Bioinformatics approaches to investigate the phytochemicals of Centella asiatica against the main protease of SARS-CoV-2

The recent pandemic caused by the novel coronavirus SARS-CoV-2 has impacted global health by increasing mortality and unexpected infection rate. Extensive clinical research is undergoing to repurposing the old drug against this virus. So, this is an emerging need to develop therapy against the virus. Plant-derived natural products have proven to be potent therapeutics for several infections and diseases. Centella asiatica, is a native plant of the Indian subcontinent, has been vastly using as folk medicine against diseases including infectious diseases. So, using bioinformatics approach we identied and checked the phytochemicals of the plant as inhibitors against the main protease (Mpro), the key regulatory enzyme of the SARS-CoV-2 lifecycle. Computer-aided drug designing methods were performed to reveal the best nine drug-like phytochemicals those theoretically have the higher binding anity of inhibiting Mpro. This outcome may direct to the development of potent therapeutics against the SARS-CoV-2 and demands experimental validation.


Introduction
Coronaviruses (CoVs) are enveloped viruses containing a single positive-stranded RNA. SARS-CoV-2 is the novel lineage of a group β-coronavirus (Licciardi et al., 2020;Lou et al., 2020;Al-Qahtani, 2020;. The Full length of this novel coronavirus (SARS-CoV-2) genome ranges from 29891 to 29903 nucleotides (nt) long Chen et al., 2020). The disease caused by this novel coronavirus was named COVID-19 (Corona Virus Disease 2019) by World Health Organization (WHO) Gorbalenya et al., 2020). First cases of COVID-19 infection reported in the Wuhan city of Hubei province in China in December 2019. According to WHO in nearly four months, COVID-19 has reached 223 countries, with a record of con rmed cases of 89,048,345 individuals and con rmed the death of 1,930,265 infected people till 12. 01.2021. [https://www.who.int/emergencies/diseases/novel-coronavirus-2019]. SARS is categorized as zoonotic diseases as they are transmitted by intermediated hosts, for example, palm civets and dromedary camels (Lu et al., 2020). WHO announced that still there is lacking of evidence about the protection of recovered from COVID-19 from secondary infection having antibodies [https://www.who.int/emergencies/diseases/novel-coronavirus-2019/technical-guidance/criticalpreparedness-readiness-and-response-actions-for- covid-19]. Recent reports on immunological responses showed that individuals after SARS-Cov-2 infection do not have a uniformly robust antibody response . Hence most of the people's naturally occurring immunity after Covid-19 infection is thought to be suboptimal and short-lived and unreliable to achieve herd immunity.
Although the research is progressing to develop new vaccines or therapy against this virus for global mass population, however, currently inadequate quantities of vaccines are available to stop the outbreak and there is no known effective treatment. Therefore, extensive research is ongoing to develop better therapeutics using novel drug discovery approaches (Zhavoronkov et al., 2020). Plant phytochemicals provide a rich resource for computational drug discovery. Many studies targeted plant products as SARS CoV-2 inhibitors and applied computer-aided drug design techniques to nd their higher binding a nity. For example, phytochemicals of Azadirachta indica (Neem) (Borkotoky and Banerjee, 2020) and alkaloids as well as terpenoids from African medicinal plants (Gyebi et al., 2020) were computationally analyzed while in silico screening, techniques were performed for nding potential inhibitors.
Mpro, being a signi cant CoV enzyme, can be a potential drug target as it plays a central role in interceding the replication and transcription of the virus (Yoosook et al., 2000;Jin et al., 2020;Anand et al., 2002). Mpro is the main essential enzyme in the viral life cycle that may work with or without the aid of closely related human homologs, making it an attractive target to ght against COVID-19 by antiviral drug designing (Pillaiyar et al., 2016). Plethora of evidence suggested that the herbal remedy of Centella asiatica has been used as a therapeutic agent against a wide range of diseases and medical conditions in South Asia for centuries (Brinkhaus et al., 2000;Gohil et al., 2010;Singh et al., 2010). Accumulating experimental proof that in both in vitro and in vivo models suggested that C. asiatica exhibits wound healing (Sawatdee et al., 2016), neuroprotective (Gray et al., 2017), antioxidant (Ari n et al., 2011), anticancer activities (Ha z et al., 2020 and anti-in ammatory effects (Cao et al., 2018). Series of in vitro and in vivo model, preclinical with clinical studies have been carried out with an emphasis on the protective effects of C. asiatica against cardiovascular diseases as well as cardiovascular-related clinical conditions like as atherosclerosis, hypertension, hyperlipidemia, hyperglycemia, oxidative stress, and in ammation. Many of these studies found that C. asiatica protects against cardiovascular diseases and their related conditions (Razali et al., 2019). In virus inhibition logarithm test to investigate the antiviral e cacy, aqueous extract of C. asiatica was showed high antiviral activity to inhibit type 2 herpes simplex virus (HSV-2). Alcohol extracts also showed excellent results (Zheng, 1989). Another study has proved that C. asiatica contains active drug like components with intracellular anti-HSV-2 activities (Yoosook et al., 2000). So, we selected C. asiatica phytochemicals to nd the lead compounds with capabilities to inhibit Mpro. Hence, in computer-aided drug designing and discovery, molecular docking can be the vastly used best alternative in silico technique for nding the binding modes of lead compounds to the target protein like Mpro of SARS-CoV-2 (Azam et al., 2014;Shushni et al., 2013;Azam et al., 2012;Ahmed et al., 2012).
The selective treatment choices are almost impossible as there is no targeted therapeutics against COVID-19 (Yoosook et al., 2000). So, for the sake of nding lead compounds for clinical use, we started combined computer-aided drug design, virtual drug screening, high-throughput screening and molecular dynamics simulation study to identify the lead compounds that can inhibit the COVID-19 virus main protease (Mpro). This study can potentiate the development of better therapeutics against SARS-CoV-2 infection.

