Introduction

It is not unusual that a Coronaviridae virus has infected humans. In 2002–03, there was a Severe Acute Respiratory Syndrome (SARS) outbreak, followed by that, in 2012 the Middle East Respiratory Syndrome Coronavirus (MERS-CoV) emerged and affected the Middle Eastern countries (Rabaan et al. 2020). The tremendous evolutionary adaptability of coronaviruses to the environment and host specificity eventually gave rise to SARS-CoV-2, which has been inflicting chaos in the world with a series of new mutations (Zheng 2020). COVID-19 has surfed successive waves in the preceding two years, with new variants surfacing one after the other, leaving the infection vulnerable to rumours because of its poor understanding (Maher et al. 2021). According to the World Health Organisation, there have been a total of 5 Variants of Concern (VOCs), namely Alpha, Beta, Gamma, Delta and Omicron (Chugh et al. 2022).

Owing to the mutability of this virus, many variants of SARS-CoV-2 have originated and have been more infectious and transmissible than before. Therefore, one of the major obstacles in controlling the Pandemic has been the absence of an efficient therapeutic strategy against the existing and emerging variants (Aleem et al. 2022).

Computational biology has been proved to be a crucial tool in indicating the potential phytocompounds from medicinal plants and repurposed pharmaceuticals in the fight against SARS-CoV-2 (Scherman and Fetro 2020; Toor et al. 2021). Within the initial few months of the COVID-19 pandemic, there was a rise in clinical trials of repurposing drugs such as Hydroxychloroquine, Remdesivir, Ritonavir, Lopinavir, Ivermectin, Interferon and several other immunomodulators and anti-inflammatory drugs. However, most of them either had serious side effects or did not seem to affect the virus significantly (Toor et al. 2021; Hall and Ji 2020; Khanna et al. 2021; Martinez 2021; Yadav et al. 2021). In parallel, research into herbal cures (phytocompounds) also grew rapidly due to their minimal adverse effects and widespread acceptance (Basu et al. 2020; Pk et al. 2020). Furthermore, the use of plant extracts in traditional medicine and novel drugs have been productive multiple times over the past few centuries (Hakobyan et al. 2016; Rolta et al. 2021).

With the growing number of new variants, it has become necessary to analyse the differential effects of the available therapeutics and to assess their efficacy on each variant of SARS-CoV-2. Therefore, a rapid and cost effective in silico method to identify existing molecules or phytocompounds could help in building a repository for clinical trials. Moreover, it would decrease the time for discovery of new drug candidates and may provide significant help for drug development that can be interpreted into clinical applications to combat SARS-CoV-2 (Dotolo et al. 2021; Rudrapal and J. Khairnar S et al. 2020).

In this study, various drugs and phytocompounds have been selected for molecular docking with Spike glycoprotein, a trimeric club shaped structural protein of the SARS-CoV-2 virus, which facilitates viral fusion with the host cell (Chugh et al. 2022). As Spike gene has a nucleotide mutation rate of 8.066 × 10–4 substitutions per site per year, while the SARS-CoV-2 genome has a rate of 6.677 × 10–4 substitutions per site per year, so certain mutations, particularly in the Spike glycoprotein, have been believed to enhance viral infectivity and transmissibility, resulting in emergence of several variants classified as Variants of Concern (VOC) and Variants of Interest (VOI) by the World Health Organization (Hu et al. 2021; Wang et al. 2020; Singh et al. 2021a, b, c, d). Hence, in this comparative study, molecular docking was performed with Spike glycoprotein from 13 variants of SARS-CoV-2. This may prove to be beneficial in curing COVID-19 as the best phytocompound and drug, effective on all the variants have been compared and analysed using in silico studies.

Materials and methods

Ligand preparation

Based on previous studies on RNA viruses, 43 drugs and 35 phytocompounds were selected for virtual screening and molecular docking study against SARS-CoV-2 Spike glycoprotein (Toor et al. 2021; Hall and Ji 2020; Khanna et al. 2021; Poratti and Marzaro 2019). The 3-dimensional structures of all the ligands were retrieved from DrugBank (https://go.drugbank.com/k) and PubChem database (https://pubchem.ncbi.nlm.nih.gov/). Further, the files were converted into protein data bank (PDB) format using OpenBabel and all the selected ligands were prepared using AutoDock Vina 1.5.6 tools. Table 1 shows the selected 78 drugs and phytocompounds (see Table 2).

