Integration of molecular docking, molecular dynamics and network pharmacology to explore the multi‐target pharmacology of fenugreek against diabetes

Abstract Fenugreek is an ancient herb that has been used for centuries to treat diabetes. However, how the fenugreek‐derived chemical compounds work in treating diabetes remains unclarified. Herein, we integrate molecular docking and network pharmacology to elucidate the active constituents and potential mechanisms of fenugreek against diabetes. First, 19 active compounds from fenugreek and 71 key diabetes‐related targets were identified through network pharmacology analysis. Then, molecular docking and simulations results suggest diosgenin, luteolin and quercetin against diabetes via regulation of the genes ESR1, CAV1, VEGFA, TP53, CAT, AKT1, IL6 and IL1. These compounds and genes may be key factors of fenugreek in treating diabetes. Cells results demonstrate that fenugreek has good biological safety and can effectively improve the glucose consumption of IR‐HepG2 cells. Pathway enrichment analysis revealed that the anti‐diabetic effect of fenugreek was regulated by the AGE‐RAGE and NF‐κB signalling pathways. It is mainly associated with anti‐oxidative stress, anti‐inflammatory response and β‐cell protection. Our study identified the active constituents and potential signalling pathways involved in the anti‐diabetic effect of fenugreek. These findings provide a theoretical basis for understanding the mechanism of the anti‐diabetic effect of fenugreek. Finally, this study may help for developing anti‐diabetic dietary supplements or drugs based on fenugreek.


| INTRODUC TI ON
Diabetes is the third biggest health threat after cardiovascular disease and cancer, according to the International Diabetes Federation. 1,2 Statistics from the World Health Organization indicate that 415 million people had diabetes in 2015, and the prevalence of diabetes is predicted to rise to 693 million by 2045. 2 Depending on the pathogenesis, diabetes can be broadly divided into Type 1 diabetes mellitus (T1DM), T2DM, gestational diabetes and other specific types; of these, T2DM accounts for more than 90% of cases. 3 T2DM is a chronic disease in which the negative feedback cycle between insulin activity and insulin secretion becomes dysregulated, resulting in abnormal glucose metabolism. 4 In addition, the long-term accumulated effects of abnormal glucose metabolism lead to cardiovascular injury, kidney disease, diabetic foot and other complications, which can have serious impacts on the mental and physical well-being of T2DM patients. 5 Lifelong medication is usually necessary for people with diabetes to control their blood glucose, and thus their health care costs are three times more than that of people without diabetes. In 2015, 67.3 billion USD was spent worldwide to treat diabetes, accounting for 12% of global health spending. 6,7 Therefore, there is an urgent need to develop effective ways to prevent and treat diabetes.
Fenugreek (Trigonella foenum-graecum L.) is a plant in the Luminaceae family and is native to India and North Africa. 8 It is cultivated in many parts of the world, including China, Australia, northern Africa, Europe and Argentina. 9 Fenugreek has been used as both a food and medicine in Iran, India and China for about six millennia. 10 Pharmacological studies have revealed that fenugreek contains alkaloids, flavonoids, polysaccharides, steroidal saponins and volatile oils, as well as other active constituents with lipid-lowering and hypoglycaemic effects. [11][12][13] Fenugreek is widely used as a natural dietary supplement for the treatment of diabetes, 14 for which it has been shown to be safe and effective. 15 Many studies have demonstrated that fenugreek-derived alkaloids not only show strong β-glucosidase inhibition, cancer suppression, antioxidant and anti-inflammatory activities but also exert effects such as cardiovascular protection, cholesterol reduction and adipogenesis inhibition. 16 Moreover, the flavonoids, polysaccharides and saponins derived from fenugreek have been shown to play a role in controlling blood glucose levels as well as treating diabetes and its complications in both patients and animal models. 17 However, the exact mechanism remains unclear.
Details on the reactions between specific active constituents in fenugreek and cells, genes and proteins in the body remain limited.
Therefore, an association analysis of the physicochemical properties of fenugreek along with potential targets and their possible mechanisms is needed. Recently, network pharmacology has emerged as an exciting new subfield of drug development because it combines analysis of biomedical big data with systems medicine. 18,19 By introducing biomedical big data, the network relationships between active compounds in traditional drugs and their target proteins can be constructed, thereby revealing the mechanism of synergistic therapeutic action in traditional medicines. Thus, network pharmacology has helped to transition the drug discovery field from the conventional 'one target, one drug' framework to a 'network target, multi-compound treatment' approach. 20,21 Fenugreek has long been used as a traditional Chinese herbal medicine for the treatment of diabetes. Recently, Oh et al. 22 investigated the anti-diabetic effect of garlic shell based on molecular docking and network pharmacology and reported the bioactive constituents and signalling pathways against diabetes, demonstrating the usefulness of these techniques in identifying the bioactive constituents of traditional herbal formulas, the potential targets of action and the mechanisms of interaction between them.
In the present study, we used molecular docking, molecular dynamics (MD) simulation and network pharmacology to investigate the active constituents of fenugreek and identify their potential target proteins, with the aim of elucidating the underlying mechanisms of its anti-diabetic activities. The study workflow is shown in

