In silico discovery of novel quinazolin-4-one derivatives as potent 3CLpro inhibitors using 2D-QSAR, molecular docking, and molecular dynamics

2.1. Data collection, molecular structures preparation, and molecular descriptor calculation

In this study, 73 quinazolin-4-one derivatives were taken from the published work of Zhang et al. [13]. Their 2D structures were first drawn using ChemDraw and then converted into 3D using ChemDraw 3D 23.0 [24]. To prepare the molecules for modelling, geometry optimization and partial charge assignments were done using the Merck molecular force field 94 (MMFF9) force field with the steepest descent algorithm, running 1000 steps. For building the GA-MLR model, molecular descriptors were calculated from the optimized 3D structures using alvaDesc (version 1.0.8) [25]. To reduce redundancy and avoid multicollinearity, descriptors with constant values above 95% were removed, and from each pair of highly correlated descriptors (correlation coefficient > 0.9), only one was kept. The experimental IC₅₀ values were converted to their negative logarithmic form (-log IC₅₀). The full list of the molecular structures and their corresponding IC₅₀ values has been provided in Table S1.

2.2. QSAR model generation and validation

In QSAR modelling, selecting the most relevant and informative descriptors is an essential step in drug discovery. Therefore, a stepwise MLR in XLSTAT Excel Statistical Analysis was employed to reduce the 1351 computed descriptors to 15 descriptors (Table S2), which were then used in an ordinary least squares MLR analysis using QSAR-INSubria (QSARINS) software [25,26] to regress pIC₅₀ values with the molecular features of compounds. The dataset was divided into a 75% training set (55 molecules) and a 25% test set (18 molecules), and GA-MLR models were developed by examining all subsets of up to eight descriptors over 10000 generations with a mutation probability of 0.05, while all other parameters remained at their default values. To ensure the robustness and predictive power of the model for new compounds, we carried out a comprehensive validation workflow consisting of internal cross-validation, Y-randomization, external validation on the test set, and applicability domain assessment, as detailed in our earlier publications [27-29].

2.3. Fragment database preparation

The FragGrow web server’s “Direct-Grow” mode was employed for structure-based drug design to modify the lead compound [30,31]. This mode was selected because it allows for the direct growth of molecules by replacing specific substructures within the original compound. Initially, the 3D structure of compound labelled 51 (Figure 2), saved in protein data bank (PDB) format, along with the target protein structure, was uploaded to the fragGrow platform to initiate the fragment growing process. Following this, five distinct regions were identified and selected as potential growth vectors, which were based on chemical bonds involving hydrogen atoms and acyclic single bonds that were located within 9 Å from the target protein. To optimize the fragment search, a comprehensive set of structural and feature constraints was meticulously defined, concentrating on various factors such as interactions with specific protein sites, size, topology, and physicochemical properties of the desired fragments. This careful selection process was crucial for narrowing down the search space and ensuring that the retrieved fragments would be suitable for further development. Once the constraints were established, suitable fragments were retrieved from the indexed 3D fragment database maintained by FragGrow. These fragments were then merged with the original compound-51 through an automated process facilitated by the platform. Figure 2 illustrates the lead compound structure, highlighting the five key substitution positions (A-E). This structure was chosen because it exhibits the highest 3CLpro inhibition activity, making it an excellent starting point for designing new compounds that retain the central scaffold while aiming for improved potency. Analysis of the starting molecule revealed that regions A, B, and C engaged in fewer interactions, particularly at position A. To ensure diversity within the library of tested molecules, we selected all significant positions, including D and E, for systematic evaluation.

Close

Figure 2.

Lead-51 scaffold highlighting key substitution positions (A-E).

Export to PPT

This approach enabled an iterative evaluation and optimization of the generated ligands based on predicted binding affinities, which is vital for assessing their potential efficacy. Ultimately, these steps aimed to enhance the compound’s therapeutic properties, contributing significantly to the rational drug design process in medicinal chemistry of SARS-CoV inhibitors. The comprehensive capabilities of FragGrow serve to streamline the process of ligand optimization, providing valuable assistance to researchers in drug discovery endeavors.

