic dyad that interacts with the bound ligand. Other amino acids in the proposed active site Ser 46 , Leu 141 , Asn 142 , Glu 166 , Pro 168 , Gln 189 , Thr 190 , and Ala 191 [38,39] were found involved in the stabilization of the enzyme active site.    Figure 2, Table 1). The Glycyrrhizin binding to this active site may disrupt the native enzyme structure and affects its activity. This distal pocket has also been reported as a promising inhibitor binding site [40,41]. This site consists of both a loop region and β-strands.
The initial MD simulation in the case of Glycyrrhizin (active site binding) shows that an equilibrium was achieved after 25 ns, so simulations were limited to 30 ns. The RMSD and RMSF values show the flexible residues in two regions, one from residues Arg 40 -Asp 56 (Domain I) and from Ile 136 to Asp 153 (Domain II) ( Figure 3). This shows the ligand interactions with the active site residues of the enzyme. The fluctuations between Tyr 237 and Gly 251 show that Glycyrrhizin binding to the active site may also induce a conformational change in other parts of the enzyme. This conformational change may be involved in reducing the enzyme activity. The Rhodiolin-M Pro complex showed huge fluctuation around Ile 281 to Val 296 while minor fluctuations in a loop region around Val 91 to Pro 96 ( Figure 3). The radius of gyration (R) at the end of the MD simulation shows the compactness and stability of the Glycyrrhizin-and rhodiolin-enzyme complexes ( Figure S1).

Papain-Like Protease
Like the M Pro , Papain-like Protease is involved in diverse functions that make it a potential drug target [2]. The ligands with the best docking with PL Pro are listed in Table 1. Most of the ligands indirectly interacted with the active site triad Cys 111 , His 272 , and Asp 286 [43] by binding around the enzyme active site ( Figure 4). These triad residues are involved in enzyme activity [43]. Baicalin showed a strong binding with the binding energy  Figure 4). Baicalin shows π-π interactions with Tyr 268 , while H-bonding is observed with Cys 155 , Ly 157 , and Tyr 171 ( Figure 4). Previously, a binding site in crystal structure of papain-like protease (PL Pro ) (PDB ID 6WX4) was defined by the residues; Trp 106 , Asn 109 [44]. This shows although Biacalin does not directly interact with the catalytic triad, it binds in the vicinity of the enzyme active site very strongly and impairs its proper functioning. The worst docking ligands include 6,9,12-octadecatrienoic acid (binding energy −4.85 kcal.mol −1 ), docosanoic acid (binding energy −4.64 kcal.mol −1 ), and acetohydroxamic acid (binding energy −4.13 kcal.mol −1 ). These compounds were unable to bind the proposed active site (Table S1).
MD simulations of 30 ns were run for Baicalin, Hesperidin, and Solophenol, where considerable RMSD and RMSF fluctuations were found in the active site residues of PL Pro ( Figure 5). In the case of Baicalin, strong binding and low dissociation constant for the Baicalin-enzyme complex is also confirmed by MD simulations. N-terminal region from Val 21 to Pro 46 is the most flexible region, while regions including Tyr 71 to Asp 76 , Leu 101 to Gln 121 , Gly 266 to Ile 276 , and Thr 281 to Lys 292 seem to be involved in Biacalin-PL pro interactions ( Figure 5). The Hesperidin-enzyme complex Radius of gyration (R g ) for Solophenol increased in 15-30 ns of MD simulation but Baicalin and Hesperidin, due to strong binding, show the R g similar to the native enzyme ( Figure S2).