Materials And Methods
2.1 Retrieval of the structure of SARS-CoV-2 Mpro and phytochemicals list of C. asiatica The human novel coronavirus main protease protein (Mpro) 3D (three-dimensional) crystal structure (accession id-6LU7) was retrieved from Research Collaboratory for Structural Bioinformatics (RCSB) Protein Data Bank (Berman et al., 2000). Unnecessary objects such as default given ligands and water molecules were removed from the protein data bank (PDB) le (6LU7) using discovery studio 4.5 and PyMOL -2.3.3 (BIOVIA, 2015;Yuan et al.,2017). In case of PyMOL, rst it has to be uploaded to the software and then clicked sequence information and found the unnecessary objects during preparation. Then the le was saved as a PDB le format and used for further studies. Centella asiatica showed prospective results in many studies against HSV 1 and 2 by itself and/or its compounds. After listing a total of 107 compounds, which are the phytochemicals of C. asiatica from the literature (Brinkhaus et al., 2000;Razali et al., 2019;Azerad, 2016;Chandrika and Kumara, 2015;Roy et al., 2013;Zahara et al., 2014;Seevaratnam et al., 2012;Bylka et al., 2014), online database PubChem was searched for 2D structures of the phytochemicals in SDF (Structure Data File) le format (Cheng et al., 2014). Afterwards, the SDFs were converted into 3D structure le PDB by using the Online SMILES Translator and Structure File Generator (Oellien and Nicklaus, 2004). After saving all the SDFs into PDB, the phytochemicals were ready as ligands for docking study with the target Mpro.

Molecular Docking Simulation
Active site of the Mpro identi ed before going to molecular docking simulation study through CASTp (http://sts.bioe.uic.edu/castp/index.html?2r7g) and cross checked by another server named metaPocket (https://projects.biotec.tu-dresden.de/metapocket/). In case of CASTp, rst the PDB structure of the Mpro has to be uploaded to the server and then clicked calculation to nd the active site amino acids of Mpro. Molecular docking simulation was carried out using PyRx 0.8 virtual screening software (Yuliana et al., 2013). For simulating the best interaction, the docking was performed setting the center in axis x-(-26.1601), axis y-12.5823 and axis z-59.0673 with the dimension was in axis x-51.3732 Å, axis y-66.9737 Å and axis z-59.6071 Å. After docking simulation, the protein data bank partial charge & atom type (pdbqt) le format, given by PyRx as output, was saved for further protein-ligand interaction analysis.

Investigation and visualization of docking simulation result
After docking simulation, PyRx 0.8 was produced with nine possible binding positions as output for each compound. Poses with the highest negative binding free energy (kcal/mol) were selected as the best pose for corresponding ligand binding. After docking simulation was performed, 9(nine) compounds were selected based on docking energy ranking. After that, the best-predicted poses were visualized and analyzed by using Discovery Studio 4.5 and PyMOL -2.3.3 for hydrogen bonding.

