Identification of Potential Dipeptidyl Peptidase (DPP)-IV Inhibitors among Moringa oleifera Phytochemicals by Virtual Screening, Molecular Docking Analysis, ADME/T-Based Prediction, and In Vitro Analyses

Moringa oleifera Lam. (MO) is called the “Miracle Tree” because of its extensive pharmacological activity. In addition to being an important food, it has also been used for a long time in traditional medicine in Asia for the treatment of chronic diseases such as diabetes and obesity. In this study, by constructing a library of MO phytochemical structures and using Discovery Studio software, compounds were subjected to virtual screening and molecular docking experiments related to their inhibition of dipeptidyl peptidase (DPP-IV), an important target for the treatment of type 2 diabetes. After the four-step screening process, involving screening for drug-like compounds, predicting the absorption, distribution, metabolism, excretion, and toxicity (ADME/T) of pharmacokinetic properties, LibDock heatmap matching analysis, and CDOCKER molecular docking analysis, three MO components that were candidate DPP-IV inhibitors were identified and their docking modes were analyzed. In vitro activity verification showed that all three MO components had certain DPP-IV inhibitory activities, of which O-Ethyl-4-[(α-l-rhamnosyloxy)-benzyl] carbamate (compound 1) had the highest activity (half-maximal inhibitory concentration [IC50] = 798 nM). This study provides a reference for exploring the molecular mechanisms underlying the anti-diabetic activity of MO. The obtained DPP-IV inhibitors could be used for structural optimization and in-depth in vivo evaluation.


Introduction
Diabetes mellitus (DM) is a metabolic disorder characterized by long-term high blood glucose levels [1]. Over time, DM can severely damage the heart, kidneys, and nervous system. DM is divided into two categories (types 1 and 2) based on an absolute or relative lack of insulin among patients. Patients with type 2 DM account for about 90% of patients with DM. Recent data from the World Health Organization (WHO) indicate that DM affected >422 million people worldwide (8.5% of the global population) in 2014, which may increase to 592 million by 2035. In 2016, DM ranked seventh regulating mitochondrial respiration [20]. Most studies on the hypoglycemic activity of MO extracts have not identified an individual component with a specific molecular mechanism.
In this study, a virtual library of MO phytochemicals was established, and potential DPP-IV inhibitors were discovered by virtual screening based on drug-like properties and molecular docking evaluation principles, and the inhibition of DPP-IV was confirmed by in vitro experiments. Three potential DPP-IV inhibitors of MO origin were discovered for the first time, and the study revealed the possible anti-diabetic molecular mechanism. The three DPP-IV inhibitors could be used as the basis for further structural optimization and in vivo research.

Results and Discussion
A virtual library of 111 compounds that isolated from MO was established using a database search (Table S1 in Supplementary Materials). First, based on Lipinski's "rule of five", molecules with less reasonable physicochemical properties were discarded [21], leading to the selection of 64 candidate molecules with good drug-like properties: molecular weight < 500, number of hydrogen bond donors < 5, number of hydrogen bond acceptors < 10, ALogP < 5, and no more than one violation of the above criteria.
Next, the "ADME/T descriptors" (absorption, distribution, metabolism, excretion, and toxicity) and "toxicity prediction" modules were used to predict the pharmacokinetic and toxicity parameters of the 64 candidate molecules. We excluded molecules that are difficult for the intestine to absorb, easily penetrate the BBB, inhibit CYP2D6, have a high plasma protein binding rate, have poor water solubility, and are toxic (high probability of carcinogenicity and mutagenesis), leaving 23 candidate compounds [22]. The relationship between the two-dimensional polar surface area (PSA_2D) and the calculated value of AlogP98 for the 23 compounds is shown in Figure 1, with the HIA and BBB penetration model 95% and 99% confidence ellipses. Predicting the value of AlogP98 can determine the hydrophilicity of the compound. AlogP98 > 5 may be related to the absorption or permeability of the compound. PSA is another key attribute related to drug bioavailability, as compounds with PSA <140 Å 2 can be passively absorbed and so have high oral bioavailability [23]. As shown in Figure 1, the 23 compounds all fell within these ranges. In this study, a virtual library of MO phytochemicals was established, and potential DPP-IV inhibitors were discovered by virtual screening based on drug-like properties and molecular docking evaluation principles, and the inhibition of DPP-IV was confirmed by in vitro experiments. Three potential DPP-IV inhibitors of MO origin were discovered for the first time, and the study revealed the possible anti-diabetic molecular mechanism. The three DPP-IV inhibitors could be used as the basis for further structural optimization and in vivo research.