RNA-Dependent RNA Polymerase (RdRp)
The RNA-dependent RNA Polymerase (RdRp), the main replication enzyme for SARS-CoV-2, was screened against a library of phytochemicals (Table S1). Glycyrrhizin showed the strongest binding to RdRp. Interestingly, Glycyrrhizin strongly interacts with RdRp at two different binding sites with binding energies of −10.27 and −9.96 kcal·mol −1 with dissociation constants of 0.03 and 0.05 µM (Table 1, Figure 6 Figure 6). The binding site of RNA-dependent RNA Polymerase (RdRp) has already been defined by using cryo-EM structures [45] where key catalytic site residues are Lys 500 , Ser 501 , Asn 507 , Lys 545 , Arg 555 , Asp 618 , Ser 759 , Asp 760 , Asp 761 , Cys 813 , Ser 814 , and Gln 815 . It seems that Glycyrrhizin interacts with these residues with a high binding affinity (multiple H-bonds) and may impair the RdRp interactions with the RNA.
Hesperidin, though not interacting with the proposed RdRp active site, binds near to the enzyme active site, where Thr 556 , Lys 621 , Arg 624 , and Ser 628 form H-bonds and Phe 793 shows π-π interactions. Hesperidin showed a binding energy of −9.53 kcal·mol −1 and with a dissociation constant of 0.1 µM. Baicalin, Naringen, and Oleuropein bind to a completely different site in RdRp, where Thr 141 , Asn 781 , and Ser 784 are key residues ( Figure 6). Baicalin showed the best binding energy of −9.01 kcal·mol −1 with a dissociation constant of 0.12 µM. Active site residues are Phe 35 Figure 6).
A 20 ns MD simulation validates the Glycyrrhizin, Hesperidin, and baicalin interactions with RdRP. Glycyrrhizin and Baicalin show RMSD changes in the interface region consisting of residue from Val 258 to Leu 270 , in the finger region Asp 481 to Tyr 515 , Thr 738 to Tyr 770 (Figure 7). The RMSD and RMSF fluctuations seem larger in Baicalin in comparison to Glycyrrhizin and Hesperidin. In the thumb region, substantial RMSD changes were observed in all three complexes. All three ligand-enzyme complexes seem more stable in terms of potential energy than native enzymes ( Figure S3).

Spike Glycoprotein
The phytochemical ligands were screened against Spike Glycoprotein's with both open and closed states of the protein. The ligands with the best docking for spike glycoprotein are listed in Table 1. Glycyrrhizin showed the best binding energy of −9.29 and −9.49 kcal·mol −1 and a dissociation constant of 0. 16 (Table S1). The worst docking ligands include Nervonic acid (binding energy −3.92 kcal·mol −1 ), octadec-9-enyl icosanoate (binding energy −3.87 kcal·mol −1 ), and Tetracosanoic acid (binding energy −3.71 kcal·mol −1 ) for close state spike glycoprotein (Table S1).
MD simulation with a closed state of spike protein shows that Glycyrrhizin-enzyme complexes are more stable than Hesperidin-enzyme complexes in terms of potential energy and R g ( Figure S4). RMSD and RMSF fluctuations in N-terminal domain subunit 1 validate the hesperidin interactions with the residue present in this region (Figure 9). Glycyrrhizin and Hesperidin also show large fluctuations in the S2 areas of spike protein (729-769 and 955-1035 a.a).

Helicase (Nsp13) Protein
Phytochemicals ligands were docked against Helicase (Nsp13) protein and Glycyrrhizin, β-Glycyrrhetinic Acid, Solophenol A, Hesperidin, and Baicalin were found to be the best docking ligands (  Figure 10). Allosteric binding for Glycyrrhizin was observed in a region between RecA1 and RecA2 domains. The same was found to be true in the case of MD simulation, where fluctuations in RMSF and RMSD were observed in a region between RecA1 and RecA2 ( Figure 11). The Glycyrrhizin derivative, 18,β-Glycyrrhetinic acid, also showed strong interactions with the second-best binding energy of −9.91 kcal·mol −1 and a dissociation constant of 54 nM. The active site residues are Ala 4 , Val 6 , Arg 15 (Figure 10). The worst docking ligands include octadec-9-enyl icosanoat (binding energy −4.30 kcal·mol −1 ), docosanoic acid (binding energy −4.16 kcal·mol −1 ), and acetohydroxamic acid (binding energy −4.11 kcal·mol −1 ) for open state spike glycoprotein. Some of these compounds were even unable to bind the proposed active site (Table S1).

