Unveiling nicotiflorin analogs with improved oral drug-like metrics as candidates for the development of dengue virus NS2B-NS3 protease inhibitors: An in silico study

Abstract

Kaempferol 3-O-β-rutinoside (nicotiflorin) has been identified as a dengue virus (DENV) NS2B-NS3 protease (NS2B-NS3pro) inhibitor. However, nicotiflorin possesses suboptimal oral drug-like metrics regarding violation of the Lipinski’s rule of five (Ro5) and low in silico-derived bioavailability score, rendering it a less favorable lead candidate for the development of oral drugs. Here, ligand-based virtual screening, structure-based virtual screening, and molecular dynamics (MD) simulation approaches were applied to identify nicotiflorin analogs with improved oral drug-like metrics, compared to those of nicotiflorin, as leads for the development of anti-DENV inhibitors targeting NS2B-NS3pro. The SwissSimilarity server was used to search chemical databases for nicotiflorin analogs, which were subsequently combined to form a small molecule library. The library was screened against DENV-2 NS2B-NS3pro, where the fitness score of nicotiflorin (control) was used to set the selection cutoff for shortlisting six compounds as screening hits. Three compounds (cpd1, cpd2, and cpd3) that remained stably bound to NS2B-NS3pro for up to 100 ns in MD simulations were identified as promising oral drug leads. Cpd1, cpd2, and cpd3 demonstrated effective binding to DENV-2 NS2B-NS3pro and obtained improved ligand efficiency scores of 0.12, 0.13, and 0.11, respectively, compared with a score of 0.09 obtained by nicotiflorin. Unlike nicotiflorin, the three analogs passed the Ro5, and each obtained an acceptable in silico-derived bioavailability score of 0.55, which is threefold better than the 0.17 score for nicotiflorin. Furthermore, the three nicotiflorin analogs exhibited RMSDs, H-bonding, binding free energies, and ADMET properties comparable to those of nicotiflorin. The overall results suggest that cpd1, cpd2, and cpd2 possess improved oral drug-like metrics compared to those of nicotiflorin and could serve as promising leads for the development of oral anti-DENV inhibitors targeting NS2B-NS3pro.

Keywords

kaempferol 3-O-β-rutinoside nicotiflorin dengue virus NS2B-NS3 protease ligand-based virtual screening structure-based virtual screening molecular dynamics ADMET

1. Introduction

Dengue fever is a mosquito-borne neglected tropical disease caused by four serotypes of the dengue virus (DENV): DENV-1, DENV-2, DENV-3, and DENV-4.^1,2 According to the World Health Organization (WHO), half of the world’s population is at risk of dengue, with approximately 100–400 million infections estimated annually.³ Global dengue statistics from the European Center for Disease Prevention and Control (ECDC) indicate over 4 million cases and 2,500 fatalities in the first half of 2025 across 101 countries.⁴ The typical first manifestations of dengue fever include elevated temperature, intense cephalalgia, retro-orbital discomfort, myalgia, arthralgia, nausea, emesis, lymphadenopathy, and exanthema. Severe dengue typically presents with symptoms such as intense abdominal pain, continuous vomiting, hemorrhaging from the gums or nostrils, and profound fatigue.^5–7 Unfortunately, no drug has been approved for dengue fever; therefore, treatment emphasizes supportive care to alleviate symptoms.^8–10 Notwithstanding, several essential DENV structural proteins (e.g., envelope and capsid proteins) and non-structural proteins (e.g., NS1, NS3, NS5, NS4B, and NS2B-NS3 protease) have been identified as drug targets that could be exploited to treat dengue infection.^11,12

The NS2B-NS3 protease (NS2B-NS3pro) of DENV is an important component of the viral life cycle, as it is responsible for cleaving viral polyproteins into functional proteins. NS2B-NS3pro is a serine protease consisting of the NS2B cofactor and the NS3 domain, which harbors the active site, catalytic triad, and protease activity (Figure 1). The NS2B cofactor is essential for activating the protease activity of the NS3 domain by inducing an active closed conformation.^13–18 NS2B-NS3pro cleaves a peptide bond after two dibasic amino acids, recognizing specific motifs such as Arg-Arg, Lys-Arg, or Arg-Lys at the P1 and P2 positions of the polyprotein substrate. Cleavage typically occurs before a smaller amino acid, such as Gly, Ser, or Ala, at the P1’ position.¹⁵ The overall sequence of NS2B-NS3pro is approximately 63%-74% conserved among the four DENV serotypes, and the catalytic triad of the active site, consisting of Ser-His-Asp amino acid residues, which forms the main catalytic machinery, is fully conserved.¹⁹ Based on the aforementioned features, NS2B-NS3pro, specifically the active site, has been recognized as a drug target for the discovery, design, and development of inhibitors against DENV.^20–24 Such inhibitors may function by competing with the polyprotein substrate for binding to the active site based on affinity. Thus, inhibitors with superior affinity may exclude the polyprotein substrate from binding to the active site, thereby impeding its cleavage into the viral proteins required for replication and proliferation. The lack of functional proteins truncates the viral life cycle, leading to inhibition or death. Despite its attractiveness as a drug target, no approved NS2B-NS3pro drugs are currently available.

Dwivedi et al.²⁵ identified the bioflavonoid kaempferol 3-O-β-rutinoside (nicotiflorin) as a DENV inhibitor targeting NS2B-NS3pro, exhibiting dose-dependent activity against DENV-2, with 77.7% viral inhibition at 100 μM, and without significant cytotoxicity. However, nicotiflorin possesses suboptimal oral drug-like metrics regarding violation of Lipinski’s rule of five (Ro5) and a low in silico-derived bioavailability score, rendering it a less favorable lead candidate for the development of oral drugs. Oral drugs are generally considered cheaper to administer than intravenous or injectable alternatives, primarily because they do not require specialized medical staff, facility resources, or complex sterile administration, making them particularly appropriate for the predominantly poor and less privileged dengue-endemic regions.²⁶ Nicotiflorin violates three of the Ro5, including molecular weight (MW), number of hydrogen bond (H-bond) donors (nHD), and number of H-bond acceptors (nHA). Furthermore, nicotiflorin has an in silico-derived bioavailability score of 0.17, which is very low compared to the minimum requirement of 0.55.²⁷ In this study, in silico techniques, including ligand-based virtual screening (LBVS), structure-based virtual screening (SBVS), and molecular dynamics (MD) simulations, were applied to identify nicotiflorin analogs with improved oral drug-like metrics, compared to those of nicotiflorin, as leads for the development of anti-DENV inhibitors targeting NS2B-NS3pro. The SwissSimilarity web-based server^28,29 was used to search chemical databases for analogs of nicotiflorin, which were subsequently combined to form a small molecule library. The small-molecule library of nicotiflorin analogs was screened against DENV-2 NS2B-NS3pro, where the fitness score of nicotiflorin (control) was used to set the selection cutoff for shortlisting six compounds as screening hits. Three compounds (cpd1, cpd2, and cpd3) that remained stably bound to NS2B-NS3pro for up to 100 ns in MD simulations were identified as promising oral drug leads. Cpd1, cpd2, and cpd3 demonstrated effective binding to DENV-2 NS2B-NS3pro and obtained improved ligand efficiency scores of 0.12, 0.13, and 0.11, respectively, compared with a score of 0.09 obtained by nicotiflorin. Unlike nicotiflorin, the three analogs passed the Ro5, and each obtained an acceptable in silico-derived bioavailability score of 0.55, which is threefold better than the 0.17 score for nicotiflorin. Furthermore, the three nicotiflorin analogs exhibited RMSDs, H-bonding, binding free energies (BFEs), and ADMET properties comparable to those of nicotiflorin. The overall results suggest that cpd1, cpd2, and cpd2 possess improved oral drug-like metrics compared to those of nicotiflorin and could serve as promising leads for the development of oral anti-DENV inhibitors targeting NS2B-NS3pro.

2. Methods

2.1. Retrieval and preparation of NS2B-NS3pro receptor model

The three-dimensional (3D) coordinates of the NS2B-NS3pro (PDB ID: 2FOM³⁰) of the DENV-2 serotype, with the NS2B cofactor as chain A and the NS3 protease as chain B, were retrieved from the Protein Data Bank (PDB). 2FOM was selected because the structure belongs to DENV-2 (frequently linked to severe outbreaks) and consists of both the NS3 protease and the NS2B cofactor.^30–35 The targeted active site, particularly the catalytic triad of the protease, is conserved among the other four DENV serotypes.²⁴ The SWISS-MODEL server^36,37 was used to reconstruct missing nonterminal amino acid residues and atoms after water molecules were removed from the structure. The NS2B-NS3pro structure was protonated and energy-optimized using the protein preparation tool of BIOVIA Discovery Studio 2016 and saved as a mol2 file.

2.2. Ligand-based virtual screening and construction of nicotiflorin-analogs library

LBVS was performed to identify structural analogs of nicotiflorin using the SwissSimilarity server.^28,29 Using nicotiflorin as the reference input structure, drugs, bioactives, and commercial and synthesizable small-molecule databases were screened using 2D (FP2, ECFP4, MHFP6, Pharmacophore, ErG, Scaffold, and Generic Scaffold), 3D (Electroshape and E3FP), and, where applicable, combined (2D and 3D) methods. The SMILES information of all compounds with a nicotiflorin similarity score of ≥ 0.6 was merged and converted into a 3D SDF file containing 1967 entries. Using the ligand preparation tool in BIOVIA Discovery Studio 2016, the compounds were processed, which involved protonation, energy minimization, and removal of duplicates, leading to the generation of 1357 prepared compounds. To identify lead compounds with improved drug-like properties compared with those of nicotiflorin, the 1357 compounds were filtered based on Ro5,³⁸ allowing only one violation of the rule. Finally, a small-molecule library of 147 nicotiflorin analogs was generated and saved as a mol2 file. Nicotiflorin was prepared and added to the library as a control.

