The leaves of A. pavonina were first washed with distilled water to remove dust and then air-dried for 7 days in a well-ventilated area. Next, the dried leaves were ground into a fine powder. A 5 g powdered sample was extracted with 150 mL of 100% methanol solvent to produce the crude extract containing a wide range of bioactive compounds. The extract was filtered using Whatman No. 4 filter paper, and the solvent was completely evaporated using a rotary evaporator to produce a dense residue. The completely evaporated extract was then weighed and stored at 4°C.

The phytochemicals present in the methanolic leaf extract of A. pavonina were detected through GC-MS (Shimadzu, Japan; Model GC–MS QP 2020) analysis with an Electronic Ionization (EI) detector. A fused silica capillary column (SH-Rxi-5Sil MS 30m (L) X 0.25 mm (ID), X 0.25 um (DF)) was used at a temperature of 80.0°C. The samples were taken in a fixed split mode at 220.00°C. The oven was preheated at 80.0°C for 2 min, then 150.0°C for 3 min, and finally 280.0°C for 5 min. The ionization voltage was adjusted to 70 eV and the electron multiplier was set to 450 V. The unknown spectra of GC-MS were then compared to the spectra of existing chemicals housed in the NIST08s, NIST08 and NIST14 libraries. The names of the compounds and their concentrations were determined at different retention times (Kanthal et al., 2014). GC-MS Post Run software was used for data analysis.

Throughout this study, we bridge the gap between chemical composition and antidiabetic activity, starting with the identification of compounds through GC-MS and proceeding to the assessment of their potential interactions with 5hhw by molecular docking.

Phytochemicals screened through molecular docking were analyzed for absorption, distribution, metabolism, excretion and toxicity (ADMET) to determine whether they possessed the anticipated features to be used as lead molecules. To analyze ADMET profiles and determine molecules’ correspondence to Lipinski’s rule of five, the pkcsm (Pires et al., 2015) and SwissADME (Daina et al., 2017) web servers were used.

Molecular dynamics simulations were carried out using YASARA (Yet Another Scientific Artificial Reality Application) dynamics software and the AMBER (Assisted Model Building with Energy Refinement) 14 force field (He et al., 2020). Initial cleanup and optimization of the docked complexes were performed, as well as optimization of the hydrogen bond network. The topology files of ligands were auto-generated by General AMBER Force Field (GAFF) with AM1BCC charges. A cubic simulation cell was constructed using the TIP3P solvation model with a periodic boundary condition (Kadaoluwa Pathirannahalage et al., 2021). The distance between the cubic box and the edge of the protein was set to 12.5 Å, and the dimensions of the cubic box were 80 × 80 × 80 Å. The simulated system was neutralized at a pH of 7.4, 310 K, and 0.9% NaCl, representing physiological conditions (Kr et al., 2015). The particle mesh Ewald (PME) method was used to determine long-range electrostatic interactions with an 8.0 Å cutoff radius (Simmonett and Brooks, 2021). A time step of 2.0 fs was chosen for the simulation. The primary energy minimization was carried out using steepest gradient algorithms with simulated annealing techniques (5,000 cycles). The RMSD, Rg, SASA, hydrogen bonding and MMPBSA binding free energy were evaluated using the simulated trajectories (Baildya et al., 2021; Islam et al., 2021). The simulation was conducted for 100 ns, and the simulation trajectories were saved after every 100 ps (Patil et al., 2021).

Plant-derived bioactive compounds have the potential to lower the complications associated with diabetes (Tran et al., 2020), and IR is a promising target for designing drugs for the curative treatment of diabetes mellitus. In this study, we identified a total of 17 phytochemicals present in A. pavonina leaf extract through GC-MS analysis (Figure 2; Table 1). Compound 3,7-Cyclodecadiene-1-methanol, .alpha.,.alpha.,4,8-tetramethyl was found to be the most abundant (29.15%) in this extract, followed by 2-(4-Ethenyl-4-methyl-3-prop-1-en-2-ylcyclohexyl)propan-2-ol (18.29%) and 2-Naphthalenemethanol, decahydro-.alpha.,alpha.,4a-trimethyl-8-methylene-, [2R-(2.alpha.,4a.alpha.,8a.beta.)] (14.15%) (Table 1). On the other hand, sabinene (27.9%) was the major compound of A. pavonina leaf oil, followed by D-limonene (14.79%), as detected through GC-MS analysis (El-Shazly et al., 2020). Although there was a significant difference in the concentrations of the compounds, some of the compounds were present at very low concentrations (<1%) in this leaf extract (Table 1).