In silico ADMET analysis
The ADMET (absorption, distribution, metabolism, excretion, and toxicity) prediction and analysis is the basic step to turn a compound into a drug. This is the process that can generate lead molecules that carries the higher binding a nity to show the satisfying result in ADMET performances in clinical trials. As the ADMET denotes absorption, distribution, metabolism, excretion, and toxicity properties of a druglike molecule, the prediction and analysis of the ADMET properties of the selected drug-like phytomolecules were performed by the online-based tools. All these ADMET properties were carried out and also compared the results given to all of our selected drug candidates through these 4 mostly used online servers like Molinspiration , admetSAR (Yang et al., 2019), SwissADME (Daina et al., 2017), and pkCSM (Pires et al., 2015). The SMILES (Oellien and Nicklaus, 2004) format of the listed phytochemicals of C. asiatica were uploaded to those online servers to calculate the ADMET properties and tabulated accordingly.

Calculation of toxicity potential
Various attributes of the drug-related properties like tumorigenicity, mutagenicity, irritation, reproductive effect, drug-likeness, and drug-score prediction were analyzed using Osiris Property Explorer (Sander, 2001). The SMILES (Oellien and Nicklaus, 2004) format of the drug-like molecules of C. asiatica were uploaded to Osiris Property Explorer (Sander, 2001) to calculate the toxicity properties and tabulated accordingly.

Molecular Dynamics Simulation
The molecular dynamics (MD) simulation of the receptor and epitope complex was carried out through YASARA dynamics (Krieger et al., 2004) tools where AMBER14 force eld (Case et al., 2005) was employed. The co-crystalized protein structure was used as a control in this study and, also ligand free apo protein was utilized to compare with nine docked complexes. Initially, the complex was optimized and cleaned and thereafter a cubic simulation cell was created with a box size of (96.0795× 96.0795×96.0795). The cell density of the system was 1.012 gm/cm 3 where TIP3P or (Transferable intermolecular potential 3 points) the model was applied. The periodic boundary condition was maintained. The physiological condition of the simulation system was set as (298K, pH 7.4, 0.9% NaCl) and the Particle Mesh Ewald method was applied to de ne long-range electrostatic interaction with a distance of 8Ǻ (Krieger et al., 2006). The initial energy minimization of the system was done through the steepest gradient approach by employing a simulated annealing method. Finally, the molecular dynamics simulation was performed for 50ns and RMSD (Root Mean Square Deviation), RMSF (Root Mean Square Fluctuation), Rg (Radius of Gyration), SASA (Solvent Accessible Surface Area), MM-PBSA (Molecular Mechanics-Poisson Boltzman Surface Area) (Krieger and Vriend, 2015).

Molecular Docking Simulation
The molecular docking analysis performed using PyRx 0.8 for 107 compounds, which are the phytochemicals of C. asiatica. PyRx 0.8 produced nine possible binding positions as output for each compound. Out of nine possible ligands binding positions, the best one was chosen for each compound based on the lowest docking energy. From 107 compounds, only 9 compounds were selected based on the binding energy ( Figure 1). Amongst the 107 compounds screened, 9 compounds, D1 to D9 (Figure 2), showed the strongest binding energy values with ≤ −9.0 kcal/mol. Based on the binding energy, the 9 compounds, have a very strong binding interaction with Mpro compared to other C. asiatica compounds, demonstrating their potential to be used as promising inhibitors. The docking energy values between ligands and proteins shown through hydrogen bonds and hydrophobic bonds as well as their interacting amino acid residues are presented in table 1 and gure 3 and supplementary gure 1. The non-bond interaction distance between amino acids, interaction category, and type of interactions is also displayed in table 2 and supplementary table 1. The anti-SARS-CoV-2 activity of the C. asiatica compounds into the Mpro of SARS-CoV-2 was found in the following order: D1 ≥ D5 ≥ D2 = D8 = D9 ≥ D3 = D4 = D6 = D7. The docking energy was negative, in the range of -9.0 to -9.5 kcal/mol (Table 1). Interactions with amino acid included Lys5, Gln127, Cys128, Arg131, Lys137, Thr199, Trp218, Phe219, Arg222, Lys236, Tyr239, Leu271, Leu272, Gly275, Met276, Arg279, Leu282, Leu287, Glu288, Asp289 and Glu290 (Table 1). The docking simulation results are excellent evidence of the anti-SARS-CoV-2 activity of the C. asiatica compounds. The highest number of hydrogen bonding interactions was observed in D4 and the lowest number of hydrogen bonding interactions was observed in D2 (Table 1). All compounds showed the conventional hydrogen bond and alkyl bond except the D9 compound. D9 showed the conventional hydrogen bond, carbon-hydrogen bond, alkyl bond, and Pi-Pi-T-shaped bond (Table 2). Detailed molecular interactions of the ve lead compounds D1, D2, D5, D8 and D9 ( Figure 3 and Table 2), which showed the lowest binding free energies ranging −9.5 to −9.1 kcal/mol, revealed that D1, D2, D5, D8, and D9 formed 9, 4, 6, 6 and 3 conventional hydrogen bond interactions and 5, 4, 1, 5 and 1 alkyl bond respectively while D9 compound formed 2 carbon-hydrogen bond and 1 Pi-Pi-T-shaped bond ( Table 2). The most common interacting amino acid residues are Lys 5, Arg 131 and Tyr 239 (Table 1). All three interacting amino acid residues were found in three compounds.