Table 1 List of 46 drugs and 36 phytocompounds
Full size table
Table 2 Summary of SARS-CoV-2 variants
Full size table

Retrieval of Spike sequences of SARS-CoV-2 variants

The crystal structures of Spike glycoprotein region for 4 different SARS-CoV-2 variants were downloaded from the protein data bank (Wild Type (6VYB) (Walls et al. 2020), Alpha (7LWT) (Gobeil et al. 2021), Beta (7LYQ) (Gobeil et al. 2021), Omicron (7QO7) (Ni et al. 2021)) (https://www.rcsb.org/). Remaining 9 were modelled using Swiss Model based on their amino acid sequences in FASTA format (Gamma (MW642248), Delta (QWO57033), Zeta (QVE55301.1), Iota (QTP80309.1), Theta (QVR41797.1), Epsilon (QPJ72086.1), Eta (QWO17721.1), Kappa (QTY54081.1), Delta Plus (QWS06686.1)), retrieved from NCBI (National Center for Biotechnology Information) virus (https://www.ncbi.nlm.nih.gov/). The information about all the selected variants and their mutations is summarized in Supplementary Table 2 and Table S1, respectively. The Ramachandran plots of the strains are shown in Supplementary Fig. S1.

Target protein preparation

The AutoDock Vina tool 1.5.6 was used for preparation of the protein structures. The binding site for protein–ligand interaction of the target Spike protein from different variants were determined through grid box generation by adjusting the grid parameter x, y, z coordinates value. The grid values of all the 13 variants are provided in Supplementary Table S2.

Virtual screening and molecular docking

Molecular docking study of the selected ligands (43 drugs and 35 phytocompounds) against Spike protein of 13 variants was done using AutoDock Vina tool 1.5.6, following the protocol described by Verma et al. (Verma et al. 2021). After the completion of the docking search, the best conformation with the lowest docked energy was chosen and the protein–ligand complex was analyzed using Discovery Studio (https://discover.3ds.com/d) to examine the list of interactions within the complex. Following that, the suitable compounds were selected for ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) analysis.

ADMET prediction of drugs and phytocompounds

ADMET screening was done to determine the absorption, toxicity, and drug-likeliness properties of ligands (Dong et al. 2018). The 3D structures of ligands were uploaded on SWISSADME (Molecular Modeling Group of the Swiss Institute of Bioinformatics), Molinspiration cheminformatics (a spin-off of Bratislava University), and ProTox-II (Prediction of TOXicity of chemicals) web servers (Charite University of Medicine, Institute for Physiology, Structural Bioinformatics Group, Berlin, Germany) for ADMET screening. ProTox-II web server was used to predict toxicity profile of the chemical (http://tox.charite.de/protox_II) (Singh et al. 2021a, b, c, d; Banerjee et al. 2018). The toxicity of a ligand is measured in terms of toxicity endpoints such as mutagenicity, carcinogenicity, etc. It can also be measured both quantitatively such as LD50 (lethal dose) values, where Class I (LD50 ≤ 5) and II (5 < LD50 ≤ 50) are considered fatal if swallowed and Class VI (LD50 > 5000) is non-toxic, and qualitatively, such as binary (active or inactive) for certain cell types and assays or indication area such as cytotoxicity, immunotoxicity and hepatotoxicity (Parasuraman 2011).

Molecular descriptors and drug likeliness properties of compounds were analyzed using the tool Molinspiration server (http://www.molinspiration.com), based on Lipinski’s Rules of five (Frey and Bird 2011).