The presence of a mutation in the 5hhw domain of the IR prevents glucose uptake and metabolism, resulting in DMT2 (Cignarelli et al., 2019; Zhang et al., 2022). In our current study, we evaluated the A. pavonina leaf extract-derived bioactive compounds through in silico approach by targeting 5hhw protein to find out potential therapeutic drug candidates for DMT2. Based on their binding energy and number of interacting amino acid with the target protein 5hhw we have selected 3 compounds from 17 compounds for further in silico analysis (Supplementary Table S1). We used cis-(R)-7-(3-(azetidin-1-ylmethyl)cyclobutyl)-5-(3-((tetrahydro-2H-pyran-2-yl)methoxy)phenyl)-7H-pyrrolo[2,3-d]pyrimidin-4-amine (CID: 67035535) as a reference ligand because it was complexed with the crystal structure of the insulin receptor kinase domain. In the docking study, the binding energy for compound diazoprogesterone (CID: 104633) was −9.2 kcal/mol compared to the reference ligand CID: 67035535 (−10.5 kcal/mol). This was followed by the compounds 2,4,4,7a-Tetramethyl-1-(3-oxobutyl)octahydro-1H-indene-2-carboxylic acid (CID: 536510) and 2-Naphthalenemethanol, decahydro-.alpha.,.alpha.,4a-trimethyl-8-methylene-, [2R-(2.alpha.,4a.alpha., 8a.beta.)] (CID: 6432456), which exhibited binding energies of −6.9 kcal/mol and −6.6 kcal/mol, respectively (Table 2). The complex 5hhw + CID: 104633 exhibited two hydrogen bonds at A:THR1181 (2.88 Å) and A:GLY1179 (2.21 Å) (Figure 3A; Table 2). Similarly, the complex 5hhw + CID: 536510 was stabilized by three hydrogen bonds at A:GLU1186 (2.17 Å), A:TYR1189 (1.99 Å), and A:ASP1183 (2.27 Å) (Figure 3B; Table 2). However, the interaction between 5hhw and compound CID: 6432456 did not reveal any hydrogen bond but rather five alkyl bonds in the A chain of the 5hhw protein (Figure 3C; Table 2). In contrast, the interaction between 5hhw and reference ligand (CID: 67035535) exhibited three hydrogen bonds at A:LYS1057 (2.20 Å), A:GLU1104 (2.48 Å), and A:MET1106 (2.59 Å). The binding of these compounds to 5hhw may activate the IR by auto-phosphorylation and phosphorylation of intracellular substrates, thus initiating the insulin pathway and facilitating glucose uptake by the cell.

An in silico study has found that plant-derived polyphenols, such as flavonoids, can trigger the activation loop of IR, suggesting that they can be considered activators of IR (Lim et al., 2021). However, some in vitro and in vivo studies suggest that A. pavonina-derived phytochemicals possess antidiabetic activity (Wickramaratne et al., 2016; Vieira et al., 2018). The phytochemical galactomannans, isolated from A. pavonina, can reduce glycemia, total cholesterol, and triacylglycerol levels in streptozotocin-induced diabetic mice (Vieira et al., 2018). The methanolic leaf extract of A. pavonina significantly inhibits the alpha-amylase activity (IC₅₀ value 16.16 ± 2.23 µg/mL) (Wickramaratne et al., 2016). It has been found that the seed aqueous extract of A. pavonina can reduce the development of diabetic nephropathy in streptozotocin-induced diabetic mice (Pandhare et al., 2012).

The compounds screened through molecular docking were chosen for ADMET analysis to evaluate their biocompatibility, pharmacokinetics, and toxicity, and to identify the lead compound. According to Lipinski’s rule, a lead compound should have certain properties, such as molecular weight ≤500 da, hydrogen bond acceptor ≤10, hydrogen bond donor ≤5 and LogP value ≤ 5. At least two violations are considered according to this rule (Abbasi et al., 2019). In the present study, it was found that all compounds met Lipinski’s rule (No violation). The bioavailability score of a compound determines its physiological activity (Jia et al., 2020). All three compounds were found to be physiologically active (>0.00), and their bioavailability scores were the same (0.55) (Table 3). The water solubility profile showed that compound CID: 536510 (−2.543) was water soluble, while compounds CID: 104633 (−4.101) and CID: 6432456 (−4.9) were moderately soluble according to the solubility scale, which is insoluble < −10 < poorly < −6 < moderately < −4 < soluble < −2 < very <0 < highly (Wang et al., 2021). The blood-brain barrier (BBB) permeability of a drug is crucial for its ability to reach the brain (Daina and Zoete, 2016). The BBB and CNS permeability of compounds CID: 104633, CID: 536510, and CID: 6432456 were 0.084 and −1.843, 0.361 and −1.104, and 0.634 and −1.858, respectively. Furthermore, all screened compounds showed positive (No) results in the human ether-a-go-go (hERG) I inhibitor test, and no toxicity was found in terms of the AMES test, however, compound CID: 536510 was found as hepatotoxic (Table 3).