E-Channel (Envelop Small Membrane Protein)
The phytochemical ligands were docked against E-channel protein are listed in Table 1 Figure 12). The worst docking ligands include malic acid (binding energy −4.05 kcal·mol −1 ), oxalic acid (binding energy −3.24 kcal·mol −1 ), and acetohydroxamic acid (binding energy −3.10 kcal·mol −1 ) for open state spike glycoprotein. Some of these compounds were even unable to bind the proposed active site (Table S1).

Non-Specific Interactions of Selected Ligands against Human Blood Proteins
Glycyrrhizin, Hesperidin, and Baicalin were docked to a local library of selected human blood proteins (a total of 100 blood proteins) to map non-specific interactions of these ligands with non-specific proteins (Table S2). Glycyrrhizin showed the highest binding affinity against DdB1 (damage-specific DNA binding protein) with a binding energy of −11.36 kcal·mol −1 and a dissociation constant of 1.68 nM. The interactions residues included Asn 16 (Figure 13a). Baicalin also interacts with DdB1 at the same binding site (Figure 13b). In non-specific interactions, Hesperidin shows the highest binding affinity (binding energy −10.89, and dissociation constant 10.4 nM) for Integrin alpha V Beta 6 head protein normally involved in cell adhesion (Figure 13c). The top four non-specific interacting partners of Glycyrrhizin, Hesperidin, and Baicalin are detailed in Figure 13a-c. In contrast, the binding energies and dissociations constants for all 100 non-specific proteins are given in Table S2.