Molecular dynamics simulation

MD simulation was done with GROMACS 2018.3 (simulation in duplicate) (Abraham et al. 2015) software which was installed in ubuntu 18.04 LTS, to study the stability of protein–ligand complexes over the period of 100 ns. Docked structures of the protein–ligand complexes (Apigenin with Delta plus Variant) were used in the simulation study. The target protein was processed and the topology file was prepared using pdb2gmx and GROMOS54a7_atb. Force field was downloaded from the automated topology builder website and incorporated into GROMACS. The ligand topology file was prepared using the automated topology builder (ATB) version 3.0. The solvent addition was done in a cubic box using a box distance 1.0 nm from closest atom in the protein. To neutralise the device, the Cl- ions calculated from the genion module for each protein were used. The energy was minimised using the steepest descent algorithm with 50,000 steps and a cumulative force of 5 kJ mol-1, as well as the Verlet cut-off scheme with Particle Mesh Ewald (PME) columbic interactions. During the equilibration process, position restraints were used. Following that, NVT equilibration was performed at 300 K with 100 ps in 50,000 steps, using the leapfrog integrator and NPT equilibration was performed with Parrinello-Rahman (pressure coupling), 1 bar reference pressure, and 100 ps in 50,000 steps. The LINCS algorithm was used to constrain the length of all bonds. For long-range electrostatics, the Particle-mesh Ewald (PME) algorithm was used. The protein–ligand complex's MD was run for 100 ns (in duplicate). Following efficient completion of Molecular dynamic simulation, the root mean square deviation (RMSD) of backbone residues, the number of hydrogen bonds, root mean square fluctuations (RMSF), Radius of gyration (ROG) & Solvent accessible surface area (SASA) were calculated (Verma et al. 2021; Darden et al. 1999; Hess et al. 1997; Páll et al. 2015).

Estimation of free energy of binding

The free energy of binding calculation was done using the standalone program, G-MMPBSA (Rolta et al. 2021; Egan et al. 2000; Kumari and Kumar 2014; Kushwaha and Kaur 2021; Pant et al. 2020) based on the molecular mechanics Poisson-Boltzmann surface area (MM-PBSA) method. The average binding energy calculations were done by a python script provided in G-MMPBSA program.

Results

Molecular docking of drugs and phytocompounds with different variants of SARS-CoV-2

A total of 35 phytocompounds and 43 drugs were used for molecular docking with the Spike protein of all 13 SARS-CoV-2 variants (Table 1). On the basis of their binding affinities with all the variants, 10 drugs were chosen, namely Dpnh (NADH), Flavin Adenine Dinucleotide (FAD) Adeflavin, Liquiritin, Glycyrrhizic acid, Raltegravir, Ritonavir, Doxycycline, Ivermectin, Abemaciclib and Nafamostat (Table 3), out of which Liquiritin showed comparable affinities with all the 13 variants (between -7.0 and -8.1 kcal/mol). Similarly, 15 phytocompounds were chosen, namely, Emodin, Artemisinin, Aloe-emodin, Anthrarufin, Alizarine, Dantron, Rhein, Cucurbitacin B, Apigenin, Curcumin, Fisetin, Quercetin, Isorhamnetin, Genistein and Luteolin (Table 4), out of which Apigenin showed similar binding affinities for all variants (between −6.8 and −7.3 kcal/mol). List of chosen drugs and phytocompounds with their respective Accession Numbers/Pubchem ID is summarized in supplementary Table S3.

Table 3 Docking score of drugs with 13 variants of SARS-CoV-2
Full size table
Table 4 Docking score of phytocompounds with 13 variants of SARS-CoV-2
Full size table

The binding affinity of all 35 phytocompounds and 43 drugs along with their interacting amino acids were visualised using Discovery Studio as mentioned in supplementary Table S4.

Toxicity prediction of drugs and phytocompounds

The 10 drugs and 15 phytocompounds were analysed by Molinspiration, Protox II and SWISSADME to check for Lipinski’s rule, toxicity and ADME respectively. ADME data showed that most of the selected drugs were water-soluble, but only a few had significant GI absorption as shown in Table 5. In the case of phytocompounds, all of them showed good water solubility and high GI absorption except Cucurbitacin B (Table 5).

Table 5 ADME prediction of drugs and phytocompounds by swiss ADME server
Full size table

Toxicity data generated using the Protox II online server showed that among all the drugs, Ivermectin and Raltegravir are Class II and III drugs respectively, while the other drugs belong to Class IV–VI. As for phytocompounds, Cucurbitacin B is a Class II drug, Fisetin and Quercetin are Class III drugs, and the others are categorised as Class IV-VI drugs. Also, Apigenin and Genistein are both non-toxic and each has an LD50 value of 2500. Toxicity data of drugs and phytocompounds are summarised in Table 6. Drug likeliness estimation of active drugs and phytocompounds was done by Molinspiration online server. According to, in-silico druglikeliness prediction Liquiritin showed zero violoations; while Raltegravir, Doxycycline, Nafamostat, Abemaciclib, were found to violate only one of the rules making them suitable candidates for further analysis. On the contrary, in phytocompounds only Cucurbitacin B showed 1 violation, whereas the other phytocompounds followed all rules of drug-likeliness data as summarised in Table 7.