Molecular dynamics simulations were conducted to assess the stability of the hit complexes during the 100 ns simulation period. The RMSD value was utilized as the standard measure of structural distance between coordinates (Velázquez-Libera et al., 2020). The three best hit compounds were selected for molecular dynamics simulations to analyze their stability and identify potential inhibitors in a time-dependent manner for 100 ns Five parameters, including the root mean square deviation (RMSD), radius of gyration (Rg), solvent-accessible surface area (SASA), hydrogen bonds and MMPBSA binding free energy of the hit complexes were investigated.

In this study, all four complexes demonstrated satisfactory RMSD, although there was some fluctuation in the data at the very beginning of the simulation period (Figure 4A). The initial RMSD values of the four complexes were below 1 Å, but after a few ns, the RMSD values of all complexes increased. Complex C2 (5hhw + CID: 536510) exhibited more fluctuation than complexes C1 (5hhw + CID: 104633), C3 (5hhw + CID: 6432456), reference ligand-bound complex R (5hhw + CID: 67035535), and Apo protein after 20 ns. This might be due to the conformational variability of the C2 complex (Saha et al., 2023). After this period, both C1 and C2 maintained a stable RMSD value for the rest of the simulation period, whereas R and Apo maintained stability with C3 until 80 ns and then R showed a downward trend with a slight increase at the end. However, the C3 complex exhibited an increased RMSD value until 60 ns, and then finally followed the stable trend with the other complexes (Figure 4A).

During the 20 ns of the simulation period, the Rg values of C2, C3, R and Apo protein showed an increasing trend, while C1 exhibited a decreasing movement. From 20 ns to 75 ns, all complexes demonstrated a stable profile. Afterward, C1 and C2 showed an upward movement, while C3, R and Apo showed a downward trend. However, at the end of the simulation period, they converged at nearly the same point (Figure 4B). The Rg profiles of C3, R, and Apo maintained almost the same trend, while C1 and C2 exhibited slight fluctuations during the simulation period due to the folding and changes in the conformation of the complexes (Rodrigues et al., 2018). The SASA profiles of C1, C2, C3, R and Apo were satisfactory, and they exhibited similar trends and stable profiles throughout the entire simulation period (Figure 4C). Similarly, the hydrogen bonding in these complexes also followed the same trend as the SASA profile throughout the simulation time (Figure 4D).

Molecular Mechanics/Poisson-Boltzmann Surface Area (MMPBSA) was used to calculate the binding free energy of protein-ligand interactions (Lyndem et al., 2023). The MMPBSA binding free energy of all complexes during the last 20 ns simulation period is shown in Figure 5A. The average MMPBSA binding free energies of complexes C1, C2, C3, and R were −54.69 ± 2.40, −43.53 ± 0.28, −31.64 ± 0.88 and −22.40 ± 2.72 kJ/mol, respectively (Figure 5B). The negative MMPBSA binding free energy indicates the stability of the C1, C2, C3 and R complexes (Figure 5A).

Snapshots of molecular dynamics simulations were extracted at 0, 20, 40, 60, 80 and 100 ns to confirm the binding of ligands to the receptor. The findings indicate that the ligands remained bound to the receptor throughout the simulation period (Figure 6). Overall, the results of the molecular dynamics simulations suggest that CID: 104633, CID: 536510 and CID: 6432456 could serve as lead compounds.

In Figure 7, the interaction between the protein and ligand after a 100 ns simulation period is depicted (Saha and Nath Jha, 2023). The results revealed that the number of interactions between the complexes 5hhw + CID: 104633, 5hhw + CID: 536510, 5hhw + CID: 6432456 and 5hhw + CID: 67035535 was 4, 3, 4 and 9 respectively (Table 4), which closely resembled the docking interactions presented in Table 2.