2.3. GOLD software and validation

The GOLD (Genetic Optimization for Ligand Docking) program^38–40 was used for the docking and VS processes. GOLD adopts a genetic algorithm method to predict ligand poses within the binding site of a receptor and uses one of four scoring functions, including CHEMPLP, GoldScore, ChemScore, and ASP, to rank the predicted poses. The GoldScore function ranks the ligand pose by generating a GOLD fitness score (Equation (1)), which is the sum of the receptor-ligand H-bond (S_{hb_int}), receptor-ligand van der Waals (S_{vdw_ext}), ligand intramolecular H-bond (S_{hb_int}), which is disabled by default, and ligand intramolecular strain (S_{vdw_int}) scores. The suitability of the GoldScore function for predicting ligand poses against DENV NS2B-NS3pro was ascertained by assessing its ability to reproduce the experimental pose of a peptide bound to the active site of DENV3 NS2B-NS3pro (PDB ID: 3U1I). After removing the peptide and redocking it onto the active site using the GoldScore function, the NS2B-NS3pro-peptide complex of the best predicted binding mode was superimposed on the X-ray crystallographic structure to compare the ligand poses. As shown in Figure 2, where the carbon atoms of the experimental and docking poses are colored green and orange, respectively, the superimposition generated an all-ligand atom root mean square deviation (RMSD) of 0.90 Å, which is indicative of the ability of the GOLD program to regenerate the experimental pose and is thus suitable for this study.

G O L D F i t n e s s S c o r e = S_{h b_e x t} + S_{v d w_e x t} + S_{h b_int} + S_{v d w_int}

(1)

2.4. Structure-based virtual screening and hits selection

After loading the prepared NS2B-NS3pro structure onto the GOLD wizard, the amino acid residues Ile36(B), Gly37(B), Trp50(B), His51(B), Val52(B), Val72(B), Lys73(B), Asp75(B), Leu128(B), Phe130(B), Ser131(B), Pro132(B), Gly133(B), Ser135(B), Tyr150(B), Gly151(B), Asn152(B), Gly153(B), Val154(B), Val155(B), and Tyr161(B) were used to define the binding site (Figure 1) for the VS exercise. The prepared library, including the control, was uploaded to the GOLD wizard and screened against NS2B-NS3pro using GoldScore as the scoring function and the default settings for all other parameters. Prior to hit selection, the poses of the control were examined, resulting in the identification of two unique binding modes, which were saved as PDB files and subjected to stability analysis using MD simulations.

To select hit compounds from the SBVS process, the bestranking. lst file, which contained the top-ranking pose of each ligand, was sorted in decreasing order based on the GOLD fitness scores. The fitness score of nicotiflorin, which served as the control, was used as the cutoff threshold for selection, where all ligands above it were selected as hits and those below were rejected. Six compounds, namely, compound 1 (cpd1), compound 2 (cpd2), compound 3 (cpd3), compound 4 (cpd4), compound 5 (cpd5), and compound 6 (cpd6), met the selection criterion and emerged as VS hits. For each hit, the NS2B-NS3pro-ligand complexes of all predicted docking poses with fitness scores greater than the control were saved as PDB files for further analysis and shortlisting using MD simulations. This idea of assessing many predictions of a docking program to determine the best binding mode is recommended, considering that the top-ranked pose may not always be accurate.⁴¹ As illustrated in Supplemental Figure S1, two and three poses were obtained for cpd1 and cpd2, respectively, whereas one pose was obtained for each of cpd3, cpd4, cpd5, and cpd6. Two unique binding modes of nicotiflorin were also observed. The NS2B-NS3pro-ligand complexes were then subjected to MD simulations categorized into phases I, II, and III, distinguished by simulation times of 20, 50, and 100 ns, respectively. The progression of a complex to the next phase was dictated by ligand stability, which was assessed using RMSD and the simulated pose relative to the initial complex. Compounds that remained stably bound to NS2B-NS3pro until phase III (100 ns) were selected as the final hits. The all-atom RMSD of each ligand was calculated as a function of time following least-squares fitting of all the atoms of NS2B-NS3pro using the gmx rms tool in GROMACS. In each case, the ligand pose in the final simulation structure was used as the simulation pose for comparison with the docking pose via superimposition. The RMSDs and poses of the six compounds are shown in Supplemental Figures S2 and S3, respectively. Three compounds, cpd1, cpd2, and cpd3, emerged as the final hits and potential anti-NS2B-NS3pro candidates. The detailed procedure for selecting the final hits based on MD simulations is presented in the Supplementary Material. A summary of the overall procedure for identifying cpd1, cpd2, and cpd3 as hits is presented in Figure 3.

2.5. Molecular dynamics simulations

The GROMACS 2025.1 simulation suite^42,43 was used to perform MD simulations. In each instance, the gmx pdb2gmx tool, along with the Amber 99SB force field⁴⁴ and TIP3P water model,⁴⁵ was used to build the topology of NS2B-NS3pro. The antechamber utility in the AMBER 21 suite^46,47 was used to construct the topology of the compound and assign atom types and restraint electrostatic potential (RESP) charges.⁴⁸ The NS2B-NS3pro-compound complex or apo form of NS2B-NS3pro was placed at the midpoint of a triclinic box, ensuring a minimum separation distance of 12 Å between the NS2B-NS3pro surface and the edges of the box. After solvation of the simulation box with an explicit TIP3P water model,⁴⁵ the net charge of the system was neutralized, ensuring an ionic concentration of 0.1 M by adding the necessary amounts of Na⁺ and Cl^- ions. Long-range charge-charge interactions were resolved using the Particle Mesh Ewald (PME) method,⁴⁹ whereas all bonds that involved hydrogen atoms were restrained at the constant equilibrium lengths using the linear constraint solver (LINCS) algorithm.⁵⁰ The steepest descent algorithm was used to minimize the energy of the system. Subsequent to applying position restraints to the non-hydrogen atoms of NS2B-NS3pro and the compound, initial velocities were randomly assigned to all atoms of the system. An equilibration run of 5 ns at 1 bar and 300 K in the NPT ensemble was performed using the C-rescale barostat for pressure coupling and V-rescale thermostat for temperature coupling.⁵¹ After removing the position restraints, a production run was continued in the NPT ensemble up to the required simulation time, with data saved every 1 ps.

2.6. Binding free energy (BFE) and per-residue contribution (PRC) calculations

The g_mmpbsa tool⁵² was used to calculate the BFEs and per-residue contributions (PRCs) of NS2B-NS3pro to the BFEs of NS2B-NS3-ligand complexes. The tool computes BFE as the sum of four energy terms (Equation (2)): the van der Waals energy (E_vdw), electrostatic energy (E_elec), electrostatic contribution to the solvation free energy (G_polar), and nonelectrostatic contribution to the solvation free energy (G_non-polar).

B F E = E_{v d w} + E_{e l e c} + G_{p o l a r} + G_{n o n - p o l a r}

(2)

2.7. ADMET analyses

The ADMETlab 3.0 server^53,54 was used to perform ADMET analyses on nicotiflorin, cpd1, cpd2, and cpd3. Acetaminophen, commonly prescribed for managing fever and pain associated with dengue fever,⁵⁵ was included in the ADMET analyses for comparison, although it neither targets NS2B-NS3pro nor inhibits DENV. The selection of properties for assessing the absorption, distribution, metabolism, excretion, and toxicity of the drug candidates was based on those commonly reported in the DrugBank.

3. Results and discussion

3.1. Docking scores, ligand efficiencies, and chemical structures of nicotiflorin and the virtual screening hits

Following the LBVS and SBVS using nicotiflorin as the reference ligand and control molecule, six nicotiflorin analogs were identified as hits: cpd1 (HMDB0040803), cpd2 (ZINC000002554914), cpd3 (CHEMBL470048), cpd4 (CHEMBL2326706), cpd5 (ZINC000071789593), and cpd6 (CHEBI169100). GOLD fitness scores of 53.81, 57.18, 54.28, 54.47, 59.00, 60.25, and 52.44 were obtained by cpd1, cpd2, cpd3, cpd4, cpd5, cpd6, and nicotiflorin, respectively (Table 1). As shown in Table 1, all compounds obtained better fitness, H-bond, and LE scores than nicotiflorin. Nicotiflorin obtained a better Van der Waals interaction score than the six compounds owing to its high molecular weight. Despite docking scores potentially favoring compounds with high molecular weights,⁵⁶ the six nicotiflorin analogs, irrespective of their low molecular weight compared with nicotiflorin, obtained better fitness scores (Table 1). Cpd1, cpd2, cpd3, cpd4, cpd5, and cpd6 obtained superior LE scores of 0.12, 0.13, 0.11, 0.12, 0.12, and 0.13, respectively, compared with 0.09 for nicotiflorin. The LE, which was calculated as the fitness score divided by the molecular weight, is a normalized measure of how effectively a compound binds to a target relative to its molecular size. The fitness and LE scores of the six nicotiflorin analogs, compared with those of nicotiflorin, suggest that they effectively bind to NS2B-NS3pro and possess improved lead-likeness.⁵⁷

Table 1.

Docking scores, molecular weights, and ligand efficiencies of nicotiflorin and its analogs.