2.4. Molecular docking

Molecular docking is one of the most widely used techniques in rational drug design [32]. It plays a crucial role in understanding the binding interactions between a drug and its target protein, primarily through active site determination [33]. In the current study, molecular docking was employed to investigate the interaction between the designed compound and the active site of the 3CLpro of SARS-CoV-2. The crystallographic structure of 3CLpro was retrieved from the (PDB ID: 6M2N). The choice of the structure was based on its high resolution (2.20 Å) and the presence of the co-crystallized compound (Baicalein) [34,35]. The complex is presented in a tetrameric format, resulting from the interaction between the target and its inhibitors. For protein preparation, water molecules and the co-crystallized ligand were removed. Polar hydrogens and Kollman charges were added using AutoDock tools [36]. The selected hit ligands, identified through QSAR modelling, were prepared by assigning Gasteiger charges and defining rotatable bonds and were saved in PDBQT format. Docking simulations were performed using AutoDock Vina [37]. The grid box and docking parameters were summarized in Table S3. The grid box was defined based on the coordinates of the native co-crystallized ligand in 6M2N, ensuring that the box was precisely cantered on the catalytic dyad (HIS41 and CYS145) and encompass the entire substrate-binding pocket, including all key residues involved in ligand recognition (GLY143, ASN142, GLU166, MET49, LEU141, SER144, HIS163, MET165, ASP187, and CYS44). This configuration provides optimal coverage of the active site and ensures that all relevant binding interactions can be captured during docking. The validation of the docking protocol was confirmed by calculating the root mean square deviation (RMSD) between the redocked pose and the native crystallographic pose. Successful validation is typically indicated by an RMSD value of less than 2 Å. Following the docking of the hit compounds with the active site of 3CLpro, the resulting protein–ligand complexes were analyzed using BIOVIA discovery studio visualizer [38]. This analysis allowed for the identification of key interactions between the ligands and the target protein.

2.5. ADMET analysis and molecular dynamics simulations

Following the molecular docking study, compounds with binding affinities greater than both lead and reference compounds were shortlisted for further pharmacokinetic study following the molecular dynamics simulation. ADMET properties were predicted using the ADMETlab 2.0 online platform [39]. Each compound was chosen based on favorable physicochemical profiles, good metabolic stability, and high intestinal absorption, with no hepatotoxic, carcinogenic, or mutagenic effects. From this analysis, three compounds with the best safety and pharmacokinetic profiles were selected for molecular dynamics simulations.

The structural stability of the selected compounds against the 3CLpro target was assessed by carrying out conventional all-atom MD simulations using GROningen MAchine for Chemical Simulations (GROMACS) 2021.3 software [40]. The input topology and system parameters were generated using the CHARMM Graphical User Interface (CHARMM-GUI) web server [41] with the CHARMM36 force field [42]. Each ligand-protein complex was placed in a rectangular periodic box and solvated using the transferable intermolecular potential with 3 Points (TIP3P) water model [43]. Sodium (Na⁺) and chloride (Cl⁻) ions were added to neutralize the system and maintain a physiological 0.15 M salt concentration, using Monte Carlo ion placement. The Verlet cut-off scheme was applied for short-range interactions, using a 12 Å cut-off for both van der Waals and Coulombic interactions. Long-range electrostatics were treated using the Particle-Mesh Ewald (PME) method [44]. Bond constraints, including those involving hydrogen atoms, were applied using the LINear constraint solver (LINCS) algorithm [45]. Energy minimization was performed using the steepest descent algorithm with a maximum of 50,000 steps and a convergence threshold of 10.0 kJ mol^-1 nm^-1. Subsequently, a two-phase equilibration was carried out: (i) NVT equilibration for 500 ps using the Nose-Hoover thermostat [46] at 303 K, and (ii) NPT equilibration for 500 ps at 1.01325 bar using the Parrinello-Rahman barostat [47]. The final production MD run was performed for 150 ns to analyze the dynamic behavior and stability of the ligand-protein complexes. Post-simulation analysis included the calculation of RMSD, Radius of Gyration (Rg), Root mean square fluctuation (RMSF), and Hydrogen bonds (HB), solvent accessible surface area (SASA) to evaluate conformational stability, compactness, flexibility, and solvent exposure. All trajectories were processed using GROMACS tools, and the results were visualized with Xmgrace software [48].