ADMET Properties of Selected Ligands
Lipinski's Rule of Five [46], Ghose filter (Amgen) [47], Veber's (GSK) [48] rules are used to predict ADME properties. According to the pharmacokinetic properties, all compounds show Gastrointestinal low absorption except 18-β glycyrrhetinic acid (GA), lopinavir, and euchrestaflavanone A, which have high absorption. These compounds have the least BBB permeability, and no CYP inhibition was observed ( Table 2).
The binding site for M Pro has already been defined by X-ray crystallographic structure (PDB ID 6WQF) obtained at room temperature [38] where His 41 and Cys 45 form a catalytic dyad to interact with bound ligand. Other amino acids involved in the stabilization of the active site were Ser 46 , Leu 141 , Asn 142 , Glu 166 , Pro 168 , Gln 189 , Thr 190 , Ala 191 . This active site is situated in a cleft between domains I and II [60,61]. In our study, Glycyrrhizin shows two binding sites; one includes the conventional active site of the enzyme, while the second interactions include an allosteric binding site. A previous docking analysis of Glycyrrhizin against M Pro has shown a binding energy value −7.81 kcal·mol −1 where it interacts with the proposed active site of the enzyme [62]. In another study, glycyrrhizic acid, Glabridin and Liquiritigenin show strong binding interactions (with a binding energy of −7.0 to −8.0 kcal·mol −1 ) with M Pro conventional active site [63]. The docking analysis of M Pro with FDA-approved anti-viral compounds and library of active phytochemicals [64] shows Nelfinavir potent against M Pro . Many natural compounds are found to be potential inhibitors of M Pro, including Leucoefdin [65], Leupeptin [66], Rutin [67], cannabisin-A, isoacetoside [68], epigallocatechin gallate, and epicatechin gallate [69]. Recently, Glycyrrhizin has been found to indirectly inhibit the SARS-CoV-2 replication by Inhibiting M Pro enzyme activity [70]. In our study, many ligands, including Glycyrrhizin, have been found to interact with Mpro allosteric binding sites (Table 1 and Table S1, Figure 2). In a previous study, 2400 FDA-approved drugs have been screened against M Pro allosteric binding sites, where selinexor, bromocriptine, Dihydroergotamine, nilotinib, entrectinib, digitoxin, and diosmin have shown promising binding to the enzyme [71].
PL Pro consists of an N-terminal ubiquitin-like (Ubl) domain (1-60 a.a) and a catalytic region with a right-handed thumb-palm-fingers architecture. The PL Pro binding site is found in the thumb and palm domain and is characterized by the presence of a catalytic triad (Cys 111 , His 272 , and Asp 286 ) [72]. In our study, Baicalin and Hesperidin have been found to be potential PL pro inhibitors that interact with their proposed active site ( Figure 4). GRL0617 with PL pro shows binding to the same site [73], whereas π-π interactions with Tyr 268 have shown definite inhibition of PL pro activity. Natural compounds like Caesalpiniaphenol A, and Sappanone B, also interact with Try 268 . Corylifol A, chromen, darunavir, sofosbuvir and some other drugs were screened against PL Pro . These drugs were found to bind near the proposed catalytic triad [74]. Phytochemicals from Vitex negundo L. are also found active against PL Pro [75]. Along with natural compounds, many approved antibacterial and antiviral drugs also have been repurposed [76][77][78].
Glycyrrhizin, Hesperidin, and Biacalin show strong interactions with RdRp. Glycyrrhizin binds in the enzyme's potential active site pocket while making H-bonding with Asp 760 and other residues ( Figure 6). Asp 760,761 have been proposed as active site residues, while Tyr 619 , Cys 622 , Ser 759 , Ala 762 , Glu 811 , Cys 813 , and Ser 814 are found in potential binding sites for Ribavirin, Remdesivir, and other antivirals interactions [86]. In a screening with flavonoid compounds, Delphinidin 3-O-beta-D-glucoside 5-O-(6-coumaroylbeta-D-glucoside) complex with RdRp has been found most stable, where Asp 760,761 have been found in the ligand-binding site [87]. Lanreotide, Argiprestocin, Demoxytocin, and Polymyxin B1 also interact with Asp 760 . Previously, polyphenols with binding energy <7.0 have been reported to interact with RdRp, while Remdesivir showed binding energy of 7.9 kcal·mol −1 [88] and interacts with a similar ligand-binding site as for Glycyrrhizin, Hesperidin. This site includes Asp 760,761 , and Glu 811 for Remdesivir [89,90]. In another study, approved antivirals including Ribavirin, Remdesivir, Sofosbuvir, Galidesivir, and Tenofovir have shown strong interactions with RdRp, where Ribavirin have shown binding energy of −8.7 kcal/mol [86]. Many compounds from the ZINC database have been screened against RdRp and 40 ns MD simulations were performed for Rifabutin, ZINC09128258 and ZINC09883305 [91].