Compound	Fitness score (FS)	H-bond score	Van der waals score	Molecular weight (MW) (g/mol)	Ligand efficiency (FS/MW)
Cpd1	53.81	11.22	33.77	446.40	0.12
Cpd2	57.18	7.84	38.52	446.40	0.13
Cpd3	54.28	4.51	37.88	474.41	0.11
Cpd4	54.47	5.71	36.66	474.15	0.12
Cpd5	59.00	10.00	36.43	474.12	0.12
Cpd6	60.25	10.84	37.66	474.12	0.13
Nicotiflorin	52.44	2.50	42.52	594.52	0.09

The chemical structures of nicotiflorin and its identified analogs are shown in Figure 4. Nicotiflorin is a plant-derived flavonoid glycoside consisting of the flavonoid kaempferol linked to a rutinoside sugar moiety (Figure 4(a)) and possesses several medicinal properties.⁵⁸ All the selected nicotiflorin analogs possess the kaempferol moiety, and similarity scores of 0.997, 0.704, 0.995, 0.997, 0.695, and 0.996 were recorded between nicotiflorin and cpd1, cpd2, cpd3, cpd4, cpd5, and cpd6, respectively. The RMSD and pose analysis results, as illustrated in Supplemental Figures S2 and S3, respectively, show that cpd1, cpd2, and cpd3 maintained stable binding to NS2B-NS3pro for 100 ns during the simulations. Consequently, the remainder of this study focused on cpd1, cpd2, and cpd3 as potential inhibitors of DENV NS2B-NS3pro with improved lead and drug likeness. Thus, compared with nicotiflorin, cpd1, cpd2, and cpd3 represent better starting points for optimization to generate more effective DENV NS2B-NS3pro inhibitors, reducing the risk of failure later in development.

3.2. Chemical interactions between NS2B-NS3pro and nicotiflorin/nicotiflorin analogs

The BIOVIA Discovery Studio 2016 software was used to analyze ligand binding modes and interactions with NS2B-NS3pro. The NS2B-NS3pro-cpd1, NS2B-NS3pro-cpd2, and NS2B-NS3pro-cpd3 complexes involve several chemical interactions, as illustrated in Figure 5. As illustrated in Figure 5(a), nicotiflorin forms H-bonds with the amino acid residues Val72(B), Phe130(B), and Gly153(B) of NS2B-NS3pro. The NS2B-NS3pro residues, including Trp50(B), His51(B), Lys73(B), Lys74(B), Asp75(B), Leu128(B), Ser131(B), Pro132(B), Ser135(B), Tyr150(B), Gly151(B), Asn152(B), Val154(B), and Tyr161(B), engage in numerous nonbonded interactions with nicotiflorin. Figure 5(b) shows that cpd1 forms H-bonds with Asp75(B), Phe130(B), Ser135(B), and Gly153(B) of NS2B-NS3pro. A π-π (T-shaped) stacking is also formed between Tyr161(B) of NS2B-NS3pro and cpd1. Cpd1 also makes nonbonded interactions and contacts with the amino acid residues His51(B), Leu128(B), Ser131(B), Pro132(B), Tyr150(B), Gly151(B), and Asn152(B) of NS2B-NS3pro. Cpd2 (Figure 5(c)) forms H-bonds with residues Asp75(B), Ser135(B), Tyr150(B), Gly151(B), and Gly153(B) and establishes nonbonded interactions and contacts with His51(B), Leu128(B), Asp129(B), Phe130(B), Ser131(B), Pro132(B), and Asn152(B) of NS2B-NS3pro. Regarding cpd3, H-bonds are formed with the amino acid residues Asp75(B), Phe130(B), Ser135(B), and Gly153(B) of NS2B-NS3pro, as illustrated in Figure 5(d). Residues Trp50(B), His51(B), Val72(B), Leu128(B), Ser131(B), Pro132(B), Tyr150(B), Gly151(B), Asn152(B), and Tyr161(B) of NS2B-NS3pro establish nonbonded interactions and contacts with cpd3. Importantly, cpd1, cpd2, and cpd3 interact with the catalytic residues of NS2B-NS3pro, namely His51, Asp75, and Ser135, and are likely to exhibit significant potency and selectivity by directly blocking the enzyme mechanism rather than merely hindering substrate binding.

3.3. Ligand RMSD and simulation convergence analyses

Three independent simulations were performed for the NS2B-NS3pro-nicotiflorin, NS2B-NS3pro-cpd1, NS2B-NS3pro-cpd2, and NS2B-NS3pro-cpd3 complexes. The RMSD of each compound was calculated in all three simulations of the corresponding system and was used to ascertain ligand stability. Using the gmx rms tool in GROMACS, the all-atom RMSD of each compound was calculated as a function of the simulation time following least-squares fitting of all the atoms of NS2B-NS3pro. The RMSD data are shown in Figure 6, where the black, red, and violet curves indicate the results for three independent simulations. The results for each ligand, as shown in Figure 6, indicate that the paths of the three RMSD trajectories are similar, indicating the replicability of the simulation data. Furthermore, it can be observed that in each case, the RMSD trajectories remained steady from at least 50 ns of simulation time until the end. This phenomenon suggests that the simulations reached equilibrium by 50 ns; thus, it is reasonable to rely on the data within the 50-100 ns range of the simulations to make inferences and conclusions. The RMSD of each ligand was calculated after designating the initial NS2B-NS3pro-ligand complex structure as the reference and selecting the atoms of NS2B-NS3 for least-squares fitting, allowing the monitoring of the relative distance between the ligand and the binding site over the simulation period. By using the results of the last 50 ns of the three trajectories of each compound to calculate average RMSDs, values of 2.85 ± 0.29 Å, 2.08 ± 0.13 Å, 2.57 ± 0.16 Å, and 2.97 ± 0.15 Å were recorded for nicotiflorin, cpd1, cpd2, and cpd3, respectively (Table 2). Overall, similar to nicotiflorin, all three selected compounds had average RMSDs < 3.20 Å, indicating stability. Considering the error margins, the average RMSDs of nicotiflorin, cpd1, cpd2, and cpd3 (2.85 ± 0.29 Å, 2.08 ± 0.13 Å, 2.57 ± 0.16 Å, and 2.97 ± 0.15 Å, respectively) indicate that cpd1 is the most stable, whereas the stabilities of cpd2 and cpd3 are comparable to that of nicotiflorin.

Table 2.

Average parameters for the last 50 ns of three independent simulations.

Analysis	NS2B-NS3-nicotiflorin		NS2B-NS3-Cpd1		NS2B-NS3-Cpd2		NS2B-NS3-Cpd3		Apo
Analysis	NS2B-NS3	Nicotiflorin	NS2B-NS3	Cpd1	NS2B-NS3	Cpd2	NS2B-NS3	Cpd3	NS2B-NS3
RMSD (Å)	3.24 ± 0.26	2.85 ± 0.29	3.19 ± 0.25	2.08 ± 0.13	3.35 ± 0.29	2.57 ± 0.16	2.85 ± 0.29	2.97 ± 0.15	2.65 ± 0.20
Rg (Å)	16.78 ± 0.15	-	16.68 ± 0.19	-	16.58 ± 0.16	-	16.40 ± 0.15	-	16.54 ± 0.11
RMSF (Å)	0.92 ± 0.07	-	0.80 ± 0.01	-	0.86 ± 0.09	-	0.81 ± 0.05	-	0.87 ± 0.04
H-Bond	2.95 ± 0.18		2.49 ± 0.35		1.27 ± 0.08		1.74 ± 0.20		-

3.4. Hydrogen bond analyses

The evolution of H-bond count between a protein and ligand as a function of simulation time can be used to evaluate the strength and stability of the interaction. A stable binding mode or interaction is defined by a consistent H-bond count, whereas an unstable binding mode or interaction results in a variable H-bond count. To ascertain the evolution of H-bonds between NS2B-NS3pro and the compounds (nicotiflorin, cpd1, cpd2, and cpd3) during the simulations, the H-bond count was computed as a function of the simulation time for each system. The calculations were performed using the default donor-acceptor angle of 30° and donor-acceptor separation distance of 3.5 Å of the gmx hbond utility in GROMACS. The results of the H-bond calculations are shown in Figure 7, where the black, red, and violet curves in each case represent the H-bond counts from three independent simulations. The consistent H-bond count as a function of time in each instance suggests the stability of the interactions between NS2B-NS3pro and the compounds. The average H-bond values for the NS2B-NS3pro-nicotiflorin, NS2B-NS3pro-cpd1, NS2B-NS3pro-cpd2, and NS2B-NS3pro-cpd3 complexes were calculated using the last 50 ns H-bond counts obtained from the three independent simulations of each system. Average H-bond values of 2.95 ± 0.18, 2.49 ± 0.35, 1.27 ± 0.08, and 1.74 ± 0.20 were obtained for the NS2B-NS3pro-nicotiflorin, NS2B-NS3pro-cpd1, NS2B-NS3pro-cpd2, and NS2B-NS3pro-cpd3 complexes, respectively (Table 2). The NS2B-NS3pro-nicotiflorin and NS2B-NS3pro-cpd1 complexes obtained average H-bond counts > 2, and the average values overlap, considering the standard errors. The NS3pro-cpd2 and NS2B-NS3pro-cpd3 complexes, on the other hand, obtained average H-bond values < 2; however, considering the standard errors, the average H-bond count for the NS2B-NS3pro-cpd3 complex is slightly greater than that of the NS3pro-cpd2 complex.

