1. INTRODUCTION
Diabetes is a global healthcare burden, affecting millions of people worldwide and leading to rising prevalence and associated complications such as cardiovascular diseases, kidney failure, and vision loss. It is broadly classified into two categories, Type 1 diabetes, caused by autoimmune destruction of pancreatic β-cells, and Type 2 diabetes, characterized primarily by insulin resistance with progressive β-cell dysfunction [1]. Approximately 589 million adult patients (20–79 years) are diagnosed with diabetes in 2024, and it is predicted to rise to 853 million in 2050 [2,3]. While India holds second rank for diabetes patients after China, as reported by IDF, India accounts for 1 in 7 of all adults living with diabetes worldwide [4].
Diabetes is a complex, chronic, and progressive metabolic disorder characterized by hyperglycemia, insulin resistance, ineffective glucose disposal, and varying degrees of pancreatic beta cell dysfunction. AMP-activated protein kinase (AMPK) is a primary indicator of cell’s bio-energetic status and is known to play a substantial role in the regulation of metabolic homeostasis [5]. The heterotrimeric AMPK is a typical serine/threonine kinase encompassing a catalytic α subunit and two regulatory β and γ subunits that can exist as multiple isoforms, in turn giving rise to 12 divergent complexes with varying tissue specificity [6]. Of note, the α2β2γ1 AMPK complex is primarily skeletal muscle-specific and known to facilitate insulin-independent glucose uptake. Hence, it could undeniably serve as a trustworthy target for the treatment of metabolic disorders, especially type-2 diabetes mellitus (T2DM) [1,7]. Chronic insulin resistance and β-cell malfunction are the key hallmarks of T2DM that contribute to abnormal glucose homeostasis [8]. Despite the potency and efficacy of the conventional anti-diabetic agents, they elicit certain untoward complications such as hypoglycemia, weight gain, cardiotoxicity, and β-cell apoptosis. Indeed, lack of target specificity is one of the key attributes of the latter effects. Necessitating a search for viable target-specific therapeutic strategies, a decrease in the intracellular ATP to a suboptimal concentration increases AMP or ADP, which eventually turns on AMPK [9]. Notably, the binding of AMP to the γ-subunit of AMPK is shown to promote threonine phosphorylation (Thr 172) indirectly (allosterically) or directly, triggering the upstream kinases like liver kinase B1 (LKB1) and calcium/calmodulin-dependent protein kinase 2, leading to its activation. In addition, AMP also causes a conformational change that further protects AMPK from inactivation through dephosphorylation by phosphatases [10]. Indeed, AMPK activation occurs by a variety of exogenous/endogenous stimuli, including starvation (CAMKK 2- 2-mediated), hypoxia (CAMKK 2- 2-mediated), detachment of cells from the matrix (CAMKK 2- 2-mediated), exercise, and hormones such as leptin and adiponectin [11].
Having evolved as an ingenious system in higher eukaryotes, AMPK critically monitors the cellular energy reserves, especially favouring multiple energy-consuming pathways such as fatty acid and cholesterol synthesis, fatty acid oxidation, mitochondrial biogenesis, and autophagy [12]. Therefore, developing agents that offer favourable safety, selectivity, and efficacy remains a highly anticipated pursuit. Interestingly, some natural and synthetic small molecules such as arctigenin, quercetin, berberine, trans-resveratrol, curcumin, metformin, canagliflozin, 5-aminoimidazole-4-carboxamide, rosiglitazone, 2,4-dinitrophenol, and salicylate are known to allosterically modulate AMPK either directly or indirectly, thereby eliciting beneficial effects in disease conditions associated with metabolic dysregulation [13,14]. Some of these are known to occupy a hydrophobic binding pocket termed as allosteric drug and metabolite (ADaM) site that is present between the α kinase domain and β carbohydrate binding module. Of note, tissue-specific AMPK isoforms raise the possibility of selective targeting, especially wherein α1β1γ1 are ubiquitously expressed, γ2 complexes are restricted to the heart, while γ3, in association with α2β2 complexes, are confined to glycolytic skeletal muscle. Essentially, α2β2γ1 and α1β2γ1 are skeletal muscle-specific complexes that steer insulin-independent glucose disposal [15,16]. Hence, targeting AMPK complexes bearing α2β2 isoforms could be an opportunistic strategy for treating metabolic disorders, principally type 2 diabetes mellitus. It is interesting to note that direct activators have a reduced tendency for lactic acidosis compared to the indirect activators. However, studies reveal that some small molecules like PT-1, A 769662 (a thienopyridone-based prototype direct β1AMPK activator), and MK-8722 (an imidazolopyridine- based pan AMPK activator) that directly activate AMPK manifest respiratory chain inhibition, 26S proteasome/Na+/K+-ATPase dysfunction, and cardiac hypertrophy as multiple off-target effects [11,17,18]. Despite the extensive knowledge on the structural and regulatory aspects of AMPK, the precise structural determinants of β2 activation with respect to the direct activators (especially, ADaM agonists) remain unsettled.
