Quantum chemical studies on the binding domain of SARS-CoV-2 S-protein: human ACE2 interface complex

A two-layer ONIOM(B3LYP/6-31G (cid:1) :PM7) method is used to model the binding of several drug/drug-like molecules ( L ) at the SARS-CoV-2S-protein: human ACE2 protein interface cavity. The selected molecules include a set of thirty-five ligands from the study of Smith and Smith which showed a high docking score in the range of (cid:3) 7.0 to (cid:3) 7.7kcal/mol and another set of seven repurposing drugs, viz. favipiravir, remdesivir, EIDD, galidesivir, triazavirin, ruxolitinib, and baricitinib. The ONIOM model of the cavity ( M ) showed a highly polarized electron distribution along its top-to-bottom direction while L s with lengths in the range 1.0 (cid:3) 1.5nm fitted well inside the cavity in a head-to-tail fashion to yield ML complexes. The ligands showed a large variation in the ONIOM-level binding energy (E b ), in the range (cid:3) 2.7 to (cid:3) 85.4kcal/mol. The E b of ML complexes better than (cid:3) 40.0 kcal/mol is observed for myricetin, fidarestat, protirelin, m-digallic acid, glucogallin, benserazide hydrochlorideseradie, remdesivir, tazobac-tum, sapropterin, nitrofurantoin, quinonoid, pyruvic acid calcium isoniazid, and aspartame, and among them the highest E b (cid:3) 85.4kcal/mol is observed for myricetin. A hydroxy substitution is suggested for the phenyl ring of aspartame to improve its binding behavior at the cavity, and the resulting ligand 43 showed the best E b (cid:3) 84.5kcal/mol. The ONIOM-level study is found to be effective for the interpretation of the noncovalent interactions resulting from residues such as arginine, histidine, tyrosine, lysine, carboxylate, and amide moieties in the active site and suggests rational design strategies for COVID-19 drug development. A quantum chemical treatment of the noncovalently-bonded drug and drug-like molecules at the interface of SARS-CoV2: human ACE2 receptor is described. It predicts that myricetin, fidare-stat, protirelin, m-digallic acid, glucogallin, benserazide hydrochlorideseradie, remdesivir, tazo-bactum, sapropterin, nitrofurantoin, quinonoid, pyruvic acid calcium isoniazid, and aspartame have high binding affinity at the interface and also suggests a design strategy to modify aspar-tame for improved binding affinity.


Introduction
The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), a pathogen responsible for the enduring international epidemic outbreaks of respiratory disease popularly known as the Coronavirus Disease 2019 (COVID-19) has been causing one of the unprecedented deadliest pandemics of the world (Guarner, 2020). Coronaviruses possess pleomorphic enveloped particles containing single-stranded RNA associated with a nucleoprotein within a capsid comprised of matrix protein (Mousavizadeh et al., 2021) which encode four major structural proteins, viz. the spike (S) protein, nucleocapsid (N) protein, membrane (M) protein, and the envelope (E) protein (Belouzard et al., 2012;van Boheemen et al., 2012) shows strong interaction with host cell protein, and has been considered as a first-line therapeutic target for coronavirus vaccine and antiviral therapy development (Li et al., 2003;Mousavizadeh et al., 2021). Drug repurposing (also called drug repositioning, reprofiling, or rediscovery) (Scherman et al., 2020) approaches have been used for accelerating the discovery of already existing drugs that can cure COVID-19. Such approaches lower the risk of failure, shorten the time frame and reduce the cost for drug development in comparison to the de novo drug discovery process (de Oliveira et al., 2021;Pushpakom et al., 2019). Xu et al. (Xu et al., 2020) used the computer-guided homology modeling method to the crystal structure of SARS coronavirus S-protein (PDB accession: 6ACD) to develop the 3-D structural model for SARS-CoV-2. ACE2 is a Type-I membrane protein expressed in various human tissues mainly in the lung, kidney, intestine, and heart (Donoghue et al., 2000;Li et al., 2020). Previous