3.5. Receptor backbone RMSD, Rg, and RMSF analyses

The backbone RMSD, radius of gyration (Rg), and root mean square fluctuation (RMSF) of the nicotiflorin-bound, cpd1-bound, cpd2-bound, and cpd3-bound NS2B-NS3pro were calculated using the gmx rms, gmx gyrate, and gmx rmsf utilities in GROMACS, respectively, and compared with those of the apo form of NS2B-NS3pro to ascertain the effect of the ligands on receptor conformation. These metrics were calculated using the trajectories of all three independent simulations for each system. Backbone RMSDs and Rgs of NS2B-NS3pro were calculated using the CC_αN atoms, whereas backbone RMSFs were calculated using only the C_α atoms. For the NS2B-SN3pro backbone RMSD analyses, the backbone atoms were least-squares fitted, and the calculations were executed on the same atoms. The backbone RMSD estimates the conformational difference by measuring the deviation of each explored structure from the initial structure during the simulations, whereas the backbone Rg determines the overall compactness of each generated structure. The RMSF, on the other hand, determines the average fluctuation of the atoms of the receptor, which could indicate the most flexible regions. The NS2B-NS3pro backbone RMSD, Rg, and RMSF results are shown in Supplemental Figures S4–S6, respectively. It is worth noting that the backbone RMSFs of only the ligand-binding/catalytic NS3 domain are shown in Supplemental Figure S6. For comparison, stable curves in each category of the calculations are presented in Figure 8, where curves colored cyan, magenta, blue, yellow, and green represent the results for the nicotiflorin-bound, cpd1-bound, cpd2-bound, cpd3-bound, and apo forms of NS2B-NS3pro, respectively.

The last 50 ns trajectories of the three simulations of each system were used to calculate the average NS2B-NS3pro backbone RMSD, Rg, and RMSF. Average backbone RMSDs of 3.24 ± 0.26 Å, 3.19 ± 0.25 Å, 3.35 ± 0.29 Å, 2.85 ± 0.29 Å, and 2.65 ± 0.20 Å were obtained for the nicotiflorin-bound, cpd1-bound, cpd2-bound, cpd3-bound, and apo forms of the receptor, respectively (Table 2). Regarding the average Rg, the nicotiflorin-bound, cpd1-bound, cpd2-bound, cpd3-bound, and apo forms of NS2B-NS3pro recorded values of 16.78 ± 0.15 Å, 16.68 ± 0.19 Å, 16.58 ± 0.16 Å, 16.40 ± 0.15 Å, and 16.54 ± 0.11 Å, respectively (Table 2). The NS3 domain of the nicotiflorin-bound, cpd1-bound, cpd2-bound, cpd3-bound, and apo forms of NS2B-NS3pro obtained RMSF values of 0.92 ± 0.07 Å, 0.80 ± 0.01 Å, 0.86 ± 0.09 Å, 0.81 ± 0.05 Å, and 0.87 ± 0.04 Å, respectively (Table 2). All ligand-bound forms of NS2B-NS3pro, with the exception of the cpd3-bound form, exhibit average backbone RMSDs that are marginally higher compared with the backbone RMSD of the apo form. Compared with the apo form of the NS2B-NS3pro, no significant variations in the average backbone Rg and RMSF values of the ligand-bound forms are observed. Ligand binding may induce conformational changes in the receptor or alter its dynamics, which could help explain the mechanism of action of the inhibitor.⁵⁹ However, ligand-induced conformational changes or dynamics that involve unfolding the native structure of the receptor may indicate an unstable receptor-ligand complex or unfavorable interactions between the receptor and ligand. The overall backbone RMSD, Rg, and RMSF results suggest that the binding of nicotiflorin, cpd1, cpd2, and cpd3 to NS2B-NS3pro did not induce any major conformational changes that may alter its function. The results further suggest that the complexes formed by NS2B-NS3pro and the compounds are stable, as no conformational distortions are observed in the NS2B-NS3pro.

3.6. Free energy surface and ligand-binding mode analyses

To determine the most energetically favorable ligand-binding modes explored during the simulations, free energy surface (FES) analyses were performed on nicotiflorin, cpd1, cpd2, and cpd3 using the gmx covar and gmx anaeig utilities in GROMACS. To ensure adequate sampling of ligand-binding modes, a 150 ns trajectory generated by combining the final 50 ns trajectories of the three simulations of each system was used for the analyses. In each case, the FES calculations were performed on the ligand after least-squares fitting of all the atoms of NS2B-NS3pro. Eigenvectors, which served as gmx anaeig input data for the projection of principal components 1 (PC1) and 2 (PC2), were generated using the gmx covar tool. FES diagrams were then created using the PC1-PC2 projections generated by gmx anaeig. The FES diagrams for nicotiflorin, cpd1, cpd2, and cpd3 are presented in Figure 9(a)–(d), respectively. In the FES diagrams, the purple regions represent the energy minima, which indicate the favorable binding modes of the compounds to NS2B-NS3pro. As shown in Figure 9, the FES diagrams of nicotiflorin (Figure 9(a)) and cpd3 (Figure 9(d)) are each characterized by a distinct energy minimum. In contrast, the FES diagrams of cpd1 (Figure 9(b)) and cpd2 (Figure 9(c)) each display two energy minima, indicated as bins 1 and 2, with varying population densities and energies. The purple and blue colors of bins 1 and 2, respectively, in the FES diagram of cpd1 indicate that the binding mode in bin 1 is more energetically favorable than that in bin 2. The FES diagram of cpd2 shows that the binding modes in bin 1 and bin 2 have similar energies; however, bin 2 is more populated and therefore more representative than bin 1.

To evaluate the ligand-binding modes explored during the simulations and ascertain the stability of the docking poses, complexes of NS2B-NS3pro and each of nicotiflorin, cpd1, cpd2, and cpd3 were retrieved from the corresponding bins in Figure 9 and compared with the docking structures via all-ligand atom superimposition and RMSD calculations. By superimposing the simulation pose of nicotiflorin (cyan sticks) on the docking pose (green sticks), an RMSD of 1.75 Å was obtained (Figure 10(a)). The superimposition of the two simulation poses of cpd1, extracted from bins 1 and 2 of the FES diagram, onto the docking pose yielded RMSDs of 0.71 Å and 0.59 Å, respectively. The results of the superimpositions involving cpd1 are shown in Figure 10(b), where the docking pose is shown as green sticks and the simulation poses extracted from bins 1 and 2 are illustrated as magenta and LT-magenta sticks, respectively. As shown in Figure 10(c), the simulation poses (blue and LT-blue sticks) extracted from bins 1 and 2 of the FES diagram of cpd2 yielded RMSDs of 1.66 Å and 1.73 Å, respectively, after superimposition on the docking pose (green sticks). An RMSD of 1.23 Å is recorded (Figure 10(d)) after superimposing the simulation pose (yellow sticks) of cpd3, extracted from the corresponding FES diagram, onto the docking pose (green sticks). Overall, the RMSD values yielded by all the superimpositions are < 2 Å, indicating that the predicted binding modes of nicotiflorin, cpd1, cpd2, and cpd3 for NS2B-NS3pro are stable. The parallels between the simulation and docking poses (Figure 10) corroborate the stability of the compounds.

3.7. Binding free energy and per-residue contribution analyses

The BFEs and PRCs of NS2B-NS3pro to the BFEs of NS2B-NS3-ligand complexes were calculated using the g_mmpbsa tool.⁵² In each instance, 1000 frames were extracted from a 150 ns trajectory generated by merging the last 50 ns of the three independent simulations of the system. An optimal dielectric constant of 6 and default settings for all other parameters of the tool were used for the calculations. BFE calculations on the NS2B-NS3pro-nicotiflorin, NS2B-NS3pro-cpd1, NS2B-NS3pro-cpd2, and NS2B-NS3pro-cpd3 complexes produced values of -106.34 ± 23.791 kJ/mol, -84.64 ± 15.94 kJ/mol, -92.16 ± 16.74 kJ/mol, and -79.98 ± 15.72 kJ/mol, respectively. The PRCs of NS2B-NS3pro in relation to the BFEs of the NS2B-NS3pro-nicotiflorin, NS2B-NS3pro-cpd1, NS2B-NS3pro-cpd2, and NS2B-NS3pro-cpd3 complexes are illustrated in Figure 11(a)–(d), respectively. In Figure 11, only amino acid residues with favorable energy contributions ≤ -1.5 kJ/mol and unfavorable energy contributions ≥ +1.5 kJ/mol are labeled. The amino acid residues His51, Lys74, Leu128, Pro132, Asn152, and Tyr161 of NS2B-NS3pro made favorable contributions, whereas Arg54, Asp75, and Phe130 made unfavorable contributions to the BFE of the NS2B-NS3pro-nicotiflorin interaction, as shown in Figure 11(a). As illustrated in Figure 11(b), His51, Leu128, Pro132, Tyr150, Val155, and Tyr161 of NS2B-NS3pro made favorable contributions to the BFE of the NS2B-NS3pro-cpd1 complex. The residues Arg54, Asp75, and Phe130 made unfavorable contributions to the BFE. Regarding the NS2B-NS3pro-cpd2 interaction, the amino acid residues Arg54, Asp75, and Asp129 of NS2B-NS3pro made unfavorable contributions to the BFE, whereas His51, Leu128, Pro132, Tyr150, Gly153, and Tyr161 made favorable contributions, as illustrated in Figure 11(c). The amino acid residues His51, Leu128, Pro132, Tyr150, Val154, and Tyr161 made favorable contributions to the final BFE of the NS2B-NS3pro-cpd3 complex (Figure 11(d)). The residues Agr54 and Phe130 made minimal unfavorable contributions to the BFE. Taking the standard errors into consideration, the BFEs of the NS2B-NS3pro-cpd1, NS2B-NS3pro-cpd2, and NS2B-NS3pro-cpd3 complexes overlap with that of the NS2B-NS3pro-nicotiflorin complex. These observations suggest that cpd1, cpd2, and cpd3 can stably interact with NS2B-NS3pro and are unlikely to show diminished affinity and activity compared with nicotiflorin.