In fact, selective targeting of β2-AMPK complex by small molecules is an emerging and rather judicious strategy to improve glucose homeostasis in type-2 diabetes mellitus [19]. To this end, a systematic and comparative screening of a series of known and structurally diverse natural and synthetic AMPK activators, along with their respective chemical analogues, against the crystal structure of skeletal muscle-specific α2β2γ1 AMPK isoform (PDB ID: 6B2E) using a combination of valid computational tools. We hypothesise that this study could help find some credible small molecule-based in silico hits that may favorably interact with the skeletal-muscle-specific β2 AMPK complexes and open new avenues for developing targeted anti-diabetic therapies.
2. MATERIAL AND METHODS
2.1. Selection of the protein target and the ligands
The crystal structure of the full-length human AMPK enzyme complex (PDB ID: 6B2E), resolved by X-ray diffraction at 3.80 Å, was obtained from the protein data bank (PDB). The missing residues in the protein were rebuilt, and native ligands were removed to reduce steric hindrance at the active site and enhance docking accuracy. This heterotrimeric complex, composed of α2β2γ1 subunits, is skeletal muscle-specific and plays a key role in insulin-independent glucose uptake. In total, 13 putative AMPK activators (both direct and indirect) of natural and synthetic types were identified through rigorous literature mining, and their three-dimensional (3D) structures were retrieved from PubChem to create the standard dataset (Table 1). Additionally, 167 structural analogues of these activators were collected to form a query dataset. Analogues were manually filtered according to Lipinski’s rule of five to ensure drug-likeness, specifically retaining compounds with a molecular weight less than 500 g/mol. All the analogues were indicated using their PubChem CIDs. Additionally, analogues were chosen to represent a diverse chemical space around the lead compound to explore structure-activity relationships effectively, consistent with knowledge-based approaches in drug discovery that emphasise combining Lipinski’s rule with broader molecular property assessments [20].
Table 1. Various synthetic compounds retrieve from PubChem database for AMPK activators and further docking the top rank compound.
| Sr. No | Compound name and PubChem CID | Chemical structure | Chemical or pharmacological class | Number of chemical analogues retrieved |
|---|---|---|---|---|
| 1. | 5-aminoimidazole-4-carboxamide riboside (AICAR) (17513) |  | 1-ribosyl-imidazolecarboxamide | 10 |
| 2. | Arctigenin (64981) |  | Lignan | 27 |
| 3. | Berberine (2353) |  | Benzylisoquinoline alkaloid | 03 |
| 4. | Canagliflozin (24812758) |  | Sodium-glucose-co-transporter 2 (SGLT-2) Inhibitor | 10 |
| 5. | Curcumin (969516) |  | Diarylheptanone | 03 |
| 6. | Metformin (4091) |  | Biguanide | 01 |
| 7. | 2,4-Dinitrophenol (1493) |  | Nitrophenol | 09 |
| 8. | Quercetin (5280343) |  | Flavonol | 03 |
| 9. | trans-Resveratrol (445154) |  | Stilbene | 08 |
| 10. | Rosiglitazone (77999) |  | Thiazolidine-2,4-dione | 05 |
| 11 | Salicylic acid (338) |  | Hydroxyacid | 14 |
| 12 | A 769662 (54708532) |  | Thienopyridone | 21 |
| 13 | MK- 8722 (89558344) |  | Imidazolopyridine | 53 |
2.2. Screening the binding affinity of the molecules
Molecular docking was performed to virtually screen all compounds against the AMPK crystal structure (PDB ID: 6B2E) using AutoDock within the PyRx platform, employing the Lamarckian genetic algorithm as the scoring function. The protein was prepared by adding Gasteiger charges and polar hydrogens, while ligand SDF files were converted to PDBQT format using OpenBabel. The active site was identified via the CASTp server, and a grid box of 40 ? × 40 ? × 40 ? was centred at coordinates (−21.7824 ?, −45.7962 ?, and −2.526 ?). Moreover, to validate the docking methodology, the dimer structure is redocked with the native ligand CG7 and the 4 ligands using the same binding site constraints. The active site was identified (−21.7824 ?, −45.7962 ?, and −2.526 ?). Docking involved 10 parallel genetic algorithm runs with up to 27,000 generations per ligand, generating multiple conformations through mutation and crossover. The top 150 interactions were ranked, and eight docking solutions with the lowest binding energies were selected per compound. Post-docking analyses were conducted using UCSF Chimaera to visualise interactions and assess binding affinity, which was quantified in kcal/mol [21].