| Screening of active compounds and related targets
A list of compounds in fenugreek was obtained from the Traditional Chinese Medicine Systems Pharmacology (TCMSP) database (http:// lsp.nwu.edu.cn/tcmsp.php/). 23 Screening of active constituents was performed following the principle that drug-likeness (DL) is ≥0.18 and oral bioavailability is ≥30%. Compounds that were not identified by screening but that have been reported as metabolic regulators were also included in the network analysis. Then, the TCMSP database was searched with the aim of further predicting the targets of the active constituents. To generate a more comprehensive collection, the simplified molecular-input line-entry system (SMILES) entry for each active constituents was retrieved from the PubChem database. Next, the BindingDB (http://bindi ngdb. org/bind/index. jsp), DrugBank (https://go.drugb ank.com/), STITCH (http://stitch. embl.de/) and SwissTargetPrediction (http://www.swiss targe tpred iction.ch/) databases were searched for related targets of active compounds based on their SMILES formulas.

| Predicting targets of diabetes
Searches were conducted of GeneCards (https: //www.genec ards. org/) and OMIM (https:// www.omim.org/), using the keywords 'diabetes' and 'hyperlipidemia,' respectively, with the aim of obtaining the related targets of diseases in Homo sapiens. Of the targets obtained from GeneCards, those with a relevance score > 40 were selected as potential targets. As for the OMIM datasets, all of the targets related to diabetes were involved.

| Drug-Compound-Target-Disease network
The overlapping targets and active compounds were imported into Cytoscape ver. 3.7.1 in order to construct the Drug-Compound-Target-Disease network and the 'Network Analyzer' function was used to analyse the topological properties of the network.

| Protein-Protein Interaction (PPI) network
Protein-protein interactions were analysed using String ver. 11.0 (https://strin g-db.org/). To further investigate the mechanism of action of fenugreek in the treatment of T2DM, the intersection targets were imported into String. The minimum combined score was set to 0.400 and protein-interaction relationships for Homo sapiens were obtained. Targets without interaction were removed and the data were saved as SIF files. Next, the PPI network was constructed by importing node1, node2 and the combined score into Cytoscape.
Finally, the core targets were screened by cluster analysis using the MCODE plug-in in Cytoscape.

| Compound-Gene-Pathway network
To clarify the relationship among active compounds, target genes and pathways, the top 20 significant pathways were intersected with the Compound-Gene-Disease network to construct the Compound-Gene-Pathway network. The results were visualized in Cytoscape.

| Molecular docking and ADMET profiling
Molecular docking can help to validate key targets in network pharmacology. 24 First, we searched the important compounds in the PubChem database, and the structure of 'MOL2' was exported.
Next, the 3D structures of the core proteins were obtained from F I G U R E 1 Study workflow comprising a network pharmacology stage and a validation and prediction stage, aimed at elucidating the mechanisms underlying the anti-diabetic effects of fenugreek.
the Protein Data Bank (https://www.rcsb.org/). Then, the progress of dehydration, hydrogenation and charge editing were monitored using AutoDock software 4.2.6, and the modified target proteins and important compounds were exported as PDBQT files. Next, docking progress was monitored separately using AutoDock Vina software 1.2.0. Finally, the docking results were imported into PyMol software for analysis and visualization. Here, we employed Metformin, a commonly used treatment drug for diabetes mellitus, as reference drug.
The SwissADME online server was used to perform machine learning to predict the drug-like properties. The six important compounds were input into the SwissADME website and the ADMET properties were selected for analysis. The toxicity of the six important compounds was predicted using the free online tool ProTox-II, and the results, including predicted LD 50 , reverse mutation assay (Ames test) and cytotoxicity, were recorded. GROMACS (2021.3) with CHARMm36 force field were employed for MD simulations. According to docking results, the top three proteinligand complexes in each group were selected for MD simulations.

| Molecular dynamics simulations
Protein-inhibitor complex from PDB datebase were employed as the reference systems. Simulate parameters were taken as described. 25,26 A total of 10 ns MD simulation was conducted. The data for every one picosecond was saved. Xmgrace were utilized to analyse RMSD and RMSF of each complex. GMX-MM/PBSA (Version 1.6.0) was recruited to compute the binding free energy calculations. 27

| In vitro validation
HepG2 cells were used to establish insulin resistance model. As the method described by Chen et al. 28

| Intersection of fenugreek and diabetes targets
After exploring the TCMSP, SwissTargetPrediction, DrugBank and other databases. 277 diabetes-related targets without duplicate values were obtained. Meanwhile, a total of 462 diabetes targets were identified in the GeneCards and OMIM databases, using diabetesrelated target screening rules. As shown in Figure 2A, a total of 71 overlapping targets were obtained by determining the intersection of the 277 compound-related targets and the 462 diabetes-related targets by using Venny.