3.8. Solubility, drug-likeness, and ADMET analyses

3.8.1. Aqueous solubility and druglikeness analyses

The SwissADME server⁶⁰ was used to assess the solubility and drug-likeness of nicotiflorin, cpd1, cpd2, and cpd3. For a candidate drug envisaged for oral administration, solubility is an important property worth considering, as it may significantly affect the absorption, distribution, metabolism, excretion, and toxicity (ADMET), bioavailability, and pharmacokinetic properties. Insoluble drugs exhibit low absorption, resulting in diminished bioavailability upon oral administration.^61–63 The solubility of each compound was assessed based on logS using the ESOL, Ali, and SILICOS-IT models.^60,64,65 All compounds under consideration, including nicotiflorin, cpd1, cpd2, and cpd3, were predicted to be soluble according to the ESOL method, moderately soluble according to the Ali technique, and soluble according to the SILICOS-IT model.

The drug-likeness of each of nicotiflorin, cpd1, cpd2, and cpd3 was evaluated based on the Ro5⁶⁶ and the Abbott bioavailability score.²⁷ The Ro5 is a guideline used to evaluate the drug-likeness of a small organic molecule intended for oral administration, considering five key properties: molecular weight (MW), logP, number of H-bond donors (nHD), number of H-bond acceptors (nHA), number of rotatable bonds (nRot), and polar surface area (PSA), which influence the ADME properties of a drug. The rule suggests that a candidate molecule has a higher chance of possessing good oral pharmacokinetics if it meets the following criteria: MW ≤ 500, logP ≤ 5, nHA ≤ 10, nHD ≤ 5, nRot ≤ 10, and PSA ≤ 140 Å, where one violation is often regarded as acceptable. The Abbott bioavailability score is a measure of the likelihood of a compound having at least 10% oral bioavailability in rats and is based on the physical properties (polarity, size, and lipophilicity) that influence membrane permeability and the Ro5. A bioavailability score of ≥ 0.55 suggests that a drug candidate is ideal and may exhibit acceptable absorption properties. Nicotiflorin fails the Ro5 test with three violations regarding MW, nHA, and nHA (Figure 4(a)), which are greater than the acceptable values. Nicotiflorin also has a bioavailability score of 0.17, which is below the recommended minimum of 0.55. Cpd1 and cpd2 pass the Ro5 with no violations and obtain acceptable bioavailability scores of 0.55 each. Cpd3 violates the nHA rule of the Ro5 (Figure 4(d)) but obtains an acceptable bioavailability score of 0.55. The results suggest that cpd1, cpd2, and cpd3 are suitable candidate drugs for oral administration.

3.8.2. ADMET analyses

3.8.2.1. Absorption

Human intestinal absorption (HIA), Caco-2 permeability, P-glycoprotein (P-gp) inhibition, and P-gp substrate were the four properties used to predict the absorption of nicotiflorin, cpd1, cpd2, and cpd3. Table 3 shows the results, where each attribute was evaluated as “excellent,” “medium,” or “poor,” indicating that the projected property falls within the optimal range (acceptable), is intermediate or average (acceptable), or is outside the ideal range (unacceptable), respectively, for a drug candidate. HIA refers to the process by which orally administered drugs cross the gastrointestinal mucosa to enter the bloodstream, whereas Caco-2 permeability is an in vitro assay that uses human colon cancer cells to simulate the intestinal barrier and measure the rate at which drugs cross it.⁶⁷ Nicotiflorin, cpd1, cpd2, and cpd3 exhibited HIA/Caco-2 permeability ratings of excellent/poor, medium/poor, poor/poor, and excellent/poor, respectively (Table 3). It is generally considered acceptable and promising in drug development to proceed with a compound that has excellent or average HIA but poor Caco-2 permeability. This scenario of a drug candidate possessing excellent or medium HIA but poor Caco-2 permeability frequently indicates that it is likely to be absorbed through carrier-mediated transport rather than passive transcellular diffusion. Caco-2 cells originate from the colon, not the small intestine, and exhibit reduced expression of numerous critical transporter proteins that are abundantly present in the human jejunum; hence, they frequently underestimate the permeability of substances that depend on these transporters. Caco-2 cells model passive transcellular diffusion of lipophilic compounds; thus, a highly hydrophilic drug candidate may have low Caco-2 permeability. Such hydrophilic compounds may be well absorbed if they are substrates for transporter proteins found in the human intestine.⁶⁸ For example, drugs such as gabapentin and sotalol exhibit low Caco-2 permeability; however, they have acceptable in vivo intestinal absorption. These drugs are often characterized by high solubility and low permeability and are classified as Biopharmaceutics Classification System (BCS) Class III drugs.⁶⁹ Drug candidates such as cpd2, characterized by poor Caco-2 permeability and HIA, typically face high development risks, primarily due to extremely low and erratic oral bioavailability, which often causes high rates of clinical trial failure. Nonetheless, such candidates are related to BCS Class IV drugs (low solubility and permeability) and can be optimized via advanced formulations or chemical modifications, such as the use of nanoparticles or prodrug strategies.^70,71 The results suggest that cpd1 and cpd3 are unlikely to face absorption challenges; however, the absorption characteristics of cpd2 may require enhancement to justify further development.

Table 3.

ADMET analyses of nicotiflorin, cpd1, cpd2 and cpd3.

Property	Prediction
Absorption	Nicotiflorin	Cpd1	Cpd2	Cpd3	Acetaminophen
Human Intestinal Absorption	Excellent	Medium	Poor	Excellent	Excellent
Caco-2 permeability	Poor	Poor	Poor	Poor	Excellent
P-glycoprotein inhibitor	Excellent	Excellent	Excellent	Excellent	Excellent
P-glycoprotein substrate	Poor	Poor	Medium	Poor	Medium
Distribution
Blood Brain Barrier	Excellent	Excellent	Excellent	Excellent	Medium
Plasma Protein Binding	Excellent	Excellent	Excellent	Excellent	Excellent
Fraction Unbound in Plasms	Excellent	Excellent	Excellent	Excellent	Excellent
Metabolism
CYP1A2 inhibitor	-	-	-	+	-
CYP1A2 substrate	-	-	+	-	-
CYP2C19 inhibitor	-	-	-	-	-
CYP2C19 substrate	-	-	-	-	+
CYP2C9 inhibitor	-	-	-	-	-
CYP2C9 substrate	-	+	-	+	-
CYP2D6 inhibitor	-	-	-	+	-
CYP2D6 substrate	-	+	-	+	+
CYP3A4 inhibitor	-	-	-	+	-
CYP3A4 substrate	-	-	-	-	+
CYP2B6 inhibitor	-	-	+	-	-
CYP2B6 substrate	-	-	-	-	-
CYP2C8 inhibitor	+	+	+	+	-
Excretion
Plasma Clearance (CL_plasma)	Excellent	Excellent	Excellent	Excellent	Medium
Half-life (T_1/2)	Interm. Short	Short	Short	Short	Short
Toxicity
Carcinogenicity	Excellent	Medium	Excellent	Excellent	Medium
Rat oral acute toxicity	Excellent	Medium	Excellent	Excellent	Excellent
FDAMDD	Excellent	Medium	Excellent	Medium	Medium
hERG blockers	Excellent	Excellent	Excellent	Excellent	Excellent
hERG blockers (10 um)	Excellent	Medium	Excellent	Excellent	Medium
Respiratory toxicity	Excellent	Medium	Excellent	Excellent	Medium
Human Hepatotoxicity	Medium	Medium	Medium	Medium	Medium
Nephrotoxicity	Excellent	Excellent	Poor	Excellent	Medium
Neurotoxicity	Excellent	Excellent	Excellent	Excellent	Medium
Ototoxicity	Poor	Medium	Poor	Medium	Excellent
Hematotoxicity	Excellent	Excellent	Medium	Excellent	Medium
RPMI-8226 Immunotoxicity	Excellent	Excellent	Excellent	Excellent	Excellent
A549 Cytotoxicity	Medium	Excellent	Medium	Medium	Excellent

Excellent: Predicted property is within the ideal/desirable range for a drug candidate (acceptable).

Medium: Predicted property is intermediate or average (acceptable).

Poor: Predicted property is outside the ideal/desirable range for a drug candidate (unacceptable).

+: Inhibitor/substrate.

-: Noninhibitor/nonsubstrate.

Short: 1-4 hours.

Intermediate short: 4-8 hours.

