The alga Ui were collected from the Olaikuda area (Gulf of Mannar) situated near North Mandapam, Rameswaram, Tamil Nadu, India, with the help of the Central Marine Fisheries Research Institute, Mandapam, and Rajendra Kumar Algae Project Center, Mandapam. The algal sample was washed thoroughly with water to remove dirt and debris and packed safely in polythene zip-lock bags. Upon reaching the laboratory it was dried using a tray drier (Figures 1A,C), mainly to concentrate the extract, preserve the hydrolabile compounds, and prevent the growth of bacteria and mold.

Phytochemical extraction was performed by Soxhlet extraction. The dried sample (∼60 g) was pulverized using a mortar and pestle (Figures 1B,D), and transferred into a thimble in the extraction tube. The extraction solvent used was 95% ethanol (100 ml). The all-glass Soxhlet apparatus was set up according to the standard protocol and was run for 6 h at 78°C using an isomantle. The extract was analyzed for the phytochemicals using a 7890B GC coupled with a 5977A mass selective detector (MSD). The chromatographic column used for GC was HP-5MS of dimensions 30 m × 250 μm × 0.25 μm (length, inner diameter, and film thickness, respectively). It is a bonded, cross-linked, and solvent-rinsable non-polar column made of (5%-phenyl)-methylpolysiloxane, with a capillary tubing made of fused silica (Agilent Technologies, Santa Clara, CA). The volume of the sample injected was 1 μl and the flow rate of the carrier gas (helium) was 1.0 ml.min⁻¹ with a split ratio of 1:1. The injection port temperature was 250°C. The system started with a 2 min-hold at 50°C, then ramped 3°C per minute until the temperature reached 270°C. The system was on hold at this temperature for 20 min. Simultaneously, the separated samples were fed automatically to the MSD at an interface temperature of 280°C. The electron ionization was performed at 70 eV, and the scan range of the system was 40–700 m/z. The total run time of the process was 95 min. The retention indices of the compounds were determined relative to trichloromethane, the standard compound selected for data analysis. Further, the compounds were identified by comparing their mass spectra with the data in NIST-14 Mass Spectral Data Library.

The three-dimensional chemical structures of the identified phytochemicals were obtained from PubChem (https://pubchem.ncbi.nlm.nih.gov/). These were then saved as SDF files. The energy minimization and format conversion of these structures were performed in PyRx software (Dallakyan and Olson 2015). The default energy minimization parameters were the universal force field and the conjugate gradient algorithm. Once energy minimization was completed, the structures were rewritten as PDBQT files. The target protein used in this study was S₁ receptor binding domain of the spike (S) glycoprotein. The three-dimensional structure of RBD was retrieved from a complex of ACE2 and RBD (PDB ID: 6M0J) from the Protein Data Bank (RCSB-PDB; https://www.rcsb.org/). As the first step, the optimization of protein structures was performed using AutoDock Tools by deleting chain A, water molecules, and co-crystal ligands. The missing atoms were then repaired, and polar hydrogens were added. Charges were distributed and minimized over the protein structure. The structure was then saved in PDBQT format.

An active site is defined as a groove or pocket of an enzymatic or non-enzymatic protein which facilitates ligand binding or biochemical reactions (Pravda et al., 2014). The characteristics of the active site are mainly determined by the active site residues (Srinivasan, 2020), and various studies have characterized the possible active site residues of RBD of S₁ subunit of spike protein (Figure 2). Tyr449, Tyr453, Arg454, Lys458, Ser459, Ser469, Glu471, Phe486, Asn487, Tyr489, Leu492, Gln493, Gly496, Gln498, Thr500, Asn501, Gly502, and Tyr505 were the reported active site residues (Lan et al., 2020; Kulkarni et al., 2020; Prajapat et al., 2020). These residues were further validated using the ‘Zone’ function in UCSF Chimera software (https://www.cgl.ucsf.edu/chimera/). The zone parameter was set to “<5.0 Å from currently selected atoms” (Ashraf et al., 2014), where the currently selected atoms were the atoms of chain A. The mean of the X, Y, and Z coordinates of the final atom of each interacting residue highlighted by UCSF Chimera was calculated and applied as the dimension of the grid-box center. The grid size was manually adjusted to cover the interacting residues. Further, the values of these coordinates were saved as a configuration text file which was later used for docking.

Molecular docking is an in silico approach which is used to predict the conformational binding energy of ligands to a preferred target using matching and scoring algorithms (Leach et al., 2006). In this experiment, we have used AutoDockVina (Trott and Olson, 2010) in PyRx software as the docking tool, The optimal binding energy of the ligands was obtained based on least root mean square deviation (RMSD) for each conformers of a particular ligand, and arranged in ascending order to select the best ligand(s) for further calculating the chemical behaviour using C-DFT and pharmacokinetic analyses. PyMOL (https://pymol.org/), an open-source molecular visualization software was used to identify the polar contacts (H-bonds) between the ligand and the interacting active site residue, and develop printable figures of this interaction. To analyze hydrophobic interactions between the ligand and residues, another visualization software, BIOVIA Discovery Studio Client 2020 (https://discover.3ds.com/discovery-studio-visualizer-download) was used.