Results and Discussion
A virtual library of 111 compounds that isolated from MO was established using a database search (Table S1 in Supplementary Materials). First, based on Lipinski's "rule of five", molecules with less reasonable physicochemical properties were discarded [21], leading to the selection of 64 candidate molecules with good drug-like properties: molecular weight < 500, number of hydrogen bond donors < 5, number of hydrogen bond acceptors < 10, ALogP < 5, and no more than one violation of the above criteria.
Next, the "ADME/T descriptors" (absorption, distribution, metabolism, excretion, and toxicity) and "toxicity prediction" modules were used to predict the pharmacokinetic and toxicity parameters of the 64 candidate molecules. We excluded molecules that are difficult for the intestine to absorb, easily penetrate the BBB, inhibit CYP2D6, have a high plasma protein binding rate, have poor water solubility, and are toxic (high probability of carcinogenicity and mutagenesis), leaving 23 candidate compounds [22]. The relationship between the two-dimensional polar surface area (PSA_2D) and the calculated value of AlogP98 for the 23 compounds is shown in Figure 1, with the HIA and BBB penetration model 95% and 99% confidence ellipses. Predicting the value of AlogP98 can determine the hydrophilicity of the compound. AlogP98 > 5 may be related to the absorption or permeability of the compound. PSA is another key attribute related to drug bioavailability, as compounds with PSA <140 Å 2 can be passively absorbed and so have high oral bioavailability [23]. As shown in Figure 1, the 23 compounds all fell within these ranges. Figure 1. Relationship between the two-dimensional polar surface area (PSA_2D) and the calculated value of AlogP98 of 23 candidate compounds selected after absorption, distribution, metabolism, excretion, and toxicity (ADME/T) screening, showing the corresponding blood-brain barrier (BBB) penetration and human intestinal absorption (HIA) model 95% and 99% confidence ellipses.

Figure 1.
Relationship between the two-dimensional polar surface area (PSA_2D) and the calculated value of AlogP98 of 23 candidate compounds selected after absorption, distribution, metabolism, excretion, and toxicity (ADME/T) screening, showing the corresponding blood-brain barrier (BBB) penetration and human intestinal absorption (HIA) model 95% and 99% confidence ellipses. All 23 compounds were located in the HIA 99% confidence ellipse, and the absorption grade (Table 1) indicated that all the compounds had good absorption except one, which had moderate absorption. BBB grade predictions indicated that all compounds had medium or very low BBB permeability [24]. Regarding the CYP2D6 inhibition predictions, no compounds inhibited this enzyme and none cause serious drug interaction toxicity. As drug activity is related to free drug concentration, it is necessary to consider whether each compound may bind to plasma proteins [25]. The 23 candidate compounds all had weak plasma protein binding activity, with binding rates <90%. Regarding solubility predictions, 17 compounds had extremely high solubility, five had good solubility, and only one had low solubility. Next, molecular docking virtual screening (including LibDock and CDOCKER analyses, based on the structural matching degree analysis involving the target DPP-IV protein) were conducted for the 23 candidate compounds. The LibDock program involves a heatmap matching simulation based on the structures of immobilized molecules and the receptor protein [26]. LibDock calculates a heatmap for the active site of the receptor protein, which contains polar and nonpolar interaction sites, and then the ligands with various conformations are rigidly superimposed onto the map to determine the most suitable interaction and energy optimization [27]. For each compound, the conformation with the highest docking score can be obtained, and the compounds can then be listed by docking score. The binding site in the co-crystal structure (Protein Data Bank ID: 6B1E) of the drug vildagliptin (positive control) and the DPP-IV enzyme was selected to be the receptor binding site. The 23 candidate ligand compounds were subjected to the "prepare ligands" module to generate 352 configurations of ligands to match with the receptor. Seven out of the 23 molecules had higher LibDock scores than vildagliptin ( Table 2). The seven compounds then underwent docking screening using the CDOCKER program. CDOCKER is a semi-flexible molecular docking analysis method based on the CHARMm force field, which can produce high-precision docking results, and it provides information on the interaction binding energy and ligand-receptor docking mode [28]. According to the molecular docking results of CDOCKER (Table 3), CDOCKER interactions and binding energies for three of the seven compounds were better than vildagliptin [29]. Therefore, compounds 1-3 may be potential DPP-IV inhibitors based on the above virtual screening and docking processes. The molecular structures of the three compounds are shown in Figure 2.  The seven compounds then underwent docking screening using the CDOCKER program. CDOCKER is a semi-flexible molecular docking analysis method based on the CHARMm force field, which can produce high-precision docking results, and it provides information on the interaction binding energy and ligand-receptor docking mode [28]. According to the molecular docking results of CDOCKER (Table 3), CDOCKER interactions and binding energies for three of the seven compounds were better than vildagliptin [29]. Therefore, compounds 1-3 may be potential DPP-IV inhibitors based on the above virtual screening and docking processes. The molecular structures of the three compounds are shown in Figure 2. Understanding ligand-receptor interactions in depth provides a basis for the subsequent optimization of drug structures. We analyzed the docking modes of the three screened compounds based on the CDOCKER analysis. As a reference, vildagliptin and the three amino acid residues of the DPP-IV binding site each formed four hydrogen bonds, and vildagliptin's adamantane fragment and tetrahydropyrrolidine fragment also formed three hydrophobic interactions with DPP-IV. Regarding the docking mode of compounds 1-3, it was found that, like vildagliptin, each compound could form various hydrogen bonds and hydrophobic interactions with key amino acid residues at the DPP-IV binding site. The details are shown in Table 3. The compounds in the table are arranged according to the binding energy; the larger the value of the CDOCKER interaction energy and the lower the negative value of the binding energy (ΔG), the stronger the ligand-receptor interaction force [30]. Compound 1 is a unique urethane compound found in MO seeds. Each fragment of compound 1 can form various interaction bonds with DPP-IV ( Figure 3). The N atom and glycosyl side chain of the urethane moiety and the O atom of the chain form four hydrogen bonds with the Asn710, Glu205, and Glu206 residues of DPP-IV, respectively, and the benzene ring can form two π- Understanding ligand-receptor interactions in depth provides a basis for the subsequent optimization of drug structures. We analyzed the docking modes of the three screened compounds based on the CDOCKER analysis. As a reference, vildagliptin and the three amino acid residues of the DPP-IV binding site each formed four hydrogen bonds, and vildagliptin's adamantane fragment and tetrahydropyrrolidine fragment also formed three hydrophobic interactions with DPP-IV. Regarding the docking mode of compounds 1-3, it was found that, like vildagliptin, each compound could form various hydrogen bonds and hydrophobic interactions with key amino acid residues at the DPP-IV binding site. The details are shown in Table 3. The compounds in the table are arranged according to the binding energy; the larger the value of the CDOCKER interaction energy and the lower the negative value of the binding energy (∆G), the stronger the ligand-receptor interaction force [30]. Compound 1 is a unique urethane compound found in MO seeds. Each fragment of compound 1 can form various interaction bonds with DPP-IV ( Figure 3). The N atom and glycosyl side chain of the urethane moiety and the O atom of the chain form four hydrogen bonds with the Asn710, Glu205, and Glu206 residues of DPP-IV, respectively, and the benzene ring can form two π-π interactions with the Tyr662 and Tyr666 residues. The ethane fragment of the side chain can also interact with the Val656 and His704 residues to form two alkyl hydrophobic forces. Faiz et al. reported that compound 1 has a certain hypotensive effect [31], but there are no reports on hypoglycemic activity or DPP-IV inhibition for compound 1.