3.1. GA-MLR 2D-QSAR model

The 2D-QSAR GA-MLR model was developed using eight key descriptors selected from the full descriptor pool: TDB07i, CATS3D_08_AA, MATS5m, Inflammat-80, RDF125s, IVDE, R1u+, E1u. with detailed descriptions and chemical interpretations provided in Table S4. These descriptors were identified as having a significant impact on the pIC₅₀ values and the molecular features of the compounds. They were ultimately used to build the QSAR model. The resulting GA-MLR model equation (Eq. 1), along with its statistical performance metrics, are presented below.

${\text{N}}_{\text{training}}=\text{55},{\text{R}}^{\text{2}}=0.\text{8659},{\text{Q}}^{\text{2}}{}_{\text{loo}}=0.\text{8}0\text{54}$

${\text{N}}_{\text{ext}}=\text{18},{\text{R}}^{\text{2}}{}_{\text{ext}}=0.\text{7385},{\text{MAE}}_{\text{ext}}=0.\text{3278},{\text{CCC}}_{\text{ext}}=0.\text{8423}$

${\text{Q}}^{\text{2}}{}_{\text{F1}}=0.\text{734}0,{\text{Q}}^{\text{2}}{}_{\text{F2}}=0.\text{7324},{\text{Q}}^{\text{2}}{}_{\text{F3}}=0.\text{7389}$

${\text{RMSE}}_{\text{tr}}=0.\text{2685},{\text{RMSE}}_{\text{ext}}=0.\text{3747}$

$\text{S}=0.\text{2935},\text{F}=\text{37}.\text{1361}$

The best GA-MLR model was built based on eight molecular descriptors, as shown on the equation above.

All statistical parameters meet QSAR model validation criteria, and the graphical representation (Figure 3(a) and Figure S1) demonstrates a strong correlation between experimental and predicted pIC₅₀ values. As all randomized models exhibit significantly lower R² and Q² values compared to the original model, The Y-randomization plot confirmed that the developed 2D-QSAR model is statistically robust and not the result of a chance correlation. Similarly, the Williams plot reveals that all compounds fall within the applicability domain of the model, which indicates reliable predictions without influential outliers (Figure 3(b)). Thus, the model equation was used to predict the potential inhibition activities of the newly designed molecules.

Close

Figure 3.

(a) Correlation between experimental and predicted pIC₅₀ values; (b) William’s Plot for applicability domain assessment.

Export to PPT

3.2. Fragment-based generation and QSAR filtering

In the FBDD workflow using FragGrow. We focused on optimizing lead compound 51 by introducing structural changes across five key regions labeled (A-E) as shown in Figure 2. From each of these regions, we generated 200 new derivatives, resulting in a diverse library of 1000 compounds (Excel sheet supplementary materials).

To screen this set for the best candidates, we applied several filtering steps, where the 2D-QSAR model served as the main selection criterion. The filtering outcomes varied depending on the different regions, where: A showed the most promise, yielding 47 candidates, while B and C contributed 15 and 10 molecules, respectively. However, from position D, only 4 molecules passed the 2D-QSAR threshold and were selected for docking. Finally, none of the compounds derived from region E met the criteria for further evaluation. These findings highlight the potential of the FBDD strategy used in producing a wide range of new molecules. In addition, the strength of the established 2D-QSAR model to screen for potent compounds with high-predicted inhibitory activity was elucidated by selecting 76 compounds from the generated set of 1000 molecules.