Conceptual Density-functional theory (C-DFT) is a computational method to predict chemical behaviour of the compounds (Poater et al., 2010; Domingo et al., 2016). Density-functional theory(DFT) has been developed from Hohenberg-Kohn theorem, which is an in-silico quantum mechanical modeling strategy used to determine the properties of a many-electron systems, using spatially-dependent electron density functionals (Hohenberg and Kohn, 1964; Kohn and Sham, 1965). C-DFT, a sub-field of DFT, helps to analyze the molecular orbital energies of conformers and can give rise to cues for understanding the structure-activity relationship of the molecule (Parr and Yang, 1989; Geerlings et al., 2003; Sarkar and Chattaraj, 2021a; Sarkar and Chattaraj, 2021b). To describe the orbital properties of a molecule, ten different molecular descriptors, known as the global reactivity descriptors and its derivatives, were considered viz. total energy (E_γ; in eV), molecular dipole moment (D_p; in Debye units), the energy of the lowest unoccupied molecular orbit (LUMO) (E_LUMO; in eV), the energy of the highest occupied molecular orbit (HOMO) (E_HOMO; in eV), energy gap (ΔE; in eV), absolute hardness (η; in eV), global softness (σ; in eV⁻¹), electronegativity (χ), chemical potential (μ; in eV), and global electrophilicity index (ψ; in eV⁻¹) (Chattaraj et al., 2003; Chattaraj et al., 2006). These molecular descriptors are calculated based on the electron density of molecules using Fukui’s molecular orbital theory (Fukui 1982; Ayers and Parr, 2000). E_LUMO and E_HOMO are the primary and the most important descriptors which determine the ability of a molecule to accept or donate electrons. D_p is the measure of the total polarity of a system. It is also a positive indicator of the reactivity of the molecule. It was found that the higher the D_p, the greater the reactivity of the molecule (Roy et al., 2006; Mert et al., 2011). The derived descriptors of E_LUMO and E_HOMO are ΔE, η, σ, χ, μ, and ψ, which also account for the ability of the molecule to interact and contribute to electron sharing or transfer with the target by transiting from HOMO to LUMO. For example, if ΔE is found to be less, the molecule can easily transit from HOMO to LUMO (Chattaraj and Roy, 2007; Bostan et al., 2012). It represents the chemical reactivity and kinetic stability of the molecule; if χ is found to be less, the inhibitory effect of the ligand is higher (Zhan et al., 2003). As the first step in determining these descriptors, the selected ligands were optimized using the Becke-3-parameter, Lee-Yang-Parr (B3LYP) function (Becke 1988; Lee et al., 1988) with 6-311G(2d, p) basis set in Gaussian-16 software (http://gaussian.com/gaussian16/) (Frisch et al., 2016). B3LYP is the most popular functional used in molecular quantum mechanical modeling and is derived from a defined set of atomic/molecular energies and potentials.

The drug-likeliness and pharmacokinetic properties such as Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) of the selected ligands were predicted. The Drug Likeliness Tool (DruLiTo; http://www.niper.gov.in/pi_dev_tools/DruLiToWeb/DruLiTo_index.html), an open-source drug-likeness software developed by the Department of Pharmacoinformatics, National Institute of Pharmaceutical Education and Research (NIPER), Punjab, India was used to analyze drug likeliness by checking whether the ligands violate any of Lipinski’s Rule of Five (RO5), or would pass the Ghose and Veber filters. A reliable online tool for pharmacokinetic predictions of small molecules, pkCSM (http://biosig.unimelb.edu.au/pkcsm/), was used to predict the ADMET properties of the ligands (Pires et al., 2015), in which the canonical or isomeric SMILES of the ligands from Pub Chem were given as input.