π interactions with the Tyr662 and Tyr666 residues. The ethane fragment of the side chain can also interact with the Val656 and His704 residues to form two alkyl hydrophobic forces. Faiz et al. reported that compound 1 has a certain hypotensive effect [31], but there are no reports on hypoglycemic activity or DPP-IV inhibition for compound 1. Compound 2 is a major isothiocyanate active ingredient found in MO seeds. It can form six hydrogen bonds with DPP-IV, three O atoms, and two H atoms of the glycosyl side chain, and can interact with six residues in the DPP-IV binding site, at the His 740, Arg125, Ser630, Asn710, Tyr662, and Tyr547positions. The benzene ring of compound 2 can form a π-π stacked interaction with the Phe357 residue. However, it can be seen from Figure 4 that the H30 atom of compound 2 forms an unfavorable hydrogen bond with the Ser630 residue, and the key isothiocyanate group did not form any interaction with DPP-IV. This may have resulted in the lower binding energy for compound 2 compared to compound 1. Carrie et al. reported that the addition of 5% compound 2-rich MO extract to mouse feed can inhibit the rate-limiting step in liver gluconeogenesis, thereby directly or indirectly increasing insulin signaling and sensitivity [32]. This study indicates the reliability of our screening method, but no molecular mechanism underlying the effect of compound 2 against DM has been reported. Our findings will be helpful for related research. Compound 3 is a dipeptide with a special structure that was first found in Aspergillus penicillioides. Isshiki et al. discovered that compound 3 relieves arthritis in rats by inhibiting Cathepsin L and B enzymes [33]. Yoon et al. found that compound 3 can inhibit the c-Jun N-terminal kinase (JNK) and p38 pathways to protect against nephritis induced by lipopolysaccharide (LPS) stimulation [34]. In our study, compound 3 and DPP-IV only formed one hydrogen bond, at the TYR (547 position (Figure 5)), while three benzene ring fragments formed four π-π interactions with the His704, Tyr547, and Phe357 residues. The number and types of interactions of compound 3 were Compound 2 is a major isothiocyanate active ingredient found in MO seeds. It can form six hydrogen bonds with DPP-IV, three O atoms, and two H atoms of the glycosyl side chain, and can interact with six residues in the DPP-IV binding site, at the His 740, Arg125, Ser630, Asn710, Tyr662, and Tyr547positions. The benzene ring of compound 2 can form a π-π stacked interaction with the Phe357 residue. However, it can be seen from Figure 4 that the H30 atom of compound 2 forms an unfavorable hydrogen bond with the Ser630 residue, and the key isothiocyanate group did not form any interaction with DPP-IV. This may have resulted in the lower binding energy for compound 2 compared to compound 1. Carrie et al. reported that the addition of 5% compound 2-rich MO extract to mouse feed can inhibit the rate-limiting step in liver gluconeogenesis, thereby directly or indirectly increasing insulin signaling and sensitivity [32]. This study indicates the reliability of our screening method, but no molecular mechanism underlying the effect of compound 2 against DM has been reported. Our findings will be helpful for related research.