Table 6 Toxicity prediction of antiviral drugs and phytocompounds using Protox II server
Full size table
Table 7 Drug likeliness prediction of antiviral drugs and phytocompounds
Full size table

Based on our comparative study, Liquiritin (between −7.0 to −8.1 kcal/mol) and Apigenin (between −6.8 and −7.3 kcal/mol) passed the toxicity prediction, drug likeliness and also have a consistent binding affinity to each of the the 13 variants (Tables 3, 4).

Liquiritin showed hydrogen bonding with Thr300, Ser50, Asn315, Arg317, Gln626 and hydrophobic interactions with Cys299, Ala290, Cys289, Glu296, Lys302, Thr628, Thr272, Ser314, Trp631, Gln319 in Delta variant; and in case of Delta plus variant it makes hydrogen bonds with Gln626, Leu627, Ser314, Thr300, Thr272, Ser50 and hydrophobic interactions with Arg271, Cys299, Ala290, Thr272, Thr628, Pro629, Glu296, Lys302, Cys289. Similarly, Apigenin made hydrogen bonds with Arg1012 and hydrophobic interaction with Thr959, Ala956, Tyr1005, Leu960, Ser1001, Gln963, Thr1004, Gln1008, Gln952, Gln955 in delta strains in case of delta plus apigenin showed only hydrophobic interactions with Thr 959, Gln 1008, Gln 952, Gln 955, Ala 956, Arg 1012, Tyr 1005, Leu 960, Ser 1001, Gln 963, Thr 1004 amino acids the most important variants, are summarized in Table 8 and Figs. 1, 2,3 and 4.

Table 8 Interactions of apigenin and liquiritin with delta and delta plus variants of SARS-CoV-2
Full size table
Fig. 1
figure 1

Interactions of Apigenin with delta variant of SARS-CoV-2 Variant: in close view of delta in complex with Apigenin, purple colour is showing target protein, green colour is showing hydrophobic interactions, yellow colour is showing hydrogen bonding and red colour ligand

Full size image
Fig. 2
figure 2

Interactions of Apigenin with Delta plus variant of SARS-CoV-2 Variant: in close view of Delta plus in complex with Apigenin, purple colour is showing target protein, green colour is showing hydrophobic interactions, yellow colour is showing hydrogen bonding and red colour ligand

Full size image
Fig. 3
figure 3

Interactions of Liquiritin with Delta variant of SARS-CoV-2 Variant: in close view of Delta in complex with Liquiritin, purple colour is showing target protein, green colour is showing hydrophobic interactions, yellow colour is showing hydrogen bonding and red colour ligand

Full size image
Fig. 4
figure 4

Interactions of Liquiritin with Delta plus variant of SARS-CoV-2 Variant: in close view of delta plus in complex with Liquiritin, purple colour is showing target protein, green colour is showing hydrophobic interactions, yellow colour is showing hydrogen bonding and red colour ligand

Full size image

Furthermore, to study the stability of active Apigenin phytocompound, MD simulation for 100 ns was performed.

MD simulation of Apigenin with Delta plus mutant of SARS-CoV-2

MD simulation of Apigenin in complex with Delta plus variant of SARS-CoV-2 for 100 ns was performed to study the stability of protein–ligand complexes. MD simulation data revealed that RMSD of Apigenin, complexed with Delta plus variant of SARS-CoV-2 was stable from the start of the simulation and remained stable upto 100 ns time (Fig. 5A). RMSF of protein–ligand complex was done to study the flexibility and fluctuation in interactive residues in secondary structure of target proteins (Sivaramakrishnan et al. 2020; Kumar et al. 2014).The RMSF plot for Apigenin fit over the Delta plus protein and showed less residual fluctuation in alpha helical and beta strands. Residues ranging from 100 to 300, 730, 900 to 1150 showed the strongest interactions with Apigenin as shown in Fig. 5B. Binding free energy of protein–ligand complexes is composed of Van der Waals energy −145.285 ± 14.315 kJ/mol, Electrostatic energy −7.358 ± 6.263 kJ/mol, Polar solvation energy 60.148 ± 15.417 kJ/mol, SASA energy −14.129 ± 1.345 kJ/mol and Binding energy −106.624 ± 11.965 kJ/mol (Fig. 6).