Pharmacokinetic properties study of drug compounds
A successful oral drug is one which is quickly and completely absorbed from the gastrointestinal tract, distributed speci cally to its site of action, metabolized in a way that does not immediately remove its activity, and eliminated properly, without causing any harm to the organs in the body. Because of poor pharmacokinetics (PK) properties, approximately half of all drugs in the development fail to make it (Lipinski et al., 1997). The pharmacokinetics properties such as absorption, distribution, metabolism, excretion, and toxicity (ADMET) have become most important in the selection and improvement process of drug compounds. Therefore, early prediction of ADMET properties has signi cant contributions that increase the success rate of the C. asiatica compounds in future development processes. The pharmacokinetic properties of all the compounds are listed in Tables 3, 4 and 5. Most of the orally administered drugs have a molecular weight is less than 500 and a miLogP (logarithm of partition coe cient) is equal or less than 5. In this study, we found that all compound miLogP value is -1.287 to 1.027 (Table3). For a good oral bioavailability score, the number of the rotatable bond must be ≤ 10 and Topological Polar Surface Area (TPSA) values ≤140 Å² (Veber et al., 2002). In the present study, the number of rotatable bonds of all the compounds is ≤ 10 ( Table 3).
Molecular descriptors such as molecular weight, number of H bond donor, number of H bond acceptor, number of rotatable bonds were calculated using admetSAR; TPSA, fraction Csp3, Molar Refractivity were calculated using SwissADME; miLogP and water solubility (LogS) were calculated using Osiris property explorer.
ADMET properties such as Plasma protein binding, 3A4 Substrate, 3A4 inhibitor, 2C9 substrate, 2C9 inhibitor, eye irritation, acute oral toxicity, and honey bee toxicity were calculated using admetSAR. Human intestinal Absorption, Caco-2 permeability, Oral bioavailability, Blood-brain barrier (LogBB), Fraction unbound in plasma, CNS permeability, Volume of distribution (L/kg), Renal OCT2 substrate, Total clearance, Hepatotoxicity, AMES toxicity, Oral Rat Acute Toxicity, Oral Rat Chronic Toxicity and Maximum tolerated dose in human were calculated using pkCSM, In addition, oral bioavailability, lipophilicity, and synthetic accessibility were calculated using SwissADME; GPCR ligand, Ion channel modulator, Kinase inhibitor, Nuclear receptor Ligand, Protease inhibitor, and Enzyme inhibitor were calculated using molinspiration. This is noteworthy that all these essential parameters were found acceptable for the 9 (Nine) predominant C. asiatica compounds as potential phytochemicals against the main protease; Mpro of SARS-CoV-2.