The GC-MS data of the Ui ethanolic extract showed 55 peaks (Figure 3), and on comparison with NIST-14 library, 43 known phytochemicals were identified (Table 1). The phytochemical class analysis revealed that 18 phytochemicals were simple carboxylic acids, fatty acids, or their derivatives (palmitic acid, HIP, methylpalmitate, ethylpalmitate, butanoic acid, paullinic acid, doconexent, allyl stearate, ethyllinolelaidate, ethyllinolenate, ethylelaidate, icosapent, MHDTE, BOD4E, BOD3E, BTES, DPPP, and propyllinoleate), seven belonged to terpenoid class (damascene, cyclosativene, dihydroactinolide, 3-DOCH, phytol, CMBA, and oxymesterone), three (6.98%) each were aldehydes and its derivatives (methylglyoxal, furfural, and retinal), alcohols and its derivatives (TAA, 1-heptatriacotanol, and DTD), alkene hydrocarbons and its derivatives (cetene, 8-heptadecene, and 9-octadecene), and alkane hydrocarbons and its derivatives (myristyl chloride, TMHA, and EEBOD), two were monoglycerides (2-monopalmitin and 1-monolinolein), and one each were an organosulfur compound (DMSO), an aromatic hydrocarbon (azulene), a phenol (2,4-DtBP), and a glycoside (RBGUL). The peak corresponding to HIP showed the highest signal abundance of >2.8×10⁷, however, the mean relative peak area of phytol (21.404%) was found to be the widest, followed by 2-monopalmitin, 9-octadecene, palmitic acid, and other compounds. The details of the GC-MS analysis such as peak number(s), retention time(s), and mean relative peak area are presented in Table 1.

Hydroxychloroquine, the control ligand, showed a binding affinity of −5.7 kcal.mol⁻¹ with the optimized structure of RBD. Twenty-one (48.84%) compounds had binding energies ranging from −4.0 kcal.mol⁻¹ to −4.8 kcal.mol⁻¹. Out of the 43 compounds, only 16 were considered for studying their molecular interaction (Tables 2, 3). Interaction analysis revealed that furfural had three hydrogen bonds interacting with Arg454, Ser469, and Glu471, but its binding energy was -3.8 kcal mol⁻¹. Considering hydrophobic interactions, icosapent interacted with Arg403, Tyr453, Tyr495, Phe497, and Tyr505. The binding energy of this molecule was −4.8 kcal.mol⁻¹. Out of these 16 compounds, only the best five compounds (2,4-DtBP, doconexent, DTD, RBGUL, and retinal) were considered for C-DFT, drug-likeliness studies using DruLiTo, and ADMET properties using pkCSM. The criteria used for this selection was mainly their relative lower binding energy. The conformations were visualized using PyMOL software and depicted in Figures 4–9.

The molecular descriptors were calculated after optimization, based on the FMO theory (Table 4). The total energy of the compounds is the total electron energy of the ground state. Lower the total energy, higher is their stability. RBGUL displayed the lowest total energy with value −41.84 × 10³ eV. Molecular orbital energies such as HOMO energy (E_HOMO) and LUMO energy (E_LUMO) were calculated and analyzed (Table 5). Retinal showed the least energy gap with an energy difference of 3.04 eV. The energy gap of RBGUL (ΔE = 3.20 eV) was also found to be close enough to that of retinal. The maximum D_p was also shown by retinal (D_p = 6.33 Debye units). Considering derived descriptors, the most electronegative compound in the selected list was retinal (χ = 3.82). The electronegativity of RBGUL (χ = 3.67) was found to be highly similar to that of retinal. Absolute hardness and Global softness are criterions of overall stability of the system and also they are supporting parameters of electronegativity. In our study Retinal and RBGUL showed acceptable values of absolute hardness, 1.52 and 1.60 and softness, 0.33 and 0.31, respectively. Chemical potential of compounds is the negative value of electronegativity values, which is also an indication of high chemical activity. Therefore in this case too, retinal and RGBUL exhibited high chemical potential. High electrophilicity of retinal (4.80) and RBGUL (4.21) suggests their elevated likeliness to accept electrons. According to the above findings, RBGUL, and retinal were considered good inhibitors of S₁ RBD of SARS-CoV-2.

The drug-likeliness prediction from DruLiTo and ADMET results from pkCSM are presented in Table 6. Evaluation of drug-likeliness showed that 2,4-DtBP satisfied and passed through the Lipinski’s RO5, Ghose, and Veber filters, whereas other ligands violated atleast one of the three parameters. Absorption properties revealed that all ligands were readily absorbed intestinally. 2,4-DtBP, doconexent, DTD, and retinal showed no interference with the P-glycoprotein system, however, RBGUL was found to be both a substrate and an inhibitor in the system. Skin permeability prediction showed that 2,4-DtBP was slightly permeable. Distribution properties showed that these compounds have tendencies to cross the blood-brain barrier (BBB) and central nervous system (CNS). Metabolic properties revealed that no ligand escaped the cytochrome P450 (CYP) system of the liver completely. Amongst the five selected ligands, DTD and RBGUL showed minimum interference with the system (acted as CYP2C19 inhibitor and CYP3A4 substrate, respectively). Considering excretion and toxicity properties, no ligand acted as renal OCT2 substrate, and human ether-à-go-go-related gene (hERG)-I protein inhibitors. The compounds passed the Ames toxicity test, indicating their inability to be a mutagen and thus a carcinogen. However, hepatotoxicity was predicted with doconexent, RBGUL, and retinal. Except for RBGUL, all other selected ligands showed skin sensitization too.