studies revealed that S-protein of SARS-CoV-2 presents a strong binding affinity to the angiotensin-converting enzyme II (ACE2) which acts as a main cell entry receptor for novel corona virus (Li et al., 2003;Prabakaran et al., 2004;Zhou et al., 2020) and this observations spurred extensive research on the interaction of ACE2 with the virus. Xu et al.  used the structural superimposition and molecular rigid docking approaches  to develop the 3-D complex structure of the S-protein binding to human ACE2 ( Figure 1a) and reported the binding free energy À50.6 kcal/mol for the complex. Recently, Smith and Smith (Smith, 2020) ensued the work of Xu et al. to generate the coordinates for the Sprotein-ACE2 complex and used the model to enact an ensemble docking virtual high throughput screening study to identify small molecules that bind to either the isolated Sprotein at its host receptor region or to the S-protein-human ACE2 interface region (PDB ID:2AJF). Based on docking studies on S-protein and the S-protein:ACE2 receptor complexes, they reported binding affinities of over 9000 drugs, metabolites, and natural products (and their isomers) and derived 30 top-scoring compounds on isolated S-protein and 19 potent candidates on S-protein:ACE2 interface. The immense amount of docked structural data obtained using Autodock Vina (Trott et al., 2010) is available from the study of Smith et al. (Smith, 2020) which is beneficial for further research in this area.
Numerous publications have appeared on the modeling of drug-receptor interactions related to SARS-CoV-2 using molecular docking, molecular dynamics simulation, and other computer-aided drug design (CADD) techniques (Baildya et al., 2020;Khelfaoui et al., 2021;Linet et al., 2021;Mebs et al., 2010;Ray et al., 2021;Sahoo et al., 2021). In 2021, Singh et al. studied the efficacy of the pharmaceutical drug hydroxychloroquine and common co-drug azithromycin by using docking calculations with the S-protein:ACE2 interface and predicted the binding affinity À3.7 and À5.2 kcal/mol for hydroxychloroquine and azithromycin, respectively (Singh et al., 2021). Moreover, the effectiveness of the FDAapproved antiviral drugs such as remdesivir and hydrochloroquine has been investigated, and found that both drugs are highly effective in the control of 2019-nCoV infection in vitro . Quinlin et al. developed a proximitybased AlphaLISA assay to measure the binding of SARS-CoV-2 S-protein receptor-binding domain (RBD) to ACE2 and reported 25 high-quality primary hits out of 3384 small-molecule drugs (Hanson et al., 2020). Govinda et al. used REDIAL-2020 machine learning suite for estimating anti-SARS-CoV-2 activities from molecular structure (Bocci et al., 2021). Very recently, Hagar et al. have reported bioactivity of antiviral N-heterocycles such as Favipiravir, Ribavirin, Amodiaquine, and 2 0 -Fluoro-2 0 -deoxycytidine by using the molecular docking and DFT calculations (Hagar et al., 2020).
The present study focuses on understanding the electronic features of the binding region of the S-protein:ACE2 interface model for small molecules using a density functional theory (DFT) method in conjunction with a hybrid ONIOM (Our Own N-layer Integrated molecular Orbital molecular Mechanics) method (Dapprich et al., 1999;Karadakov et al., 2000;Svensson et al., 1996;Vreven et al., 2001) that combines DFT and semi-empirical approaches. Further, the molecules listed by Smith and Smith as the most binding at the interface are selected to create inputs for ONIOM level calculations. The ONIOM results provided non-covalent binding features of the small molecules at the interface. Also they led to a design strategy to identify derivatives having better binding energy. Since the ONIOM method offers a DFT-semi-empirical quantum chemical approach to study the binding domain in the cavity of the protein, a powerful prediction on the quantum chemical binding behavior of the ligands/drugs at the interface is possible which is expected to be more accurate than a docking study.