Fig. 5
figure 5

RMSD and RMSF graph of Apigenin with delta plus variant of SARS-CoV-2 variants for 100 ns: A RMSD and B RMSF

Full size image
Fig. 6
figure 6

Breakdown of free energy of binding estimated with gMMPBSA

Full size image

Hydrogen bond interactions of protein–ligand complexes are shown in Fig. 7A. The Radius of gyration in the range of Apigenin in complex with Delta plus is 4.4–4.9 nm, as shown in Fig. 7B. The Radius of gyration plot establishes the compactness of the Apigenin and Delta plus protein complex and confirms their stability. Solvent accessible range of Apigenin complexed with Delta plus protein is between 650 and 560 nm2; as shown in Fig. 7C.

Fig. 7
figure 7

A Hydrogen bond interactions of ligand with protein, B Radius of gyration (ROG) and, C Solvent accessible surface area

Full size image

Discussion

The mutations in the RBD region of the 18 amino acid long SARS-CoV-2 Spike glycoprotein strengthen the virus's capacity for transmission. To understand the genesis of novel variants, research has focused on the Spike glycoprotein. The Spike protein's RBD region mutations can make closer contact with hACE2, which results in a stronger binding affinity and probably enhanced VOCs infectivity (Chugh et al. 2022). Among all the previously circulating VOCs, the Delta variant, has shown adverse effects on patients and has caused twice as many hospitalizations (Edara et al. 2021). The Delta was found to be 60% more transmissible than the highly infectious Alpha variant identified in the United Kingdom in September 2020 (Duong 2021). The strain undoubtedly contributed to India's massive second wave of cases. According to data available on GISAID, it had spread to 208 countries as of December 19, 2022 (https://gisaid.org/hcov19-variants/). Delta plus, also known as the AY.1 strain which showed a rapid spread, was found to bind easily to the ACE-2 receptor, and was potentially resistant to monoclonal antibody therapy (Roy and Roy 2021). As per GISAID database, apart from these variants, the Omicron variant, which was discovered in Botswana, has now spread to 208 countries as of December 19, 2022 (https://gisaid.org/hcov19-variants/). In regional genomic surveillance, XBB, a recombinant of the BA.2.10.1 and BA.2.75 sublineages, has been reported in 35 countries with a global prevalence of 1.3%. The regional immunological landscape and COVID-19 vaccination rates appear to have an impact on establishing whether the increased immune escape of XBB is sufficient to cause new infection waves (Kurhade et al. 2022). Hence, repurposing existing drugs against potential targets of the virus could be an effective strategy to speed up the drug discovery process (Bhardwaj et al. 2021a, b, c).

Molecular docking facilitates the prediction of protein–ligand affinity and the structure of the protein–ligand complex. Additionally, it can be used to investigate the binding difference between the two molecules, which is useful information for lead optimization. In the early stages of the pandemic, Ivermectin was considered as a viable therapeutic drug against SARS-CoV-2. Despite the fact that it violated Lipinski's rule and was immunotoxic, being FDA approved for other viral infections, repurposing of this medicine became a ray of hope. It showed strongest affinity with the majority of SARS-CoV-2 variants (as justified in our study Table 3) and was found to minimise the probability of mortality in COVID-19 (Bryant et al. 2021; Caly et al. 2020; Krolewiecki et al. 2021; Zaidi and Dehgani-Mobaraki 2022; Mastrangelo et al. 2012). Also, Australia’s National COVID-19 Clinical Evidence Taskforce and the World Health Organization suggested the use of Ivermectin only in clinical trials (FAQs 2022). Later on, a review of 10 randomised controlled trials by Roman et al. concluded that Ivermectin is not a viable option for the treatment of COVID-19 patients (Roman et al. 2022). Consequently, it became a weak contender.