P-gp is an ATP-dependent membrane transporter that acts as a protective efflux pump, removing toxins and xenobiotics from cells. It plays a critical role in pharmacokinetics by limiting drug absorption in the intestine and reducing drug penetration across the blood-brain barrier. P-gp overexpression is a primary mechanism for multidrug resistance.⁷² A drug candidate acting as a P-gp inhibitor primarily helps overcome the mechanism of drug resistance and enhances drug delivery, whereas a drug acting as a P-gp substrate is actively pumped out of cells, preventing it from reaching the bloodstream and intended target. Regarding their status as P-gp inhibitors or substrates, nicotiflorin, cpd1, cpd2, and cpd3 obtained excellent/poor, excellent/poor, excellent/medium, and excellent/poor ratings, respectively (Table 3). A drug candidate that acts as both a P-gp inhibitor and substrate may primarily act as a self-modulating molecule that enhances its absorption and distribution. When a drug is both a substrate (transported out) and an inhibitor (blocks transport), the inhibitory effect can mitigate efflux, resulting in improved performance. However, the administration of these drugs in combination therapy may lead to undesirable drug-drug interactions.⁷³ Several FDA-approved antiviral drugs, including Ritonavir, Ledipasvir, and Simeprevir, act as both P-gp substrates and inhibitors. The excellent/poor, excellent/poor, and excellent/poor ratings of nicotiflorin, cpd1, and cpd3 suggest that they are good candidates for P-gp inhibitor-substrate drugs.

3.8.2.2. Distribution

The distributions of nicotiflorin, cpd1, cpd2, and cpd3 were predicted using the blood-brain barrier (BBB), plasma protein binding (PPB), and fraction unbound in plasma (FUP) pharmacokinetic properties. The BBB, a protective and highly selective membrane that prevents nearly all large-molecule medicines and more than 98% of small-molecule pharmaceuticals from entering the brain, is one of the main barriers to the distribution of drugs to brain cells.^74,75 PPB is a reversible interaction between drugs and blood proteins that regulates the amount of a “free” active drug in circulation. FUP refers to the portion of a drug unbound to plasma proteins that can be distributed into tissues, cross membranes, bind targets, and exert therapeutic effects. Highly bound drugs exhibit prolonged action; however, their distribution is limited.^76,77 As shown in Table 3, the predictions for the compounds were rated as “excellent,” “medium,” or “poor” to indicate whether the predicted property is within the ideal range (acceptable), average (acceptable), or outside the ideal range (unacceptable), respectively, of a candidate drug. Nicotiflorin, cpd1, cpd2, and cpd3 received excellent ratings for BBB, PPB, and FUP, indicating that these compounds are candidates for efficient distribution.

3.8.2.3. Metabolism

Drug metabolism involves the conversion of drugs into more polar, water-soluble, and easily excreted metabolites. This process primarily occurs in the liver and is catalyzed by an array of cytochrome P450 (CYP) enzymes.⁷⁸ Drug metabolism is a desirable process required for detoxifying the body, but it may also render drugs toxic and/or ineffective.⁷⁹ Inhibition of CYP enzymes by a drug is generally considered undesirable or risky in clinical practice, as it slows the metabolism of other drugs, leading to dangerous accumulation, toxic effects, or reduced efficacy of prodrugs. While sometimes used intentionally to boost drug levels, it commonly causes adverse drug-drug interactions.⁸⁰ Furthermore, being a substrate of CYP enzymes is generally considered neutral to slightly advantageous in drug development, as it allows the body to metabolize and eliminate drugs. However, it carries significant risks of drug-drug interactions, toxicity, or inefficacy if metabolic rates change.⁸¹ Nicotiflorin, cpd1, cpd2, and cpd3 were assessed as inhibitors or substrates of CYP enzymes using seven enzymes, including CYP1A2, CYP2C19, CYP2C9, CYP2D6, CYP3A4, CYP2B6, and CYP2C8. The results are shown in Table 3, where inhibitor/substrate is indicated by “+,” whereas noninhibitor/nonsubstrate is shown as “-.” The majority of the enzymes were not inhibited by the nicotiflorin and its analogs, except that all four compounds were labeled as potential inhibitors of CYP2C8. Cpd2 is suggested as a CYP2B6 inhibitor, whereas cpd3 is flagged as an inhibitor of CYP1A2, CYP2D6, and CYP3A4. Cpd1 and cpd3 were recognized as potential substrates of CYP2C9 and CYP2D6, whereas cpd3 was predicted to be a possible substrate of CYP1A2.

3.8.2.4. Excretion

Regarding excretion, two properties, plasma clearance (CLplasma) and half-life (t_1/2), were used for the assessment. CLplasma determines the volume of blood plasma that is completely cleared of a drug per unit time. Thus, CLplasma represents the efficiency of the body in eliminating a drug via renal excretion and hepatic metabolism.⁸² The t_1/2 of a drug, on the other hand, is the time required for its concentration in the body to decrease by 50%.⁸³ CLplasma and t_1/2 are both crucial pharmacokinetic parameters used to determine drug exposure and dosage required to maintain steady-state concentration. As indicated in Table 3, nicotiflorin, cpd1, cpd2, and cpd3 were predicted to possess excellent CLplasma, indicating that this property is within the ideal range for a drug candidate. Nicotiflorin was predicted to have an intermediate short t_1/2 (4-8 hours), whereas cpd1, cpd2, and cpd3 were predicted to have short t_1/2 (1-4 hours).

3.8.2.5. Toxicity

The toxicities of nicotiflorin, cpd1, cpd2, and cpd3 were assessed based on 13 properties, including carcinogenicity, rat oral acute toxicity, FDA maximum daily dose (FDAMDD), human ether-à-go-go-related gene (hERG) blockers, hERG blockers (10 µm), respiratory toxicity, human hepatotoxicity, nephrotoxicity, neurotoxicity, ototoxicity, hematotoxicity, RPMI-8226 immunotoxicity, and A549 cytotoxicity.^53,54 The results of the predictions are shown in Table 3, where each property is rated as “excellent,” “medium,” or “poor,” signifying that the predicted property is within the ideal range (acceptable), average (acceptable), or outside the ideal range (unacceptable), respectively, for a drug candidate. Generally, all four compounds were flagged as excellent or medium for the assessed toxicity properties, except for nicotiflorin, which failed for ototoxicity, whereas cpd2 failed for nephrotoxicity and ototoxicity.

3.8.3. Mechanisms of DENV NS2B-NS3pro and inhibition by nicotiflorin and its analogs

All the amino acid residues that form the polyprotein binding pocket in DENV NS2B-NS3pro are crucial for hydrolyzing the polyprotein substrate into functional proteins. The non-catalytic residues of the binding site are essential for the recognition of the polyprotein substrate and for facilitating its stable binding via chemical interactions. Once the viral polyprotein binds to the active site of NS2B-NS3pro, the residues of the catalytic triad (Figure 1), including His51, Asp75, and Ser135, facilitate the hydrolysis of amide bonds to generate viral proteins. As illustrated in Figure 1, His51 is anchored in the correct orientation by Asp75 via hydrogen bonding to abstract a proton from Ser135, generating an electron-rich nucleophile. The nucleophile then mounts a nucleophilic attack on the carbonyl carbon of the scissile amide bond, initiating a chain of events that leads to the hydrolysis of the bond, release of products, and restoration of the active site structure for subsequent cycles. Owing to the importance of the catalytic triad in viral survival, the amino acid residues involved are highly conserved across species.^19,84 By binding to NS2B-NS3pro, nicotiflorin blocks the polyprotein from accessing the active site. Consequently, the proteins required for viral survival and propagation are not produced, leading to viral suppression. Because the catalytic triad is conserved, anti-NS2B-NS3pro agents that interact with its constituent residues are likely to exhibit effective and broad-spectrum activity against the four DENV serotypes. Unlike nicotiflorin, cpd1, cpd2, and cpd3 establish H-bonds with the conserved Asp75 and Ser135 of the catalytic triad of NS2B-NS3pro (Figure 5) and could be promising candidates for developing potent inhibitors against the four dengue serotypes.

4. Conclusions

In silico techniques were employed to identify nicotiflorin analogs with improved Ro5 and in silico-derived bioavailability score metrics, compared to those of nicotiflorin, as potential lead candidates for the development of oral inhibitors of DENV NS2B-NS3pro. Three nicotiflorin analogs, cpd1, cpd2, and cpd3, were identified as promising candidates. Unlike nicotiflorin, cpd1, cpd2, and cpd3 passed the Ro5 criterion, and each obtained an acceptable in silico-derived bioavailability score of 0.55, which is threefold better than the 0.17 score for nicotiflorin. The Ro5 and bioavailability results suggest that cpd1, cpd2, and cpd3 possess the necessary physicochemical properties for developing orally active drugs, unlike nicotiflorin. The three nicotiflorin analogs demonstrated RMSDs, H-bonding, and BFEs comparable to those of nicotiflorin, suggesting that their application as oral drug leads may not diminish the affinity for NS2B-NS3pro. Overall, the data suggest that cpd1, cpd2, and cpd3 are superior alternatives to nicotiflorin as lead compounds for designing and developing oral anti-NS2B-NS3pro drugs to combat dengue fever. In vitro, in vivo, and biochemical studies are recommended to further characterize cpd1, cpd2, cpd3, and nicotiflorin as potential anti-dengue inhibitors that target NS2B-NS3pro.

Supplemental material

Supplemental material - Unveiling nicotiflorin analogs with improved oral drug-like metrics as candidates for the development of dengue virus NS2B-NS3 protease inhibitors: An in silico study

Supplemental material for Unveiling nicotiflorin analogs with improved oral drug-like metrics as candidates for the development of dengue virus NS2B-NS3 protease inhibitors: An in silico study by Elvis Awuni in Journal of Chemical Research.

Footnotes

ORCID iD

Elvis Awuni

Consent to participate

This article does not contain any studies with human or animal participants.

Author contributions

The author asserts sole authorship.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

All data are incorporated within the text and supplemental material.*

Supplemental material

Supplemental material for this article is available online.