Computational methods
For the SARS-CoV-2 S-protein: human ACE2 complex (PDB ID: 2AJF) (Figure 1a), Smith and Smith used the interface model shown in Figure 1b, which contains the most significant portions of both S-protein and ACE2 complex for binding with the ligands. This model is further reduced with the inclusion of 611 atoms in the binding domain ( Figure 1c). The reduced model M and the docked structure of M with the ligand L (ML complex) were optimized with a two-layer hybrid QM method, ONIOM, developed by Morokuma and co-workers (Chung et al., 2015), as implemented in Gaussian 16 (M. J. Frisch et al., 2016). Such a procedure has been widely used for the determination of the structure and properties of large molecular systems such as heterogeneous catalysts, nanomaterials, and biological macromolecules. In the ONIOM scheme (Figure 1d), the high layer is treated with the B3LYP/6-31G Ã level DFT (Becke, 1993;Stephens et al., 1994;Tirado-Rives et al., 2008) (DFT layer), and the low layer is treated with the semi-empirical PM7 method (PM7 layer) (Peng et al., 2019;Stewart, 2013). Molecular structure in the high layer is fully optimized under the ONIOM(B3LYP/6-31G Ã :PM7) scheme while with the atoms in the low layer fixed. Freezing the low layer is essential to retain the general shape of the binding cavity. Since low layer is not optimized, the energy of the high layer, which corresponds to the energy at B3LYP/6-31G Ã level is used for the binding energy calculation.

Results and discussion
Reduced interface model (M) for SARS-CoV-2 S-protein: human ACE2 complex: Two views of the ONIOM (B3LYP/6-31G Ã :PM7) level optimized geometry of M is given in Figure 2. The size of the cavity wherein the small molecule can bind is approximately 10.9 Â 10.6 Â 11.1 Å in dimension. Though the total charge of the high layer in ONIOM (sphere model in Figure 2) is zero, it contains three carboxylate (-COO -) moieties, two positively charged arginine moieties (-CN 3 H 5 þ ), and one ammonia (-NH 3 þ ) moiety. The anionic carboxylate moieties lie close to the cationic moieties. The high layer of the cavity also consists of two phenolic moieties from the tyrosine units (both belong to the S-protein), a portion from a histidine unit, and a portion from a proline unit (both from ACE2).
With the ONIOM optimized geometry, a single point DFT calculation at B3LYP/6-31G Ã level is done for M to obtain its electron density distribution. The MESP on planes passing through the middle region of the cavity is shown in Figure  S1 for the view given in Figure 2(a). These pictures show that the bottom region of the cavity where the proline portion appears is electron-rich (blue area) while the top region characterized by histidine portion is electron-deficient (red area). This suggests a highly polar binding domain for the interface. The high dipole moment 44.4 D observed for this binding cavity with a top to the bottom direction ( Figure 2b) indicates its natural tendency to attract polar molecules for strong binding.

Selected small molecules for binding:
The thirty-five top scoring ligands reported by Smith and Smith (Smith, 2020) (Smith, 2020) along with the corresponding popular/approved/scientific name. The common structural motifs representing general backbone for selected molecules of each group for the binding study are shown in Figure 3.

Group 1: O-heterocyclic molecules
All the O-heterocyclic molecules depicted in Figure S2 are THI systems (1 2 11) and they show very close structural similarity. The common structural motif observed in all is depicted in Figure S2 (first entry in column 1) and the notation 6r-66r is used to represent this motif. Here 6r represents a 6-membered ring (phenyl) and 66r represents a fused structure made up of two 6-membered rings while the hyphen indicates a single bond connection between 6r and 66r. The 66r moiety is the O-heterocycle chromone in 1 2 5 systems, 4-chromanone in 6 2 8 systems, and 4-chromanol in 9 and 10 systems. Both 6r and 66r moieties show multiple hydroxy substitutions. Also, 1 2 8, and 11 contain a carbonyl group. 12 (glucogallin), in which a 6-membered ring is connected to a 6-membered ring through three bonds. The symbol " Ù -" is used to denote the connection between the ring moieties through three bonds.