Glycyrrhizin strongly binds to the S2 subunit of the spike protein. Although the binding residues are different from the receptor-binding domain (331-524 a.a.), it is expected that this interaction may bring the conformation change in the protein that may affect the receptor binding. This conformational change is also indicated by MD simulation showing large RMSD and RMSF changes in the receptor-binding region and formation of the close complex (low R g ) (Figure 8 and Figure S4). A number of flavonoid compounds were docked to spike protein and naringin has been found as the most potent compound against SARS-CoV-2 spike protein [92]. In our study, Glycyrrhizin showed the best binding energy against the open and close state of spike glycoproteins. Most of the Glycyrrhizin's interactions have been found with the S2 subunit of the spike protein (Figure 8b). In one study, 66 compounds were found to interact with RBD of spike protein and Glycyrrhizic acid was found to be the most potent antiviral and spike protein inhibitor [93].  [97], was marked by mutation of K417N and E484K in the RBD region of the spike protein. Surprisingly, the affinity of the Glycyrrhizin slightly decreased against the spike protein variants, though it still shows considerable binding energy and dissociation constant ( Figure S6). Catechins and tamibarotene have been found to interact strongly with the UK variant and triple SARS-CoV-2 variant [98,99].
In the case of SARS-CoV-2 helicase, the interface between RecA1 and RecA2 domains contains the active site residues for helicase enzyme, including Lys 288 , Ser 289 , Asp 374 , Glu 375 , Gln 404 , and Arg 567 [100,101]. In this study, Glycyrrhizin binds to an allosteric site near the catalytic cleft, while Glycyrrhetinic acid interacts with residues in stalk and 1B domain. Triphenyamine and Darunavir have been reported to bind the same active site [102]. Rutin, xanthones, and many other polyphenols have been reported to be ATPase inhibitors [59]. A database of 14,000 phytochemicals has been docked against helicase using virtual screening; out of them, 368 compounds have been found to be potent helicase inhibitors [103]. Picrasidine M from a herb Picrasma quassioides shows the best binding to helicase, where it forms 2 H-bonds with Ser 289 and one H-bond with Gln 404 and Arg 567 [103]. E-channel blockers are demonstrated to be potent antivirals by protecting hosts cells from death. Glycyrrhizin, Glycyrrhetinic acid, and Baicalin have also shown binding to E-channel protein. Proanthocyanidins have been reported to inhibit the MPro and E-channel protein of the SARS-CoV-2 [104].
Glycyrrhizin seems to interact with four SARS-CoV-2 key proteins with high affinity, where its binding with helicase and RdRp has been found to be more stable (Figure 14). Glycyrrhizin has also been reported to interact with TMPRSS2, involved in viral penetration into the host cells [105]. It also indicates that licorice root, a potential source of Glycyrrhizin, may be used as a possible cure and a household remedy for COVID-19. The non-specific interactions of Glycyrrhizin show that it may interact with Ddb1 and other non-specific human blood proteins. Glycyrrhizin as Glycyrrhizic acid from Glycyrrhiza glabra and as a derivative in the form of β-Glycyrrhetinic acid are reported to show penetration through the blood-brain barrier and are non-carcinogenic [63] (Table 2). This study shows that natural antioxidant compounds, either partially purified or in crude form, may provide protection against SARS-CoV-2 severity and complications by interacting with its key enzymes. Although we have performed 30 ns MD simulations as in the case of many protein-ligand complexes equilibrium was achieved and complex was found stable, longer MD simulations (>100 ns) in future studies may help to better understand the behavior and stability of the complexes. In vivo studies for the selected compounds are also recommended to probe the computational result and may help in the formulation of natural antivirals with less toxicity and more efficacy.