Toxicity risks and drug score assessment
The prediction of the toxicity risks of compounds is much more convenient. In the present study, Osiris property explorer was used to calculate toxicity risk parameters such as mutagenicity, tumorigenicity, irritating effects, and reproductive or developmental toxicity effects of all the C. asiatica compound's, D1 to D9 (Table 6). The predictions of these parameters are based on the functional group similarity for the query molecule with the in vitro and in vivo validated compounds present in the database of this online program. The toxicity results can be visualized using color codes; green color shows low toxic tendency whereas red color shows a high tendency of toxicity. In the toxicity, screening results in compound D7 show a high risk of irritation. On the other hand, the rest of the compound has a low risk of toxicity (Table  6).
To assess the C. asiatica compound's overall potential to qualify for a drug, the overall drug score was calculated (Table 6), which combines toxicity risk parameters, hydrophobicity (miLogP), water solubility (LogS), molecular weight and drug-likeness of the compound. miLogP values are directly proportional to the oral hydrophobic ity of the drug. The more hydrophobic the drug, the higher is the ability of the drug to circulate longer in our body. It would not be easy to excrete such a drug. In the present investigation, the miLogP values of the drug molecules were observed to be in the range of -1.287 to 1.027 (Table 3)

Molecular Dynamics Simulation analysis
To assess the structural integrity and stability of the protein and ligand complex, molecular dynamics simulation was performed for all nine complexes. The RMSD of the C-alpha atom of the simulated complex revealed that D1, D2, D3, D8, and D9 complex had an initial lower RMSD peak and subsequently reached stability after 10ns (Figure 4). During the whole simulation time, all of the complexes did not exceed 2.5Ǻ value which indicates favorable rigidness and stabilized nature of the complex. On the other hand, the ligand free complex or apo protein and the control complex had a higher RMSD trend than D1, D2, D5, D8, and D9 which con rms the comparatively less exible nature of the docked complex. However, the D9 complex had a slightly higher RMSD pro le than other complexes which indicates less rmness whereas D5 and D1 ensured fewer uctuations. The supplementary gure S2 also con rmed that D3, D4, D6 had more stability except for D7 where Apo and control exhibited less stability. The Root Mean Square Fluctuation of the simulated system aid to recognize the exible nature of the amino acid residue. Most of the residues of D1, D2, D3, D8, and D9 con rmed less exibility which approves the stable nature of the system. The amino acid residue, Ser1, Gly2, Glu47, Gly215, Val303, Thr304, Phe305, and Gln306 exhibited the most exibility than other residues. Also, the amino acid residue Thr45(betaturn), Ser46(beta-turn), Glu47(beta-turn), Asp48(beta-turn), Met49(beta-turn), Leu50(beta-turn), Asn51(beta-turn), Pro52(beta-turn), Asn53(beta-turn), Tyr54(helix-strand), Glu55(helix-strand), Asp56(helix-strand), Leu58(helix-strand), Ile59(helix-strand), Arg60(helix-strand) from the Apo protein has more exibility in beta turn and helix region which reduces in signi cant degree upon binding with ligand. From the supplementary gure S2, it can be con rmed that the complex D3, D4, D6 and D7 has less exible amino acid than Apo and control, this structure shows less labile nature, hence indicates the stability. This result along with RMSD value re ects that all docked complexes are more stable and structurally rigid than the control complex.
We also calculated SASA value from the simulation trajectory to understand the change in the surface area. Figure 4 illustrated that, SASA pro le did not uctuate too much for D1 and D5 which approves no signi cant change in the surface area of these two complexes. However, D2 and D9 had a lower SASA pro le from 20 to 30ns which may be responsible for the shrunken of the surface area. Additionally, D8 exhibited a higher trend from 15 to 30ns which indicates the slight expansion of the surface area. The ligand free protein structure initially expands the surface area and from 0 to 18ns and thereafter stabilized where the control structure had lower SASA value than other complexes during whole simulation time. Although the control complex had contracted nature, lesser deviation in other complexes con rms less change in the protein surface area. Additionally, D3, D4, D6, and D7 had an upper peak from 10 to 20ns and stabilized subsequently (supplementary gure S2). Among the complexes, D4 displayed expansion and slightly exible nature.
The labile nature and rigidness of the system can be ensured through the radius of gyration. From gure 4, it can be observed that among all ve complexes, D1, D5, D8 lower than D2 and D9 which indicates the compact nature of D1, D5, and D8. The control and apo structure did not reveal any signi cant uctuation and quite similar trend with D1, D8 and D9 whereas D2 and D5 had an upper Rg pro le. However, In case of supplementary gure S2, D6 is less rigid than D3, D4 and D7 as it uctuates more than those complexes.
The MM-PBSA analysis (Figure 4) showed that the D9 complex displayed more positive energy than D1, D2, D5, and D8 which established better binding of D9 complex than the other four complexes. The complex D2 and D9 had more binding energy than control complex which indicates favorable binding than other complexes. Moreover, the binding energy of D1, D2, and D8 were also found signi cant and the D5 complex had the lowest binding energy among the ve complexes.
In case of Post-MD study, we also found supporting results from molecular dynamics study. The D1 complex had common binding residue in Asp289, Leu287, Asp289, Leu272, Met276 for pre and post-md complex whereas D2 complex had similar binding patterns in Lys137, Tyr239, Leu287, Leu272, Met276 residues. Also, D3 complex shows similar types of interactions with Lys137, Tyr239, Leu287, and Cys128 and also, D4 exhibits the same pattern. Moreover, Lys5, Lys137, Gln127, Glu290, Leu282 conserved for D5 and similar hydrogen bond also found in Thr199, Tyr239, and Asp289 for D6. The pre and post-MD docked complex of D7 stabilized by Lys137, Tyr289, Thr199, Leu287, and Met276 residues. On the other hand, common interactions were observed in Lys5, Thr199, Tyr239, Leu272, Met276, Cys128, Leu287, Tyr289 for D8 where D9 complex also con rmed the conservation of binding dynamics in both post and pre-MD structure (Supplementary Table S2).