3.3. Molecular docking

To begin the molecular docking study, the protocol was first validated through a redocking computation. The native ligand Baicalein was re-docked in the active site of the 3CLpro protein. The resulting binding poses closely matched the original co-crystallized conformation with an RMSD value of 0.7036 Å, as shown in Figure 4. This RMSD value is below the accepted threshold of 2 Å, confirming the reliability and accuracy of the used docking protocol.

Close

Figure 4.

Superimposition of the co-crystallized ligand (green) and the redocked pose (salmon) showing binding pose agreement.

Export to PPT

Among the 1000 designed compounds, the 76 molecules selected by the 2D-QSAR model were subjected to a molecular docking study within the active site of the target protein 3CLpro. As a result, 38 compounds were identified with higher binding affinities than both the co-crystallized ligand baicalein (-7.6 kcal mol^-1) and the lead-51 (-7.7 kcal mol^-1). Subsequently, ADMET filtering was applied to assess the pharmacokinetic and toxicity profiles of these top 38 compounds. Out of them, only 3 compounds were found to pose no toxicity risk, indicating their potential as promising drug candidates. The chemical structures of the top 3 hits, along with their predicted pIC₅₀ values and binding affinities, have been presented in Table 1. The complete list of all 38 compounds has been provided in Table S5.

Table 1. Binding affinities and pIC₅₀ values of the top three ranked hit compounds with the lead compound and Baicalein after molecular docking.

Position	Hit ID	Chemical structure	pIC50	Binding affinity (kcal.mol-1)
	Baicalein		-	-7.6
	Lead-51		7.081	-7.7
A	A-100		7.459	-8.5
	A-117		7.703	-8.3
D	D-119		7.137	-7.8

The interaction of the 3CLpro target’s residues with the selected compounds was carefully analyzed using the discovery studio visualizer tool. All the candidate drugs demonstrated strong binding affinities with key residues from 3CLpro’s active site. The molecule A-100 was stabilized by nine hydrogen bonds and five hydrophobic contacts within the active site, including hydrogen bonds with the catalytic residues HIS41 and CYS145, as well as hydrophobic interactions involving HIS163. Similarly, the compound A-117 formed nine hydrogen bonds (five conventional and four carbon-hydrogen bonds), including contacts with ASN142 and GLU166, as well as five hydrophobic and other interactions, thereby increasing its binding to the active site. The compound D-119 exhibited three conventional hydrogen bonds and two electrostatic pi-cation interactions, which were supplemented by six hydrophobic contacts, particularly engaging with the essential catalytic site residue HIS41 and close residues such as CYS145. Interestingly, the lead compound’s interaction pattern paralleled that of D-119, showing similar stabilizing characteristics via hydrogen bonds, pi-cation interactions, and hydrophobic contacts with key active site residues. Baicalein, a natural flavonoid, produced five hydrogen bonds (three conventional and two carbon-hydrogen bonds), including contacts with GLY143, GLU166, and HIS41, as well as significant electrostatic and hydrophobic interactions. particularly with HIS41 and MET49. Figure 5 illustrates all these interactions with the key residues of the target 3CLpro.

Close

Figure 5.

2D representations of the interactions of newly designed compounds (A-100, A-117 and D-119), along with the Lead-51 and Baicalein within the 3CLpro binding pocket.

Export to PPT

To complement the 2D interaction visualizations and the docking score table, a comparative interaction heatmap was generated for all compounds (Figure 6). This heatmap integrates the interaction counts for each compound across the key active-site residues of 3CLpro, allowing a direct comparison of their binding profiles. It highlights conserved interaction hotspots at HIS41, CYS145, GLU166, and ASN142, which are consistently engaged across the compounds. Moreover, the distinct intensity patterns observed for A-100, A-117, D-119, and the lead compound emphasize their stronger and more diverse interactions with these catalytic residues. Overall, the interaction analyses indicate that all selected drugs interact efficiently with one or both key catalytic residues (CYS145 and HIS41), indicating their potential as 3CLpro inhibitors. Table S6 in the supplementary materials presents a detailed overview of the interactions of the top three drug candidates with 3CLpro.