Since ancient times compounds extracted from traditional medicinal plants with strong antiviral activity have been used to treat viral infections. It has been found that phytocompounds can inactivate SARS-CoV-2 variants by binding to the Spike glycoprotein and thus inhibit their function like Curcumin, a component of turmeric (Curcuma longa), is believed to have potential properties to prevent or treat diseases such as cancer and viral infections (Manoharan et al. 2020; Rattis et al. 2021; Singh et al. 2021). Artemisinin and Emodin have also been found to interact with SARS-CoV-2 and inhibit its Spike glycoprotein (Rolta et al. 2021; Nair et al. 2021; Sehailia and Chemat 2021).

All this led to investigation of binding affinity of drugs as well as phytocompounds with the Spike glycoprotein of SARS-CoV-2 using molecular docking. Results from our study show that phytocompounds exhibit the binding affinity as high as drugs. Also, Sathya et al. has reported the promising results of Liquirtin against H1N1 and H3N2 influenza A virus which further confirms its anti-viral drug property and makes it a competitive candidate for the treatment of COVID-19 (Sathya et al. 2020). Zhu et al. also proposed that Liquiritin mimics Type I IFN, which inhibits viral replication (Zhu et al. 2020). Our study also suggested Liquirtin as one of the promising drugs, as it exhibits high and uniform binding affinity with the Spike glycoprotein of all 13 variants (between −7.0 and −8.1 kcal/mol). Although it was found to be immuno-toxic, zero violations of Lipinski's Rule make it a candidate for research.

Similarly various studies have attempted to carry out in silico validation of phytocompounds to cure various diseases (Rolta et al. 2021; Mehta et al. 2021; Salaria et al. 2022). Rolta et al. 2021 (Rolta et al. 2021) also reported that phytocompounds (emodin, aloe-emodin, anthrarufin, alizarine, and dantron) of R. emodias inhibitor of nucleocapsid phosphoprotein of SARS-CoV-2. Some of the bioactive molecules from tea have also shown promising binding affinities with other proteins of SARS-CoV-2, some of them being NSP15, NSP16 and Mpro (Main protease)(Bhardwaj et al. 2021a, b, c; Singh et al. 2021a, b, c, d; Sharma et al. 2021; Chauhan et al. 2022. Bhardwaj et al. 2021). Hakobyan et al. demonstrated the in vitro effect of Apigenin on African swine fever virus infection by interfering with the viral cell cycle at an early stage in their study, implying that Apigenin could be an effective candidate for extended in vitro and in vivo studies combining dosage effectivity (Hakobyan et al. 2016). In present analysis as well Apigenin expressed the strongest and most consistent binding affinity with all strains (between -6.8 and -7.3 kcal/mol). Additionally, Apigenin exhibited no toxicity and zero violations of Lipinski’s rule.

Multiobjective optimisation in drug discovery field implies that a drug should be potent in being active, non-toxic, orally bioavailable, free of side effects, with strong binding affinity, GI absorption (Thomford et al. 2018; Lambrinidis and Tsantili-Kakoulidou 2021). These parameters aid in the screening and recommendation of the prospective drug candidates for in vitro and in vivo studies. Considering all the important parameters, we propose that Apigenin and Liquiritin could be promising options for the treatment of COVID-19 and that they should be investigated further in vitro and in vivo to see if they can be used to build therapeutic strategies to combat future SARS-CoV-2 peaks.

Conclusion

It is imperative that the drugs and phytocompounds not only pass the toxicity prediction and drug likeliness, but they should also have a consistent binding affinity to all the variants. In the present study, 43 drugs and 35 phytocompounds candidates with potential inhibitory effects towards Spike glycoprotein of SARS-CoV-2 were chosen to perform molecular docking studies. Based on our comparative binding affinity analysis, ADMET analysis and druglikeliness profile we have shortlisted Liquiritin (among the repurposing drugs) and Apigenin (among the phytocompounds). MD simulation results confirmed the stability of Apigenin with Delta plus variant. The consistent binding affinities of repurposing drugs and phytocompounds with all the existing variants of SARS-CoV-2 indicates that these maybe effective universally against upcoming variants as well, thus making it one of the largest comparative studies.

Data availablility

All data generated or analysed during this study are included in this published article (and its supplementary information files).