2.3. Molecular dynamics simulations
Protein-ligand molecular dynamics (MD) simulations were conducted to investigate the protein and ligand systems’ dynamic behaviour and interaction mechanisms. Molecular topology and force field parameters were generated using Amber Tools, with ligand topology derived via Antechamber through quantum chemical calculations. Parmchk2 converted these parameters into GROMACS-compatible formats, while T leap integrated ligand and protein topologies for simulation setup. Simulations employed GROMACS 2021.3 with the AMBER99 force field, using the crystal structure 6B2E and its ligand complexes. Systems were solvated in a TIP3P water model within a dodecahedron box, neutralised with 0.15 M NaCl, and energy minimised via the steepest descent algorithm to a 10.0 kJ/mol convergence or 50,000 steps. Production runs were performed under NPT conditions at 300 K for 100 ns with a two-fs timestep using the leap-frog integrator [22,23]. Trajectory analyses included radius of gyration (Rg), solvent accessible surface area, root mean square deviation (RMSD), and fluctuations (RMSF). Molecular interactions were visualised using PyMOL 2.5.7 and VMD [24].
2.4. Calculation of binding free energy by MM/GBSA method
The strength of protein-ligand interactions was estimated using the Molecular Mechanics/Generalized Born Surface Area (MM/GBSA) method implemented in MMPBSA. This approach calculated the binding free energy (ΔG_bind) as the difference between the free energy of the protein-ligand complex and the sum of the free energy of the isolated protein and ligand. Each component’s free energy was decomposed into gas-phase energy, solvation free energy, and entropy contributions. Due to computational constraints, only gas-phase and solution free energies were considered in this study. The binding free energy is calculated using the following equation [25,26]:
ΔG bind = ΔG Complex − (ΔG protein + ΔG ligand) (1)
Here, ΔG complex, ΔG protein, and ΔG Ligand represents the free energy of the complex, the isolated protein, and the isolated ligand, respectively. The ligand topology file, which defines the ligand’s atomic interactions, was generated using the antechamber tools within AMBER.