Close

Figure 6.

Molecular docking interaction heatmap of newly designed compounds (A-100, A-117 and D-119), the Lead-51 and Baicalein.

Export to PPT

The QSAR model analysis highlighted the key molecular descriptors influencing the activity. The descriptor TDB07i measures the occurrence of atom pairs with high ionization potential (e.g, halogens) separated by 7 bonds. This descriptor exhibited a negative coefficient (-0.9306) in the model. This indicates that a high TDB07i value reflects numerous distant electron-withdrawing centers which correlate with reduced potency (lower pIC₅₀), likely due to diminished electronic complementarity. As a result, comparing A-100 and A-117, the removal of halogen substituents led to a decrease in TDB07i and a concomitant increase in predicted activity. Contrariwise, D-119 contains more heteroatoms separated by 7 bond intervals, indicates a higher TDB07i and slightly lower activity. The Inflammat-80 descriptor represents structural similarity to known anti-inflammatory agents and had a positive coefficient (+0.3202) in the model equation. This suggests that molecules with higher Inflammat-80 values are predicted to be more potent. This trend likely arises that scaffolds sharing pharmacophoric features, such as hydrogen bond donors and acceptors or planar aromatic rings, are favorable for target binding. In our series, the addition of aromatic planar rings in A-100 and A-117 enhanced pIC_5O, and the introduction of an additional oxygen atom in A-117 contributed more hydrogen bonding potential. The E1u descriptor quantifies global electro topological variation in the molecule and displayed a highly positive coefficient (+5.6159) in the model equation. This suggests that increased electronic and topological heterogeneity enhances target interaction and activity. The three selected hits included more diverse substituents and a greater heteroatom number compared to the lead structure.

Previous studies have shown that the designed quinazolin-4-one derivatives engage the SARS-CoV-2 main protease active site through diverse types of interactions such as hydrogen bonding, π-stacking, and hydrophobic interactions with key residues such as HIS41, CYS145, and GLU166, consistent with previously reported binding modes for potent non-covalent inhibitors [15,17]. Favorable docking scores were particularly observed with compounds forming multiple hydrogen bonds with backbone and side-chain atoms of GLU166, GLY143, and ASN142, as well as π-sulfur contacts with CYS145 and π-alkyl contacts with Met49 and Met165 [16]. The catalytic dyad especially HIS41/CYS145 plays a crucial role in achieving a high binding affinity and specificity for substrate-binding subsites (S1, S2, S4), corroborating recent reports that non-covalent quinazolinone-based inhibitors can occupy novel binding modes distinct from traditional peptidomimetic or covalent ligands [15].