Molecules 2020, 25, x FOR PEER REVIEW 6 of 12 π interactions with the Tyr662 and Tyr666 residues. The ethane fragment of the side chain can also interact with the Val656 and His704 residues to form two alkyl hydrophobic forces. Faiz et al. reported that compound 1 has a certain hypotensive effect [31], but there are no reports on hypoglycemic activity or DPP-IV inhibition for compound 1. Compound 2 is a major isothiocyanate active ingredient found in MO seeds. It can form six hydrogen bonds with DPP-IV, three O atoms, and two H atoms of the glycosyl side chain, and can interact with six residues in the DPP-IV binding site, at the His 740, Arg125, Ser630, Asn710, Tyr662, and Tyr547positions. The benzene ring of compound 2 can form a π-π stacked interaction with the Phe357 residue. However, it can be seen from Figure 4 that the H30 atom of compound 2 forms an unfavorable hydrogen bond with the Ser630 residue, and the key isothiocyanate group did not form any interaction with DPP-IV. This may have resulted in the lower binding energy for compound 2 compared to compound 1. Carrie et al. reported that the addition of 5% compound 2-rich MO extract to mouse feed can inhibit the rate-limiting step in liver gluconeogenesis, thereby directly or indirectly increasing insulin signaling and sensitivity [32]. This study indicates the reliability of our screening method, but no molecular mechanism underlying the effect of compound 2 against DM has been reported. Our findings will be helpful for related research. Compound 3 is a dipeptide with a special structure that was first found in Aspergillus penicillioides. Isshiki et al. discovered that compound 3 relieves arthritis in rats by inhibiting Cathepsin L and B enzymes [33]. Yoon et al. found that compound 3 can inhibit the c-Jun N-terminal kinase (JNK) and p38 pathways to protect against nephritis induced by lipopolysaccharide (LPS) stimulation [34]. In our study, compound 3 and DPP-IV only formed one hydrogen bond, at the TYR (547 position (Figure 5)), while three benzene ring fragments formed four π-π interactions with the His704, Tyr547, and Phe357 residues. The number and types of interactions of compound 3 were Compound 3 is a dipeptide with a special structure that was first found in Aspergillus penicillioides. Isshiki et al. discovered that compound 3 relieves arthritis in rats by inhibiting Cathepsin L and B enzymes [33]. Yoon et al. found that compound 3 can inhibit the c-Jun N-terminal kinase (JNK) and p38 pathways to protect against nephritis induced by lipopolysaccharide (LPS) stimulation [34]. In our study, compound 3 and DPP-IV only formed one hydrogen bond, at the TYR (547 position (Figure 5)), while three benzene ring fragments formed four π-π interactions with the His704, Tyr547, and Phe357 residues. The number and types of interactions of compound 3 were fewer than those of compounds 1 and 2. As a result, compound 3 had the lowest binding energy related to DPP-IV among the three compounds.
Molecules 2020, 25, x FOR PEER REVIEW 7 of 12 fewer than those of compounds 1 and 2. As a result, compound 3 had the lowest binding energy related to DPP-IV among the three compounds. After the four-step virtual screening of 111 MO phytochemicals, we identified three potential DPP-IV inhibitors and purchased these compounds. Thereafter, in vitro fluorescence detection of inhibitory activity against DPP-IV was performed, and the half-maximal inhibitory concentration (IC50) values of compounds 1-3 were calculated. As shown in Figure 6, the three MO compounds inhibited the activity of DPP-IV to a certain extent, and the inhibitory activity was consistent with the order of the CDOCKER results. The inhibitory activity of compound 1 was the strongest among the three compounds, with an IC50 of 798 nM, which is equivalent to that of the positive control, vildagliptin (IC50 = 528 nM). The inhibitory activity of compound 1 increased with the concentration, showing considerable concentration dependence. When the compound concentration reached 100 μM, the inhibition rates were 99.64%, 71.25%, 30.93%, and 23.46% for vitagliptin and compounds 1, 2, and 3, respectively. The IC50 values of compounds 