Group 3: N, O-heterocycles
These molecules ( Figure S4) are made up of both N-and Oheterocycles and they show six different structural motifs, viz. 56r (26, 27), 5r-6r (28), 5r Ù -5r (29), 65r Ù 66r (30), 5r$56r (31), and 5r-566r (32). Here the "$" denotes a spiro connection between two rings while 665r notation represents a fused three-ring structure. A common structural motif (56r) is observed in 26 and 27 and their side-chain contains a sugar moiety. The 27 is perhaps the odd one out considering the phosphate-sugar-base connectivity which is typically seen in drugs targeting RNA polymerase. The 56r moiety of the FDAapproved drug 26 (vidarabine) system is purin and the repurposing drug 27 (remedisivir) system is pyrrolo[2,1f][1,2,4]triazin-4-amine. The 5r moiety of CAN drug 29 (nitrofurantoin macrocrystalline) system is an N-heterocycle that is connected to the 5r moiety of O-heterocycle through three bonds. The 65r Ù 66r is observed in THI molecule 30 (alphadichroine), in which the 5-membered ring of a fused 65r is connected to a 6-membered ring of a fused 66r. The structural motif of NPC-approved drug 31 (fidarestat) is denoted by 5r$56r and is characterized by the presence of a spiro carbon center with a fluorine substitution on the phenyl ring. In the case of 32, the FDA-approved drug ofatumumab, a triple ring-fused, 566r structural moiety is present which is connected to a five-membered N and O-heterocycle.

Group 4: Miscellaneous collection
Ten molecules are sorted into this category (33 2 42) are shown in Figure S5 and among them, 33 2 38 are neither Nheterocyclic nor O-heterocyclic. The 33 2 36 systems are characterized by a 6r structural motif, the phenyl unit. The THI 37 (m-digallic acid) has the 6r Ù -6r configuration with two phenyl rings while another THI 38 (shikonin) with the 66r configuration is a derivative of 1,4-naphthoquinone. The side chains of these systems are composed of ester, acid, carbonyl, hydroxy, nitro, amino, and imino functional groups. For instance, the FDA-approved general-purpose sweetener aspartame (35) is a methyl ester of the dipeptide of the natural amino acids L-aspartic acid and L-phenylalanine. The antidiabetic drug phenformin (36) is a biguanide with a 2-   Seven repurposing drugs are docked by using Autodock 4.2 software and the docked structures are used for the ONIOM calculations. The SARS-CoV-2 S-protein: human ACE2 interface region, represented with the stick model shows that the ligand is surrounded by both ACE2 (olive green sticks) and S-protein (purple sticks) portions. The cavity observed at the interface is approximately 1 nm 3 size and the orientation of the entrapped ligands in this cavity is depicted with the van der Waals surface model. All the ligands fit well inside the cavity and the substituents of the ring moieties participate in noncovalent interactions with both S-protein and ACE2 portions. The ring moieties, viz. 5r, 6r, 65r, 56r, 66r, and 665r, spiro '$' connection and cage moieties of ligands give structural rigidity while the connections, viz. '-', ' Ù ', and ' Ù -' give flexibility to the connected ring moieties for adopting suitable orientations within the cavity.

Structure of ML complexes and binding energy
The binding energy (E b ) of the ligand within the interface cavity, calculated at ONIOM(B3LYP/6-31G Ã :PM7) level for Oheterocycles (THI), N-heterocycles, N, O-heterocycles, and miscellaneous collection of molecules is given in Table S1, S2, S3, and S4, respectively. Among THIs, myricetin (3) and glucogallin (12) show the best E b values À85.4 and À59.3 kcal/mol, respectively while the rest of the ten molecules have the 6r-66r structural motif show E b in the range À13.5 to À36.5 kcal/mol. As shown in Figure 5, the strongest binding observed for myricetin, a multifunctional flavonol (Agraharam et al., 2022) that belongs to the 6r-66r category can be attributed to the hydrogen bond interactions arising from the three hydroxy groups at 6r with bond distances 1.72, 1.91, and 2.06 Å and two hydroxy groups at 66r with bond distances 1.67 and 1.90 Å and one with a carbonyl group at 66r with distance 1.58 Å. Glucogallin creates five strong interactions with the glutamic acid, asparagine, arginine, lysine, and histidine residues with bond distances 1.63, 1.72, 2.03, 1.92, and 1.87 Å respectively. The superior E b data observed for myricetin and glucogallin suggest that the presence of three hydroxy groups at the 3, 4, positions of the 6r unit is advantageous for the effective binding of the ligand. However, a comparison of the E b of these systems with leucodelphinidin (10) which is also substituted with hydroxy groups at 3, 4, 5 positions of the 6r indicates that the inferior E b of this system is due to the saturated structure of the Oheterocycle in the 66r moiety. In fact, all structures showing saturated carbon on the O-heterocycle (6 2 11) show inferior binding properties.