References

Tuiskunen Bäck

Lundkvist

. Dengue viruses – an overview. Infect Ecol Epidemiol 2013; 3: 19839. https://doi.org/10.3402/iee.v3i0.19839

Witte

Venturini

Meyer

, et al. Dengue fever—diagnosis, risk stratification, and treatment. Dtsch Arztebl Int 2024; 121: 773–778.

WHO . Dengue, 2025. https://www.who.int/news-room/fact-sheets/detail/dengue-and-severe-dengue, accessed 06/06/2026.

ECDC . 12-month dengue virus disease case notification rate per 100 000 population, 2025. https://www.ecdc.europa.eu/en/publications-data/12-month-dengue-virus-disease-case-notification-rate-100-000-population-september, accessed 21/03/2026.

Zhang

Zhou

Peng

, et al. Predictive Symptoms and Signs of Severe Dengue Disease for Patients with Dengue Fever: A Meta-Analysis. BioMed Res Int 2014; 2014: 359308. https://doi.org/10.1155/2014/359308

Htun

Xiong

Pang

. Clinical signs and symptoms associated with WHO severe dengue classification: a systematic review and meta-analysis. Emerg Microbes Infect 2021; 10: 1116–1128. https://doi.org/10.1080/22221751.2021.1935327

Pourzangiabadi

Najafi

Fallah

, et al. Dengue virus: Etiology, epidemiology, pathobiology, and developments in diagnosis and control – A comprehensive review. Infect, Genet Evol 2025; 127: 105710. https://doi.org/10.1016/j.meegid.2024.105710

Wiwanitkit

. Dengue fever: diagnosis and treatment. Expert Rev Anti-infect Ther 2010; 8: 841–845. https://doi.org/10.1586/eri.10.53

Rajapakse

Rodrigo

Rajapakse

. Treatment of dengue fever. Infect Drug Resist 2012; 5: 103–112. https://doi.org/10.2147/IDR.S22613

10.

Tayal

Kabra

Lodha

. Management of Dengue: An Updated Review. Indian J Pediatr 2023; 90: 168–177. https://doi.org/10.1007/s12098-022-04394-8

11.

Jiang

Wang

, et al. Antiviral agents for dengue virus. Virol Sin 2025; 40: 865–873. https://doi.org/10.1016/j.virs.2025.11.009

12.

Obi

Gutiérrez-Barbosa

Chua

, et al. Current Trends and Limitations in Dengue Antiviral Research. Trop Med Infect Dis 2021; 6: 180. https://doi.org/10.3390/tropicalmed6040180

13.

Zhang

. Chapter 7 - Flavivirus NS2B/NS3 Protease: Structure, Function, and Inhibition. In: Gupta

(ed). Viral Proteases and Their Inhibitors. Academic Press, 2017, pp. 163–188.

14.

Salaemae

Junaid

Angsuthanasombat

, et al. Structure-guided mutagenesis of active site residues in the dengue virus two-component protease NS2B-NS3. J Biomed Sci 2010; 17: 68. https://doi.org/10.1186/1423-0127-17-68

15.

Yusof

Clum

Wetzel

, et al. Purified NS2B/NS3 serine protease of dengue virus type 2 exhibits cofactor NS2B dependence for cleavage of substrates with dibasic amino acids in vitro. J Biol Chem 2000; 275: 9963–9969. https://doi.org/10.1074/jbc.275.14.9963

16.

Noble

Seh

Chao

, et al. Ligand-bound structures of the dengue virus protease reveal the active conformation. J Virol 2012; 86: 438–446. https://doi.org/10.1128/JVI.06225-11

17.

Krishna Murthy

Clum

Padmanabhan

. Dengue virus NS3 serine protease. Crystal structure and insights into interaction of the active site with substrates by molecular modeling and structural analysis of mutational effects. J Biol Chem 2009; 284: 34468.

18.

Lee

WHK

Liu

Fan

, et al. Dengue virus protease activity modulated by dynamics of protease cofactor. Biophys J 2021; 120: 2444–2453. https://doi.org/10.1016/j.bpj.2021.04.015

19.

Chandramouli

Joseph Jeremiah

Daudenarde

, et al. Serotype-Specific Structural Differences in the Protease-Cofactor Complexes of the Dengue Virus Family. J Virol 2010; 84: 3059–3067. https://doi.org/10.1128/JVI.02044-09

20.

Starvaggi

Previti

Zappalà

, et al. The Inhibition of NS2B/NS3 Protease: A New Therapeutic Opportunity to Treat Dengue and Zika Virus Infection. Int J Mol Sci 2024; 25: 4376. https://doi.org/10.3390/ijms25084376

21.

Norshidah

Leow

Ezleen

, et al. Assessing the potential of NS2B/NS3 protease inhibitors biomarker in curbing dengue virus infections: In silico vs. In vitro approach. Front Cell Infect Microbiol 2023; 13: 1061937. https://doi.org/10.3389/fcimb.2023.1061937

22.

Nitsche

. Proteases from dengue, West Nile and Zika viruses as drug targets. Biophys Rev 2019; 11: 157–165. https://doi.org/10.1007/s12551-019-00508-3

23.

Bock

Snitko

, et al. Novel dengue virus NS2B/NS3 protease inhibitors. Antimicrob Agents Chemother 2015; 59: 1100–1109. https://doi.org/10.1128/AAC.03543-14

24.

Purohit

Sahoo

Panda

, et al. Targeting the DENV NS2B-NS3 protease with active antiviral phytocompounds: structure-based virtual screening, molecular docking and molecular dynamics simulation studies. J Mol Model 2022; 28: 365. https://doi.org/10.1007/s00894-022-05355-w

25.

Dwivedi

Bharadwaj

Afroz

, et al. Anti-dengue infectivity evaluation of bioflavonoid from Azadirachta indica by dengue virus serine protease inhibition. J Biomol Struct Dyn 2021; 39: 1417–1430. https://doi.org/10.1080/07391102.2020.1734485

26.

Alqahtani

Kazi

Alsenaidy

, et al. Advances in Oral Drug Delivery. Front Pharmacol 2021; 12: 618411. https://doi.org/10.3389/fphar.2021.618411

27.

Martin

. A bioavailability score. J Med Chem 2005; 48: 3164–3170. https://doi.org/10.1021/jm0492002

28.

Bragina

Daina

Perez

MAS

, et al. The SwissSimilarity 2021 web tool: novel chemical libraries and additional methods for an enhanced ligand-based virtual screening experience. Int J Mol Sci 2022; 23: 811. https://doi.org/10.3390/ijms23020811

29.

Zoete

Daina

Bovigny

, et al. SwissSimilarity: a web tool for low to ultra high throughput ligand-based virtual screening. J Chem Inf Model 2016; 56: 1399–1404. https://doi.org/10.1021/acs.jcim.6b00174

30.

Erbel

Schiering

D'Arcy

, et al. Structural basis for the activation of flaviviral NS3 proteases from dengue and West Nile virus. Nat Struct Mol Biol 2006; 13: 372–373. https://doi.org/10.1038/nsmb1073

31.

Bos

Zambrana

Duarte

, et al. Serotype-specific epidemiological patterns of inapparent versus symptomatic primary dengue virus infections: a 17-year cohort study in Nicaragua. Lancet Infect Dis 2025; 25: 346–356. https://doi.org/10.1016/s1473-3099(24)00566-8

32.

Hernández Bautista

Cabrera Gaytán

Santacruz Tinoco

, et al. Retrospective analysis of severe dengue by dengue virus serotypes in a population with social security, Mexico 2023. Viruses 2024; 16: 769. https://doi.org/10.3390/v16050769

33.

Hossain

Rakib

MSI

Hasan

, et al. The 2023 Dengue Outbreak in Bangladesh: An Epidemiological Update. Health Sci Rep 2025; 8: e70852. https://doi.org/10.1002/hsr2.70852

34.

Agarwal

Ganvir

Kale

, et al. Continued dominance of dengue virus serotype 2 during the recent Central India outbreaks (2019-2021) with evidence of genetic divergence. Pathog Glob Health 2024; 118: 109–119. https://doi.org/10.1080/20477724.2023.2246712

35.

Diallo

Diouf

Gaye

, et al. Dengue vectors in Africa: A review. Heliyon 2022; 8: e09459. https://doi.org/10.1016/j.heliyon.2022.e09459

36.

Waterhouse

Bertoni

Bienert

, et al. SWISS-MODEL: Homology modelling of protein structures and complexes. Nucleic Acids Res 2018; 46: W296–W303. https://doi.org/10.1093/nar/gky427

37.

Guex

Peitsch

Schwede

. Automated comparative protein structure modeling with SWISS-MODEL and Swiss-PdbViewer: a historical perspective. Electrophoresis 2009; 30: S162–S173. https://doi.org/10.1002/elps.200900140

38.

Verdonk

Cole

Hartshorn

, et al. Improved protein-ligand docking using GOLD. Proteins: Struc, Funct, Bioinf 2003; 52: 609–623. https://doi.org/10.1002/prot.10465

39.

Jones

Willett

Glen

, et al. Development and validation of a genetic algorithm for flexible docking. J Mol Biol 1997; 267: 727–748. https://doi.org/10.1006/jmbi.1996.0897

40.

Jones

Willett

Glen

. Molecular recognition of receptor-sites using a genetic algorithm with a description of desolvation. J Mol Biol 1995; 245: 43–53. https://doi.org/10.1016/s0022-2836(95)80037-9

41.

Ramírez

Caballero

. Is It Reliable to Take the Molecular Docking Top Scoring Position as the Best Solution without Considering Available Structural Data? Molecules 2018; 23: 1038. https://doi.org/10.3390/molecules23051038

42.