| Construction and analysis of the Drug-Compound-Target-Disease network
To construct the Drug-Compound-Target-Disease network ( Figure 2B), 71 common targets of fenugreek in diabetes treatment were input to Cytoscape. As shown in Figure

| PPI network of therapeutic targets for fenugreek against diabetes
A total of 71 common targets were input to the String database, and the PPI network was exported as a TSV file to Cytoscape, where the PPI network was visualized and analysed. As shown in Figure Table 3. and atherosclerosis, which suggests that fenugreek might exert its anti-diabetic effects through these pathways.

| Compound-Gene-Pathway network analysis
The Compound-Gene-Pathway network was constructed using Cytoscape. As shown in Figure 2D, the node size represents the degree value. The network comprises 110 nodes and 390 edges.
AKT1, PTGS2, MAPK1, TP53 and IL-6 interacted with most of the potential pathways, including cancer pathways, the AGE-RAGE signalling pathway, the FoxO signalling pathway, the HIF-1 signalling pathway, the NF-κB signalling pathway and lipids and atherosclerosis, which suggests that these genes might play a key role in the anti-diabetic effects of fenugreek. Figure 5 shows the action of the core target on the AGE-RAGE signalling pathway in diabetic complications.

TA B L E 2
Topology parameters of important compounds in the network.

| Molecular docking verification and ADMET profiling
The six important active constituentsβ-sitosterol, diosgenin, for- ADMET analysis is an important tool in drug discovery. The SwissADME database was used for ADMET analysis, and the results revealed that the important compounds have excellent pharmacokinetic properties. 29 As we can see in Table 6, all six selected drug candidates showed no side effects in terms of their pharmacokinetic properties in the various prediction models, including P-glycoprotein substrates, blood-brain barrier penetration, gastrointestinal absorption and human oral absorption. Furthermore, all the important active constituents showed good biocompatibility in the various toxicity prediction models, and none of them showed toxic behaviour such as reverse mutation, hepatotoxicity and cytotoxicity.

F I G U R E 3 Visualization of GO term enrichment.
Meanwhile, the binding energies of these complexes were calculated using the GMX-MM/PBSA method. From Table 7 indicates that these compounds have a strong binding affinity to the target receptor. Besides, the contribution of this binding affinity is mainly derived from van der Waals force and electrostatic interactions. Therefore, based on the collected results, diosgenin, quercetin and luteolin were further screened as the core components of fenugreek in anti-diabetes. These core components can bind with key targets to form stable complexes, which in turn exert antidiabetic effects. Figure 8A Figure 8B, by comparing to IR model, the glucose consumption was significantly increased in all of the fenugreek treated group. Thus, fenu-

| Effect of fenugreek on glucose consumption in IR HepG2 Cells
greek may contribute to improve the glucose uptake in IR-HepG2 cells.

| DISCUSS ION
Previous research has investigated the anti-diabetic effects of fenugreek. 11 In addition, extractions derived from fenugreek have been used for the treatment of diabetes, with good results. The therapeutic effects of fenugreek manifest mainly as improvements to insulin sensitivity and resistance as well as reduced fasting blood glucose and HbA1c levels. 33 These pharmacodynamic activities may be related to the flavonoids, alkaloids, volatile oils and unsaturated fatty acids derived from fenugreek. 14 However, which compositions are most effective, the potential target genes and the mechanisms of action are not fully understood, which limits the application of fenugreek.
In recent years, molecular docking technology and network pharmacology have been used to screen for active constituents and with the corresponding targets. These results suggest that the connection between the selected core active constituents and the core diabetes-related target genes is stable, indicating that the core active compounds in fenugreek can bind to the key genes involved in diabetes mellitus and induce anti-diabetic activity.
Futhermore, cell culture results demonstrated that fenugreek not only has good biological safety in vitro, but also can effectively improve the glucose consumption of insulin resistance HepG2 cells.
Thus, our results further suggest that these core active constitu- phosphorylation, which induces insulin resistance and leads to the development of T2DM. 38 During the development of diabetes, the nuclear factor is closely associated with inflammation and oxidative stress activation and plays a key role in the expression of cytokines such as TNFα, IL-6 and IL-1B as well as adhesion molecules such as ICAM1. 39 Zheng et al. 40

| CON CLUS ION
We used molecular docking, molecular dynamics simulation and network pharmacology to conduct a preliminary study of the ac-

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors have no conflicts of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available on request from authors.