3.4. ADMET analysis and molecular dynamics simulations

The pharmacokinetic and toxicity properties of all the compounds that have passed the docking filter were assessed to determine their suitability as drug candidates. The evaluation included physicochemical properties, drug-likeness, and toxicity parameters. The ADMET Table in the Excel sheet summarized the results and served to select the best safe three compounds. The physicochemical and ADME properties results of the three selected compounds have been summarized in Table 2, while Table 3 is presented their toxicity assessment. All these compounds have molecular weights between 327.09 and 511.14 Da, which is acceptable for oral bioavailability (less than 500 to 600 Da). Due to its lowest molecular weight (327.09 Da), compound D-119 is expected to improve membrane permeability and oral absorption ability. The LogP values ranging from 1.384 to 2.626 demonstrate a moderate level of lipophilicity among all compounds, this suggests that they exhibit favorable membrane permeability while avoiding excessive hydrophobicity, which could result in poor solubility or toxicity issues. This range corresponds effectively with the ideal LogP range (1-3) for oral medications. The hydrogen bond acceptors (HBA) and donors (HBD) of all compounds align with Lipinski’s rule of five, indicating their potential as orally active molecules. Compound A-117 exhibits the highest HBA value of 10, potentially affecting its solubility and binding affinity. Considering these advantageous characteristics, all compounds demonstrated accordance with the Lipinski and Pfizer rules. Suggesting a favorable drug profile and a reduced risk of failure attributed to inadequate pharmacokinetics or toxicity. The synthetic accessibility scores, which range from 2.756 to 3.417, suggest a moderate level of ease in synthesis. with compound A-100 demonstrating the highest degree of synthetic accessibility. Blood-brain barrier (BBB) permeability values are relatively low, indicating limited CNS exposure, which is favorable for non-neuroactive agents. Similarly, human intestinal absorption (HIA) values (0.007-0.061) suggest moderate to low absorption, with compound D-119 showing the highest predicted HIA. The Caco-2 permeability values (logS range: -4.893 to -5.027) are slightly below the ideal threshold (>-5.15), indicating acceptable passive diffusion through intestinal epithelial cells. Notably, compound A-100 has the best permeability among the three. In terms of cytochrome P450 interactions, all compounds are predicted to be CYP2C9 substrates, with no indication of CYP2C9 inhibition. This is advantageous, as it minimizes the risk of enzyme inhibition-related drug-drug interactions. However, metabolism via this isoform should still be considered in pharmacokinetic planning. Clearance rates varied from 0.838 to 1.382 mL/min/kg, with compound D-119 exhibiting the highest clearance. While high clearance may reduce half-life, it also implies efficient elimination, which may be beneficial depending on the therapeutic context. In terms of toxicity, none of the compounds were predicted to be mutagenic or carcinogenic. All these chemicals have AMES (bacterial reverse mutation assay) values below 0.5, which indicates that they have a low mutagenic potential. Scores for carcinogenicity are also well within the acceptable range. Importantly, according to the hepatotoxicity predictions. Compound D-119 is the most likely to lead to liver damage (0.667), whereas compound A-117 has the lowest risk (0.282), indicating that compound A-117 may have a better safety profile for future in vitro experiments.

The binding behavior of the studied ligands and the conformational stability of ligand-protein complexes were assessed using molecular dynamics simulations. In this investigation. The metrics Rg, RMSF, SASA, HB, and RMSD were computed. The RMSD was calculated to quantify the change in a protein’s backbone conformation between its native state and its final state. The protein’s stability with regard to its structural conformation is determined by the RMSF seen during a simulation. It serves as well as locating flexible regions. The degree of structural variation in a protein is proportional to how well it keeps its structure together or how unstable it is. Protein compactness is measured by the Rg, which is calculated by comparing the protein’s Rg values before and after ligand binding. Additionally, SASA was calculated to assess the extent of the protein’s surface exposed to the solvent and to monitor structural compactness during the simulation. Similarly, hydrogen bonds were analyzed to evaluate the strength and stability of interactions between the ligands and the protein over time.