2 and 3 were 157.694 μM and 191.126 μM, respectively, and both exhibited concentration dependence, but the DPP-IV inhibitory activity was moderate so no higher-concentration activity assay was performed. In this study, for the first time, MO compound 1, was found to be an excellent new type of DPP-IV inhibitor, making it a potential lead compound for the treatment of type 2 DM. The findings also suggest that the natural products' binding ability and selectivity toward the protein target still need to be improved compared to those of commercially available drugs. In the future, a series of After the four-step virtual screening of 111 MO phytochemicals, we identified three potential DPP-IV inhibitors and purchased these compounds. Thereafter, in vitro fluorescence detection of inhibitory activity against DPP-IV was performed, and the half-maximal inhibitory concentration (IC 50 ) values of compounds 1-3 were calculated. As shown in Figure 6, the three MO compounds inhibited the activity of DPP-IV to a certain extent, and the inhibitory activity was consistent with the order of the CDOCKER results. The inhibitory activity of compound 1 was the strongest among the three compounds, with an IC 50 of 798 nM, which is equivalent to that of the positive control, vildagliptin (IC 50 = 528 nM). The inhibitory activity of compound 1 increased with the concentration, showing considerable concentration dependence. When the compound concentration reached 100 µM, the inhibition rates were 99.64%, 71.25%, 30.93%, and 23.46% for vitagliptin and compounds 1, 2, and 3, respectively. The IC 50 values of compounds 2 and 3 were 157.694 µM and 191.126 µM, respectively, and both exhibited concentration dependence, but the DPP-IV inhibitory activity was moderate so no higher-concentration activity assay was performed.
Molecules 2020, 25, x FOR PEER REVIEW 7 of 12 fewer than those of compounds 1 and 2. As a result, compound 3 had the lowest binding energy related to DPP-IV among the three compounds. After the four-step virtual screening of 111 MO phytochemicals, we identified three potential DPP-IV inhibitors and purchased these compounds. Thereafter, in vitro fluorescence detection of inhibitory activity against DPP-IV was performed, and the half-maximal inhibitory concentration (IC50) values of compounds 1-3 were calculated. As shown in Figure 6, the three MO compounds inhibited the activity of DPP-IV to a certain extent, and the inhibitory activity was consistent with the order of the CDOCKER results. The inhibitory activity of compound 1 was the strongest among the three compounds, with an IC50 of 798 nM, which is equivalent to that of the positive control, vildagliptin (IC50 = 528 nM). The inhibitory activity of compound 1 increased with the concentration, showing considerable concentration dependence. When the compound concentration reached 100 μM, the inhibition rates were 99.64%, 71.25%, 30.93%, and 23.46% for vitagliptin and compounds 1, 2, and 3, respectively. The IC50 values of compounds 2 and 3 were 157.694 μM and 191.126 μM, respectively, and both exhibited concentration dependence, but the DPP-IV inhibitory activity was moderate so no higher-concentration activity assay was performed. In this study, for the first time, MO compound 1, was found to be an excellent new type of DPP-IV inhibitor, making it a potential lead compound for the treatment of type 2 DM. The findings also suggest that the natural products' binding ability and selectivity toward the protein target still need to be improved compared to those of commercially available drugs. In the future, a series of In this study, for the first time, MO compound 1, was found to be an excellent new type of DPP-IV inhibitor, making it a potential lead compound for the treatment of type 2 DM. The findings also suggest that the natural products' binding ability and selectivity toward the protein target still need to be improved compared to those of commercially available drugs. In the future, a series of derivatives could be rationally designed and synthesized according to the ligand-receptor interaction mode results in order to improve the affinity of the compounds to the target.