Discussion
The COVID-19 is affecting severely millions of people and taking thousands of precious lives every day over the globe due to its pandemic behavior making it a winning challenge for the scienti c communities across the world for the survival of the human race. But there has been no satisfactory breakthrough yet made in the treatment of COVID-19 (Lake, 2020;Yuen et al., 2020;Fang et al., 2020;Dong et al., 2020). Although some candidate drugs were studied and proposed to treat the disease, the attempts lie ambiguous due to low e cacy .
There are several advantages of drugs over vaccines (Gutteridge, 1991;Kremer and Snyder, 2003), a study found moderate evidence of in uenza vaccines with mild gastrointestinal events; trivalent inactivated in uenza vaccine (TIV) was associated with febrile seizures (Maglione et al., 2014). Moreover, developing a potent vaccine of RNA viruses are di cult as there always is a challenge to get a strong immune response yet maximal effect in the rise of antigenic changes as RNA viruses are very much prone to rapid mutation thus evolution. Although vaccines are developed targeting epitopes that are strongly conserved, moderately conserved, and poorly conserved antigenic sites in their surface protein, due to mutation these epitopes can be changed causing a serious problem in vaccine designing (Steinhauer and Holland, 1987).
In our current study, molecular docking was applied for high throughput screening and advanced analysis like molecular dynamics of the lead molecules of C. asiatica against Mpro of SARS-CoV-2. The coronavirus main protease protein (Mpro) also known as coronavirus 3C-like protease (3CLP) plays the most vital role in controlling viral replication and transcription by performing extensive proteolytic processing of replicase polyproteins, making it a potentially attractive drug target (Zhavoronkov et al., 2020;Yang et al., 2005). ADMET analysis was done to check the drug and potentiality of the drug candidates inside the human body.
When it comes to comparing the study of hydrogen (H) bonds formation, Khaerunnisa et al achieved 3 to 8 H bonds (Khaerunnisa et al., 2020), Bouchentouf and Missoum achieved the number of H bonds was 0-3 (Bouchentouf and Missoum, 2020), Gyebi et al's study achieved the number of H bonds was 0-1 (Gyebi et al., 2020), Enmozhi et al found 4 H bonds (Enmozhi et al., 2020), and nally, Lobo-Galo et al study achieved the number of H bonds was 0 to 5 (Lobo-Galo et al., 2020) but on the other hand in our study, we got 4 to 11 hydrogen bonds formed between the lead compounds and the target enzyme Mpro (Table   1).
We got incredible results when we compared the logP values and logS values of our lead compounds with similar computational drug designing studies done by others. In a previous study, Bouchentouf and Missoum showed logP value was 1.06 to 3.52, and logS value was -8.24 to -2.01 (Bouchentouf and Missoum, 2020), Gyebi et al showed ClogP value was 3.31 to 4.80 (Gyebi et al., 2020). We have found that the miLogP value is -1.287 to 1.027 and logs value is -5.319 to -4.073 in our study (Table 3). Gyebi et al showed the bioavailability score was 0.55, non-AMES toxic, and non-carcinogens (Gyebi et al., 2020) whereas in our study the bioavailability score is 0.17 to 0.55 and all the target compounds are non-AMES toxic and non-carcinogens. The acute oral rat toxicity (LD50) was found to be 2.162 mol/kg and Chronic oral rat toxicity (LOAEL) was found to be 1 log mg/ kg_bw/day (Enmozhi et al., 2020). In our study, the acute oral rat toxicity (LD50) is 2.656 to 3.452 mol/kg and chronic oral rat toxicity (LOAEL) is 1.911 to 3.846 log mg/ kg_bw/day (Table 4).
When it comes to cytotoxicity or side effect studies, our proposed compounds show excellent results in in silico prediction. In a study, C. asiatica was found as a causative factor in abnormality in reproduction such as infertility in mice and abortion in women in chronic treatment (Orhan, 2012) but the plant C. asiatica phytochemicals we have worked with, Osiris property explorer showed no predictions of any abnormalities or toxicity risk in the reproductive system. Moreover, no prediction of any mutagenic effect and the tumorigenic effect was found although one compound showed an irritating effect (Table 6). The pharmacokinetic properties of the selected molecules were carried out by using online-based software tools. All the lead compounds tested in our study showed a moderate to the high absorption rate.