The examination of the MDs results (Figure 7, Table 4) indicates that the overall average RMSD values for 3CLpro-apo, 3CLpro-Lead, 3CLpro-Baicalein, 3CLpro-D-119, 3CLpro-A-100, and 3CLpro-A-117 are 0.219, 0.251, 0.69, 0.192, 0.206, and 0.287 nm, respectively. The lowest average RMSD value suggests that 3CLpro-D-119 demonstrates enhanced stability in comparison to 3CLpro when bound to the Lead compound. The elevated RMSD values suggest that 3CLpro-A-117 exhibits reduced stability in comparison to both the 6M2N apo form and 3CLpro-Lead. Additionally, the RMSF analysis of the complexes and the apo form of Mpro was conducted by evaluating C-alpha atoms across the full duration of the simulation. The RMSF values of the Mpro bound to the Lead compound served as a baseline for assessing the flexibility of the designed compounds in the analysis of the binding of the three compounds to Mpro. The 3CLpro-D-119 complex validates the RMSD findings exhibiting the lowest RMSF in comparison to both the apo protein and the protein-lead complex. The 3CLpro-D-119, 3CLpro-A-100, and 3CLpro-A-117 complexes exhibited a reduced RMSF compared to the 3CLpro-Lead complex, indicating that D-119, A-100 and A-117 may serve as potential and strong inhibitor candidates for Mpro. The complexes 3CLpro-apo, 3CLpro-Lead, 3CLpro-Baicalein, 3CLpro-D-119, 3CLpro-A-100, and 3CLpro-A-117 yielded average Rg values of 2.193, 2.24, 2.222, 2.216, 2.205, and 2.199 nm, respectively. The 3CLpro-D-119 and 3CLpro-A-100 complexes exhibited comparable or marginally higher Rg values relative to the Mpro in its apo form while demonstrating lower values than Mpro bound to lead compound 51, suggesting that these compounds are less densely packed. To assess the extent of the protein’s surface exposed to the solvent in both its unbound and ligand-bound forms. The SASA of each system was calculated. In the apo state, the protein exhibited a SASA value of 150.913 nm², indicating a stable and compact conformation. Upon binding to the lead compound, the SASA increased to 153.196 nm², suggesting a slight expansion of the protein’s surface as a result of ligand accommodation. Baicalein presented a SASA of 153.626 nm², the highest among all tested systems, which may reflect enhanced solvent exposure due to local conformational changes upon ligand binding. The newly designed compounds D-119, A-100 and A-117 exhibited SASA values of 151.398 nm², 152.398 nm², and 150.628 nm², respectively. These values indicate that ligand binding induced minimal alterations in the overall solvent-exposed surface area of the protein. The relatively consistent SASA values across all complexes suggest that none of the ligands significantly disrupted the structural integrity or folding of the 3CLpro protease, supporting their potential as stable inhibitors. However, to deepen understanding of how these systems, protein-ligand, were stabilized, hydrogen bonds were analyzed during the 150 ns. In Figure 7(e), it was shown that the lead compound and Baicalein formed a relatively low and fluctuating number of hydrogen bonds throughout the simulation, rarely exceeding two concurrent HBs. This suggests transient or weak interactions that may not be sufficient to ensure long-term binding stability. Among the newly designed molecules, A-100 and A-117 demonstrated significantly enhanced and sustained hydrogen bonding. A-100 maintained between 2 and 4 HBs for a considerable duration particularly after 60 ns, indicating a stable interaction interface with the protein. Notably, A-117 exhibited the most consistent and highest HB count. It frequently maintains 3 to 5 hydrogen bonds, with peaks up to 6. This strong and persistent hydrogen bonding network suggests a more robust and favorable binding mode within the 3CLpro active site.

Close

Figure 7.

(a) Time-dependent RMSD of c-α backbone of the 3CLpro-apo, 3CLpro-Baicalein, 3CLpro-D-119, 3CLpro-A-100, 3CLpro-A-117, 3CLpro -Lead-51. (b) The RMSF for c-α atoms of 3CLpro-apo, 3CLpro-Baicalein, 3CLpro-D-119, 3CLpro-A-100, 3CLpro-A-117, 3CLpro -Lead-51. (c) Plot of Rg vs. time for 3CLpro-apo, 3CLpro-Baicalein, 3CLpro-D-119, 3CLpro-A-100, 3CLpro-A-117, 3CLpro -Lead-51. (d) The SASA of 3CLpro-apo, 3CLpro-Baicalein, 3CLpro-D-119, 3CLpro-A-100, 3CLpro-A-117, 3CLpro -Lead-51. (e) HB of 3CLpro-apo, 3CLpro-Baicalein, 3CLpro-D-119, 3CLpro-A-100, 3CLpro-A-117, 3CLpro -Lead-51.

Export to PPT