Our research also led to detailed in silico predictions of the toxicological properties of compounds 1-3. Table 4 shows the toxicological and chemical properties of the compounds. The numbers of hydrogen bond acceptors, hydrogen bond donors, ionization states, stereoisomers, and tautomers conformed to Lipinski's "rule of five", and the compounds exhibited good drug-like properties. Additionally, as shown in Table 1, compounds 1-3 had good ADME pharmacokinetic properties, and the prediction confidence was >95%. As shown in Table 4, compounds 1 and 3 (but not 2) were predicted to be biodegradable. All compounds were predicted to be noncarcinogenic for male and female rats and mice according to the US National Toxicology Program (NTP) classification regarding evidence of carcinogenic activity. Similarly, all compounds had high safety regarding predicted carcinogenicity according to the Ames mutagenicity and weight of evidence (WOE) results [35]. Compounds 1 and 2 had predicted hepatotoxicity, so improvements need to be made through further structural optimization. The skin sensitization evaluation indicated that all compounds are mild and nonirritating to the skin. Regarding the prediction of the median toxic dose (TD 50 ) and median lethal concentration (LC 50 ), compound 1 had a higher safe dose than vildagliptin, indicating that compound 1 is a good potential DPP-IV inhibitor. Table 4. Chemical information and toxicity properties of compounds 1-3 and vildagliptin.

Compound 1 2 3 Vildagliptin
than one violation of the following criteria: <5 hydrogen bond donors, <10 hydrogen bond acceptors, molecular weight < 500, AlogP < 5, and no more than one violation of the above criteria. After this, the "prepare ligands" module was applied to the remaining molecules to generate multiple conformations.

Second-Round Screening on the Basis of ADME/T Properties
The "ADME/T Descriptor" module of DS was used. In the parameter settings, water solubility, BBB penetration, CYP2D6 binding, liver toxicity, intestinal absorption, and plasma protein binding were selected as the research objects. In addition, more detailed toxicity predictions were performed for compounds 1-3. In the parameter settings of the "toxicity prediction" module, carcinogenicity, mutagenicity, skin irritation, TD 50 , and LC 50 were selected as the research objects. Compounds that had poor pharmacokinetic properties and were likely to have high carcinogenic and mutagenic potential were excluded.

Third-Round Screening Using LibDock
Before LibDock screening, it was necessary to determine the binding site. Thus, the vildagliptin-binding site in the X-ray crystal structure of DPP-IV in complex with vildagliptin (Protein Data Bank ID: 6B1E) was determined by co-crystallization (X: 35.8402, Y: 50.2541, and Z: 35.3156), and the radius was set to 12 Å. The ligand conformation generation method was selected as "best", the number of binding site hotspots was set to 100, and other parameters were set at their default values. The docking results were sorted by LibDock score.

Fourth-Round Screening Using CDOCKER Molecular Docking Analysis and Docking Mode Analysis
For the molecular docking analysis, the crystal structure of DPP-IV in complex with vildagliptin (Protein Data Bank ID: 6B1E) was selected as the acceptor, and it was optimized by hydrogenation and CHARMm force field calculations. The binding site was defined by the ligand atoms, and the radius range was automatically generated. The CHARMm force field and annealing simulation algorithm were used to optimize the energy of the complexes of ligands with the protein, combining them in different conformations. Parameters were set at their default values. After each compound was docked, the 10 best conformations were obtained. The compounds were screened by comprehensively considering their interaction energy and binding free energy. The analysis of the binding mode (3D or 2D ligand-receptor interaction simulation map) of each selected compound was also conducted using CDOCKER.

Conclusions
To identify the potentially anti-diabetic active components of MO, we carried out computer-assisted virtual screening of phytochemicals from MO based on the structure of the DPP-IV enzyme, and we verified the candidate compounds using an in vitro DPP-IV inhibition assay. For the first time, three natural MO components with inhibitory activity against DPP-IV were identified. Among them, the most effective compound was compound 1 (IC 50 = 798 nM), which is a urethane known as O-Ethyl-4-[(α-l-rhamnosyloxy) benzyl] carbamate. It has excellent pharmacokinetic properties and safety and is a potential lead compound against DPP-IV. Additionally, a molecular docking analysis was used to simulate the interaction mode of the candidate compounds with the DPP-IV receptor, which provides the necessary basis for subsequent structural optimization and drug research.