Conclusions
In the current pandemic situation, COVID-19 s battle with humanity is still continued. The availability of various vaccines has eased the aggravating situation, but the rise of SARS-CoV2 variants warns that humans have to live with the virus, evading its devastating effects. Potential natural cures/medications/home remedies without any side effects seems a viable solution in the current circumstances. Our study has examined and screened more than a hundred natural compounds from plants against six SARS CoV-2 proteins by using molecular docking and molecular dynamics simulations to identify potential bioactive compounds. Glycyrrhizin was found as the best ligand showing strong inhibition of five SARS CoV-2 proteins. Glycyrrhizin is present in a large amount in an inexpensive household herb licorice already registered for its magical curative properties against a number of diseases. Other phytochemicals found potent against SARS-CoV2 include Hesperidin and Baicalin present in citrus fruits and many other plants. In our study, Glycyrrhizin, Hesperidin, and Baicalin were docked against non-specific human blood proteins and have shown interactions with DNA binding proteins. Based on our findings, we suggest that further in vivo evaluation of Glycyrrhizin and its sister compounds as potential antivirals will signify their role in the treatment and management of COVID-19.

Materials and Methods
A total of 115 natural compounds with established therapeutic properties were targeted against six SARS-CoV-2 proteins by molecular docking. The compound structures were obtained from various chemical databases, including PubChem (https://pubchem. ncbi.nlm.nih.gov/, accessed on 25 March 2021), ChemSpider (https://chemspider.com, accessed on 25 March 2021) and MolPort (https://www.molport.com/, accessed on 25 March 2021) ( Table S1). The compound structures were obtained in form of The Spatial Data File (SDF) and optimized using the MM1 forcefield in YASARA Structure ver. 20.7.4 [106]. The ligand structures were merged in a single file prepared for virtual screening.  (Figure 1). Spike protein variants' structures as well as structures for all selected non-specific proteins were also obtained from PDB (Table S2a-c). The single-chain structures for all proteins were prepared using YASARA Structure ver. 20.7.4 [106] and heteroatoms were removed.

Molecular Docking
The ligand-protein interactions and binding energies were calculated by applying a virtual screening module in YASARA software version 20.7.4 [106] that uses a modified AutoDock-Lamarckian Genetic Algorithm. The parameters used for the virtual screening and molecular docking have been described earlier [107], where AMBER03-FF was used with a hundred global docking runs and by keeping the random seed value of 1000. AutoDock local search (LGA-LS) was also used for selected top five ligands to reassure best binding and energy minimization. Protein-ligand interactions were mapped in terms of binding energies and dissociation constants, while the docking scores were calculated by Equation (1) ∆G = ∆G (van der Waals) + ∆G (H-bonding) + ∆G (electrostatic) + ∆G (torsional free energy) + ∆G (desolvation energy) LigPlus [108] was used to obtain and ligand-protein interactions. The specific ligand was selected in the ligand-protein complex file and interactions including H-bonds were mapped. Non-specific Interactions of selected ligands against human blood proteins were studied by an inverse docking procedure where more than 100 human blood protein targets were screened against the selected ligands (Table S2a-c).

Molecular Dynamic Simulations
Molecular Dynamic simulations (MDS) were performed using YASARA Structure ver. 20.7.4 [106] with AMBER14 as a force field as described before [109]. The simulation cell was prepared by providing 20 Å water-filled space around the fully mobile protein with a density of 0.997 g/mL. The system was neutralized with 0.9% NaCl while maintaining 298 K temperature, pH 7.4, periodic boundaries, and 7.86 cut-off for long-range coulomb electrostatics forces. After the initial steepest descent minimization, MDS was performed at the rate of 1.25-2.50 fs time steps, and the simulation snapshot was saved every 100 ps. MDS of 10-30 ns were calculated for different proteins depending on the number of atoms in the simulation cell. The raw data were analyzed using GraphPad Prism ver. 7.0. [110]. RMSD and RMSF values were tabulated and analyzed for the fluctuations. A docking and MD simulation flow chart had been given in Figure 15.

Pharmacokinetics and Drug-Likeness
The pharmacokinetic properties and drug-likeness prediction of the top 10 li were performed by the SwissADME server (http://www.swissadme.ch/, accessed on 20 April 2021). It calculates the topological polar surface area (TPSA), logP (lipophilicity), and logS (solubility). The drug-likeness was predicted by following Lipinski, Ghose, and Veber rules and bioavailability scores [46][47][48]. The Lipinski's Rule of Five states that the absorption or permeation of a molecule is more likely when the molecular mass is under 500 g/mol, the value of log P is lower than 5, and the molecule has utmost 5 H-donor and 10 H-acceptor atoms [46]. Ghose filter (Amgen) [47] defines drug-likeness based on log P between −0.4-5.6, MW between 160-480, molar refractivity between 40-130, and the total number of atoms between 20-70. Veber (GSK) [48], the rule defines drug-likeness as rotatable bond count ≤10 and polar surface area (PSA) ≤ 140.

Conflicts of Interest:
The authors declare no conflict of interest.