Among the thirteen N-heterocyclic molecules studied here (13 2 25), pyruvic acid calcium isoniazid (14), sapropterin (18), and quinonoid (19) show strong binding with E b in the range À41.0 to À46.1 kcal/mol. The repurposing drug ruxolitinib (23) exhibits the least E b À2.7 kcal/mol which can be attributed to the sterically repulsive interactions arising from the cyclopentyl unit (supp info). Among the repurposing Nheterocyclic drugs, galidesivir has the best binding energy À34.3 kcal/mol. The strongest binding observed for sapropterin which belongs to the 66r category can be attributed to the hydrogen bond interactions arising from the carbonyl group at 66r with bond distances 1.70, 1.63 Å, and NH … OC bond with a distance 1.87 Å as well as hydrogen bonds with hydroxy groups at side chains with bond distances 1.98 and 1.74 Å ( Figure 5). Similarly, the combined effect of several significant non-covalent interactions as shown in Figure 5 stabilizes the quinonoid (19) ligand within the cavity.
Among the seven N, O-heterocyclic molecules studied here (26 2 32), NPC-approved drug fidarestat (31) with the 5r$56r structural motif shows the strongest binding with E b À76.1 kcal/mol ( Figure 5). Here two strong H-bonds are observed from the 5r portion while the 56r portion shows four such bonds, including one fluorine-to-arginine interaction at a distance of 2.07 Å. Moreover, the repurposing drug remdesivir (27) and nitrofurantoin (29) exhibit high binding with E b À46.7 and À44.8 kcal/mol, respectively. In the case of remdesivir, in addition to two strong interactions (1.84 and 1.93 Å), thirteen other interactions with bond distances in the range 2.24 to 2.81 Å are observed ( Figure 5). Similarly, nitrofurantoin (5r Ù -5r) also displays two strong interactions with bond distances 1.69 and 1.74 Å (with arginine). E b of all other molecules is in the range of À18.5 to À34.9 kcal/mol. Among this category of molecules studied, protirelin (39) and m-digallic acid with the 6r Ù -6r motif (37), under the miscellaneous class of molecules (33 2 42) show the best E b values À71.5 and À71.2 kcal/mol, respectively (Table S4). Protirelin is rich with several amide functionalities and three Nheterocycles which lead to the formation of multiple noncovalent interactions with the interface moieties ( Figure 5) and among them, interactions with distances 2.03, 1.61, 2.03, and 1.73 Å contribute significantly to the E b . In the case of m-digallic acid, all the five hydroxy groups and the carbonyl unit of the ester group are noncovalently connected with various interface moieties ( Figure 5). Among them, carboxylic acid moieties at 6r create two H-bonds with 1.67 and 1.45 Å distances, furthermore, hydroxy groups at 6r produce H-bonds with distances 1.54 and 2.04 Å are the most significant for the strong binding character ( Figure 5). Also, the miscellaneous ligands 33, 35, and 40 show high E b values À57.6, À40.9, and À47.7 kcal/mol, respectively. Compared to 33, its optical isomer 34 shows inferior binding affinity with the interface and this result also suggests that the binding interaction of the ligand with a biological receptor can be greatly influenced by the configuration of the chiral center. Aspartame (35), a sugar substitute commonly used in food supplements and beverages is a dipeptide. The high E b observed for this ligand is attributed to the noncovalent interactions arising from its carboxylic acid and methyl ester moieties (supp info).