Lindahl

Hess van der Spoel . GROMACS 2020.4 Manual. Zenodo, 2024.

43.

Van Der Spoel

Lindahl

Hess

, et al. GROMACS: Fast, flexible, and free. J Comput Chem 2005; 26: 1701–1718. https://doi.org/10.1002/jcc.20291

44.

Cornell

Cieplak

Bayly

, et al. A second generation force field for the simulation of proteins, nucleic acids, and organic molecules. J Am Chem Soc 1995; 117: 5179–5197. https://doi.org/10.1021/ja00124a002

45.

Jorgensen

Chandrasekhar

Madura

, et al. Comparison of simple potential functions for simulating liquid water. J Chem Phys 1983; 79: 926–935. https://doi.org/10.1063/1.445869

46.

Case

Cheatham

Darden

, et al. The AMBER biomolecular simulation programs. J Comput Chem 2005; 26: 1668–1688. https://doi.org/10.1002/jcc.20290

47.

Pearlman

Case

Caldwell

, et al. AMBER, a package of computer programs for applying molecular mechanics, normal mode analysis, molecular dynamics and free energy calculations to simulate the structural and energetic properties of molecules. Comput Phys Commun 1995; 91: 1–41. https://doi.org/10.1016/0010-4655(95)00041-d

48.

Bayly

Cieplak

Cornell

, et al. A well-behaved electrostatic potential based method using charge restraints for deriving atomic charges: the RESP model. J Phys Chem 1993; 97: 10269–10280. https://doi.org/10.1021/j100142a004

49.

Darden

York

Pedersen

. Particle mesh Ewald - An N.log(N) method for Ewald sums in large systems. J Chem Phys 1993; 98: 10089–10092. https://doi.org/10.1063/1.464397

50.

Hess

Bekker

Berendsen

HJC

, et al. LINCS: A linear constraint solver for molecular simulations. J Comput Chem 1997; 18: 1463–1472. https://doi.org/10.1002/(sici)1096-987x(199709)18:12<1463::aid-jcc4>3.0.co;2-h

51.

Bussi

Donadio

Parrinello

. Canonical sampling through velocity rescaling. J Chem Phys 2007; 126: 1–7. https://doi.org/10.1063/1.2408420

52.

Kumari

Kumar

Lynn

. g_mmpbsa-A GROMACS tool for high-throughput MM-PBSA calculations. J Chem Inf Model 2014; 54: 1951–1962. https://doi.org/10.1021/ci500020m

53.

Shi

, et al. ADMETlab 3.0: an updated comprehensive online ADMET prediction platform enhanced with broader coverage, improved performance, API functionality and decision support. Nucleic Acids Res 2024; 52: W422–W431. https://doi.org/10.1093/nar/gkae236

54.

Dong

Wang

Yao

, et al. ADMETlab: a platform for systematic ADMET evaluation based on a comprehensively collected ADMET database. J Cheminform 2018; 10: 29. https://doi.org/10.1186/s13321-018-0283-x

55.

Kellstein

Fernandes

. Symptomatic treatment of dengue: should the NSAID contraindication be reconsidered? Postgrad Med 2019; 131: 109–116. https://doi.org/10.1080/00325481.2019.1561916

56.

Pan

Huang

Cho

, et al. Consideration of molecular weight during compound selection in virtual target-based database screening. J Chem Inf Comput Sci 2003; 43: 267–272. https://doi.org/10.1021/ci020055f

57.

Hopkins

Keserü

Leeson

, et al. The role of ligand efficiency metrics in drug discovery. Nat Rev Drug Discov 2014; 13: 105–121. https://doi.org/10.1038/nrd4163

58.

Patel

. Medicinal Importance, Pharmacological Activities and Analytical Aspects of a Flavonoid Glycoside 'Nicotiflorin' in the Medicine. Drug Metab Bioanal Lett 2022; 15: 2–11. https://doi.org/10.2174/1872312815666220404110200

59.

Awuni

. Effect of A22 on the Conformation of Bacterial Actin MreB. Int J Mol Sci 2019; 20: 1304. https://doi.org/10.3390/ijms20061304

60.

Daina

Michielin

Zoete

. SwissADME: A free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci Rep 2017; 7: 42717. https://doi.org/10.1038/srep42717

61.

Guo

Shen

. pKa, solubility, and lipophilicity. In: Yan

Caldwell

(eds). Optimization in Drug Discovery: In Vitro Methods. Humana Press, 2004, pp. 1–17.

62.

Wang

Skolnik

. Recent advances in physicochemical and ADMET profiling in drug discovery. Chem Biodivers 2009; 6: 1887–1899. https://doi.org/10.1002/cbdv.200900117

63.

Zhong

. ADMET properties: Overview and current topics. In: Grover

(ed). Drug Design: Principles and Applications. Springer Singapore, 2017, pp. 113–133.

64.

Delaney

. ESOL: Estimating aqueous solubility directly from molecular structure. J Chem Inf Comput Sci 2004; 44: 1000–1005. https://doi.org/10.1021/ci034243x

65.

Ali

Camilleri

Brown

, et al. Revisiting the general solubility equation: in silico prediction of aqueous solubility incorporating the effect of topographical polar surface area. J Chem Inf Model 2012; 52: 420–428. https://doi.org/10.1021/ci200387c

66.

Lipinski

. Lead- and drug-like compounds: the rule-of-five revolution. Drug Discov Today Technol 2004; 1: 337–341. https://doi.org/10.1016/j.ddtec.2004.11.007

67.

Ren

Lien

. Caco-2 cell permeability vs human gastrointestinal absorption: QSPR analysis. Prog Drug Res 2000; 54: 1–23. https://doi.org/10.1007/978-3-0348-8391-7_1

68.

Larregieu

Benet

. Drug discovery and regulatory considerations for improving in silico and in vitro predictions that use Caco-2 as a surrogate for human intestinal permeability measurements. Aaps j 2013; 15: 483–497. https://doi.org/10.1208/s12248-013-9456-8

69.

Wang

Bei

, et al. Recent advances in oral drug delivery systems for BCS III drugs. Curr Issues Mol Biol 2026; 48: 63. https://doi.org/10.3390/cimb48010063

70.

Ghadi

Dand

. BCS class IV drugs: Highly notorious candidates for formulation development. J Control Release 2017; 248: 71–95. https://doi.org/10.1016/j.jconrel.2017.01.014

71.

Hashmi

Sekar

Zahra

, et al. Advanced drug delivery strategies to overcome solubility and permeability challenges: driving biopharmaceutical advancements toward commercial success. ACS Omega 2025; 10: 40769–40792. https://doi.org/10.1021/acsomega.5c06171

72.

Lin

Yamazaki

. Role of P-glycoprotein in pharmacokinetics: clinical implications. Clin Pharmacokinet 2003; 42: 59–98. https://doi.org/10.2165/00003088-200342010-00003

73.

Lin

Wang

, et al. Drug transporters and metabolizing enzymes in antimicrobial drug pharmacokinetics: mechanisms, drug-drug interactions, and clinical implications. Biomolecules 2025; 15: 864. https://doi.org/10.3390/biom15060864

74.

Pardridge

. Drug transport across the blood-brain barrier. J Cereb Blood Flow Metab 2012; 32: 1959–1972. https://doi.org/10.1038/jcbfm.2012.126

75.

Chen

, et al. The blood-brain barrier: structure, regulation, and drug delivery. Signal Transduct Target Ther 2023; 8: 217. https://doi.org/10.1038/s41392-023-01481-w

76.

Webborn

PJH

Beaumont

Martin

, et al. Free drug concepts: a lingering problem in drug discovery. J Med Chem 2025; 68: 6850–6856. https://doi.org/10.1021/acs.jmedchem.5c00725

77.

Roberts

Pea

Lipman

. The clinical relevance of plasma protein binding changes. Clin Pharmacokinet 2013; 52: 1–8. https://doi.org/10.1007/s40262-012-0018-5

78.

Zhao

, et al. Cytochrome P450 enzymes and drug metabolism in humans. Int J Mol Sci 2021; 22: 12808. https://doi.org/10.3390/ijms222312808

79.

. Overview: Evaluation of metabolism-based drug toxicity in drug development. Chem Biol Interact 2009; 179: 1–3. https://doi.org/10.1016/j.cbi.2008.11.013

80.

Guengerich

. Inhibition of cytochrome P450 enzymes by drugs-molecular basis and practical applications. Biomol Ther (Seoul) 2022; 30: 1–18. https://doi.org/10.4062/biomolther.2021.102

81.

Ogu

Maxa

. Drug interactions due to cytochrome P450. Proc (Bayl Univ Med Cent) 2000; 13: 421–423. https://doi.org/10.1080/08998280.2000.11927719

82.

Toutain

Bousquet-Mélou

. Plasma clearance. J Vet Pharmacol Ther 2004; 27: 415–425. https://doi.org/10.1111/j.1365-2885.2004.00605.x

83.

Smith

Beaumont

Maurer

, et al. Relevance of Half-Life in Drug Design. J Med Chem 2018; 61: 4273–4282. https://doi.org/10.1021/acs.jmedchem.7b00969

84.

Niyomrattanakit

Winoyanuwattikun

Chanprapaph

, et al. Identification of residues in the dengue virus type 2 NS2B cofactor that are critical for NS3 protease activation. J Virol 2004; 78: 13708–13716. https://doi.org/10.1128/JVI.78.24.13708-13716.2004

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

1.48 MB