Compound D1 showed a high absorption rate of 100% whereas compound D4 showed the lowest absorption in 25.448%. All compounds are the substrate of CYP3A4 and no carcinogenicity and AMES toxicity are present in the prediction (Table 4). ADME and toxicity analysis of these compounds suggest that they could be used for the development of new drugs to treat COVID-19 although further validation is necessary. MD simulation is an imperative method to explore the protein-ligand complex in real-time, widely used to assess the conformational variability and stability of the protein systems Bappy et al., 2020;Islam et al., 2019). The molecular dynamics simulation represents stable RMSD compared to ligand free apo structure and control which provides insights about the structural integrity of the docked complex (Dash et al., 2019;Arifuzzaman et al., 2020). Furthermore, Rg pro le was illustrated along with RMSD pro le, where RMSD describes the uctuation during simulation periods, and protein folding and degree of compactness describes through Rg. The RMSD and RMSF pro le of all 9 systems con rms the rigidness and less exibility. Moreover, post-MD binding interaction analysis supports the ndings from molecular dynamics simulation as almost every hydrogen and hydrophobic bond remains in rigid in post-MD docked complex. After MD analysis, we have also superimposed all the molecular docking complexes and found a structural similarity to them with the previous complexes ( Figure S3). Furthermore, we analyzed the hydrogen bond stability from the molecular dynamics simulation trajectory. The supplementary gure S4 demonstrated that most of the complex has more hydrogen bonds than the control complex. However, complex D4, D5, and D6 have more hydrogen bonds than other complexes which indicate the comparative stability of the complex. When it comes to molecular docking, Out of nine drug candidates, D1 showed slightly higher energy than others. However, in the case of ADMET properties in the human body, D9 expressed the best hit although D1 showed the highest absorption, out of those nine compounds. When it comes to molecular dynamics, D1, D2, D5, D8, and D9 showed more suitability as a candidate drug rather than D3, D4, D6, and D7. However, MM-PBSA analysis revealed that the D9 complex demonstrated the most favorable binding than other complexes. This study is limited to in silico modeling and analysis of the bioactive compound of C. asiatica against COVID-19. Hence experimental validation is fundamental to assess the e cacy of the selected bioactive compounds.

Conclusion
Overall our in silico drug designing ndings revealed nine potential phytochemicals from the bioactive compounds of C. asiatica against SARS-COV-2 replication and propagation associated protein-Main Protease. Based on our comprehensive computational molecular docking, pharmacokinetics study, molecular dynamics simulation, we have identi ed nine phytochemicals of C. asiatica exhibited enhanced binding a nities to target the main-protease of SARS-CoV-2 compared to approved anti-viral drugs. Most importantly, the proposed compound is also predicted to less likely to have side-effects in patients. The main advantage to use the phytochemicals of C. asiatica, it is readily available and has been used as a traditional medicine in many regions of Asia including India and China. Since vaccine development against SARS-CoV-2 requires a considerable amount of time, therefore, our proposed bioactive compounds of C. asiatica could be a potential alternative therapy against this deadly virus.
The accumulative ndings of our study make a stronger case which demands future studies to investigate the possible preclinical and clinical e cacy of these agents' e cient treatment of SARS-CoV-2.    Table 6: Drug-likeness/scores and toxicity calculations of Centella asiatica compounds based on Osiris property explorer Figure 1 Flow diagram of methodologies and pipeline applied in this study.