It may be noted that the docking score for the entire sample falls in the narrow range of À5.0 to À7.7 kcal/mol and it suggests that the data is not sensitive enough to distinguish the binding behaviour of these molecules at the interface. An indepth analysis of the noncovalent binding behaviour of the cavity is obvious from the DFT-ONIOM studies. The cumulative effect of the strength of the H-bonds created with the residues can be visible from the analysis of quantum chemical studies and this treatment is more reliable than docking studies. The observed E b values of the whole sample by the ONIOM method lie in the range of À2.7 to À85.4 kcal/mol. Among them, the strongest binding interactions are observed for myricetin and the weakest binding is shown by ruxolitinib. Figure 5 shows detailed information about the interactions for the two prominent complexes which show the highest E b from each group. In all cases, around L, more than eight interactions can be identified which support their high binding behavior in the cavity. These interactions mainly arise from arginine, histidine, tyrosine, and lysine moieties as well as from carboxylate and amide units of the interface. Noncovalent bonding with the arginine moiety is found in all cases. The myricetin (3), sapropterin (18), quinonoid (19), remdesivir (27), fidarestat (31), tazobactum (40) show interactions from two arginine moieties. 3, 14, 18, 19, 27, 29, 31, 39, and 40 show interactions from tyrosine moiety. The THI molecules m-digallic acid, benserazide hydrochlorideseradie, and glucogallin are devoid of interactions with tyrosine moiety whereas protirelin (39), pyruvic acid calcium isoniazid (14), 19, and repurposing drug 27 show one tyrosine and two argininedrug interactions. Histidine N-interactions are observed in 12, 31, 33, and 39 molecules. Unsubstituted ring structures (5r, 6r, 65r, 56r, 66r, 665r) remained aloof from interactions whereas ester, acid, carbonyl, hydroxy, nitro, amino, and imino functional groups in rings, side chains, or spacers exhibit major binding features. The binding energy profile of all ligands in the four classifications is summarized as a bar chart in Figure 6. For a drug-like molecule, Lipinski's rule of five suggest molecular mass less than 500 Dalton, high lipophilicity (expressed as LogP less than 5), less than 5 hydrogen bond donors, less than 10 hydrogen bond acceptors, and molar refractivity between 40-130. All the molecules selected for this study except 1, 3,10,12,18,19,21,32,33,34,36, and 37 obey the Lipinski's rule. Table S9 is provided for all the molecules selected for this study in which twelve molecules show a hydrogen bond donor count greater than five. Moreover, analysis of the HOMO, LUMO and HOMO-LUMO band gap energy data of all the molecules ( Figure S8) suggest their high kinetic stability (Aihara, 1999). In Figure 6, the names of thirteen molecules which show binding energy better than À40.0 kcal/mol (supp info.) are depicted and among them 3, 12, 19, 33, and 37 violate Lipinski's rule meaning that the most promising drug-like moleules, in the increasing order of binding strength are fidarestat (31), protirelin (39), tazobactum (40), remdesivir (27), sapropterin (18), nitrofurantoin (29), pyruvic acid calcium isoniazid (14), and aspartame (35).
For a better understanding of interactions within the protein cavity, ligand structure is again simplified to head (H) and tail (T). Here, H is used to represent the higher-order ring motif (56r, 65r, 66r, 665r) which is connected to T, the lower order ring motif (5r or 6r), or simple linear side chains. Generally, the ONIOM-DFT optimized complexes show better binding energy when the T portion orients towards the "A" region of the cave structure inside the pocket and H orients towards the "B" region of the binding pocket as shown in Figure 7, the head-tail 66r-6r representation of myricetin (3).
The FDA-approved sugar substitute aspartame (35) shows the lowest absolute binding energy 40.9 kcal/mol among the selected thirteen molecules. It shows only three significant H-bonds having bond distances 1.63 Å with acid moiety, 1.97 Å with ester moiety, and 2.00 Å with amines moiety. A conspicuous fact is that its 6r moiety (H), a phenyl ring is devoid of any significant interactions and E b observed for aspartame can be mainly attributed to the collective strength of the hydrogen bonds created by the T portion. This suggests that by modifying the 6r moiety, improvement in E b can be achieved. In order to test this hypothesis, we have designed three different hybrid structure 43, 44, and 45 with one hydroxyl group at the meta position, para position and three hydroxyl groups on the phenyl ring, respectively. The 45 showed the best interaction followed by 43 and 44. Due to hydrogen bond donor count 7, Lipinski's rule is violated by 45. The optimized structure of 43 within the cavity is shown in Figure 8 (b) along with the noncovalent interactions (see supporting information for 44 and 45). Compared to aspartame, the phenyl ring of 43 shows two strong significant H-bond interactions, one involving a cationic NH 3 moiety and the other involving a carboxylate unit. Further, the T portion shows multiple H-bond interactions. Altogether, the E b of the modified aspartame is improved with seven strong interactions having bond distances 1.49, 1.61, 1.52, 1.93, 2.09, 2.10, and 1.84 Å and leads to the highest binding energy À84.5 kcal/mol. The significance of hydroxy substitution in 6r moiety is obvious from this result. The hybrid design indeed strengthens the overall interaction between the ligand and receptor.
The Ls showed multiple noncovalent interactions with both S-protein and ACE2 units. Though the docking score reported by Smith and Smith is nearly the same for all Ls, the DFT results show that the binding energy of the two most strongly binding Ls is much better than the least binding L. From the sample of forty-two drug-like molecules, myricetin (3) is observed with the highest binding energy. Hydroxyl substituted 6r moiety shows significant interactions with the protein cavity. Further, we have made use of this ONIOM-DFT method to screen the ligands against the SARS-CoV-2 S-protein: human ACE2 interface. These calculations have highlighted the high binding features of thirteen molecules viz. myricetin (3), fidarestat (31), protirelin (39), m-digallic acid (37), glucogallin (12), benserazide hydrochlorideseradie (33), tazobactum (40), remdesivir (27), sapropterin (18), nitrofurantoin (29), quinonoid (19), pyruvic acid calcium isoniazid (14), and aspartame (35) which are arranged in the decreasing binding energy value. All these molecules showed more than three strong interactions around them connecting with moieties such as arginine, histidine, tyrosine, lysine, carboxylate, and amide from both S-protein and ACE2 residues. The high binding energy is attributed to the cumulative effect of all these interactions. 3, 12, 31, 33, and 37 can be excluded due to the violation of Lipinki's rule of five (Mitra et al., 2021) (given in supp info.). In many cases, better binding energies are observed when the T portion orients towards the inner 'A' region and H orients towards the outer 'B' region of the binding pocket.
Aspartame, a ligand with moderate binding energy showed rich noncovalent interactions at the interface through its tail region composed of amino, ester, and acid moieties. The strongly interacting THI myricetin has significant interactions at the interface through its tail portion, a hydroxy substituted 6r structural motif. The modified aspartame ligand 43 is emerged as the most strongly binding Ls at the interface due to the incorporation of a hydroxy substitution at meta position of the phenyl ring. The vast improvement in the binding energy observed for the hydroxylderivative of aspartame suggests that the FDA-approved sugar substitute could be easily modified for the development of antiviral drugs. Also such a strategy points towards the repurposing of food derivatives for drug discovery.
In summary, the two-layer ONIOM-DFT optimization strategy is an effective and affordable way to model the active site of drug-receptor complexes. The inner layer of ONIOM optimized structures provides the specific binding features of the molecules at a reliable quantum chemical level to understand the drug-receptor binding process. The conserved features of the binding domain can be brought out from such studies with structural and energetic details on noncovalent binding. Also, several promising ligands for the inhibition of the activity of S-protein on the ACE2 receptors are obtained.