2.5. Density functional theory (DFT) calculations
The DFT theoretical computations were performed in the gas phase using the Gaussian 09 and the B3LYP 6-311G (d, p) basis set. This process involved simulating the geometrical optimisation for each compound to determine the lowest energy molecular structure. Subsequently, frequency calculations were performed at the optimised structural geometry, during which many thermochemical quantities were also determined. In the case of all the top four top-ranked molecules/ ligands (MK-8722, berberine, PubChem CID 138508665, and PubChem CID 138508964), optimum structures are regarded as stable because the imaginary frequency is absent. The energy optimisation of the ligands is followed energies of the highest occupied molecular orbitals (EHOMO), lowest unoccupied molecular orbitals (ELUMO) calculations, which are further used for the calculation of the energy gap (ΔG) through the following equation [27]:
Energy gap (ΔG) = (ELUMO-EHOMO)(2)
2.6. Theoretical prediction of pharmacokinetics and toxicity
Most of the time, the drug development process is highly susceptible to failure, which is attributed to four important factors: poor clinical efficacy, unreliable druglike properties, stringent toxicity, and lack of proper planning [28]. Hence, a fine-tuned combination of these factors dictates a molecule’s success rate. Absorption, distribution, metabolism, excretion, and toxicity, jointly termed as ADMET, are considered the critical indicators of pharmacokinetics and safety of a drug. Computational or in silico models are deemed valid for the ADMET prediction at the preliminary stages of the drug discovery program, in which several important ADMET parameters can be effortlessly and theoretically assessed using simple, feasible, and in silico tools. In general, a thoroughly validated ADMET model could enable prioritising the molecules during lead optimisation. Here, we employed a combination of Swiss-ADME and ADMET lab 2.0 tools to predict some of the essential pharmacokinetic and toxicological parameters of the analogues (query dataset) based on a scoring function [29].
2.7. Enrichment analysis of biological processes and genes
Total 50 genes were isolated form various literature support those who are directly and indirectly involve in T2DM including INS, IRS1, IRS2, PPARG, PIK3R1, AKT1, SLC2A4, FOXO1, TNF, IL6, LEP, ADIPOQ, GCK, TCF7L2, KCNJ11, ABCC8, MTOR, HNF1A, HNF4A, IGF1, IGF2, MAPK1, MAPK3, RELA, CD36, SOCS3, CXCL8(IL8), CAV1, PTPN1, FASN, SREBF1, G6PC, PDX1, NR1H3(LXRα), JAK2, TAT3, BCL2, SOD2, PRKAA1, GSK3B, TXNIP, UCP2, FABP4, ACACA, CPT1A, SLC30A8, NEUROD1, PDHA1, NFKB1, HSPD1. Furthermore, these genes were introduced into “ShinyGO version 0.76” software and integrated with the Gene Ontology (GO) and enrichment analysis [30–33].
3. RESULTS
3.1. Ligands against the selected target by molecular docking
All the standard molecules (total 13) and their respective chemical analogues (total 167) retrieved from the PubChem database were successfully docked and scored against the selected crystal structure of AMPK (PDB ID: 6B2E). These scores enable us to predict the strength of affinity of a given binding pose or orientation, wherein the highest negative value or score indicates a better association of ligand within the active site of the receptor protein in its most energetically favoured pose or orientation Table 1 and S1. It is observed that the docking/ binding affinity scores of the standard molecules fall within the range of −4.5 to −9.6 kcal/mol, and among them, MK-8722 and berberine orderly showed better binding affinity scores of −9.6 and −8.8 kcal/mol Table 2. On the other hand, most of the analogues displayed better binding affinity scores ranging between −4.3 to −10.3 kcal/mol compared to the standard molecules Table S1, of which, PubChem CID 138508665, PubChem CID 138508964, PubChem CID 132097535, and PubChem CID 138508669 are the leading ones with the binding affinity scores of −10.3, −10.2, −10.1, and −10 kcal/mol, respectively. Here, Figure 1A depicts the binding modes of the two top-ranked molecules from the standard and query datasets compared to the crystal ligand (CG7) bound at the ADaM site. To further affirm the biological credibility of this docking methodology, the protein structure is redocked with the native ligand CG7, and their poses are compared. Here in the redocking image (Figure 1A–D), against the native CG7 (red), the redocked CG7 (magenta) shows the docking score of −10Kcal/mol. Moreover, the redocked ligand binds proximal to the native ligand, and the distance deviation between these two positions is 4.43Å. This certainly shows the credibility of the docking methodology.” It is obvious from Figure 1B that the four molecules are bound at the regions more or less adjoining the ADaM site. On the other hand, we reiterate that the resolution of 3.80 Å, while moderate, is appropriate and supported by both experimental precedents and computational modelling advances for large multi-domain complexes. Our use of MODELLER and SWISS-MODEL for reconstructing missing residues follows best practices, rigorously combining homology-based spatial restraints and iterative refinement to achieve accurate, complete atomic models.
 | Figure 1. The protein-ligand interaction of top four ligands during molecular docking study: (A1), 3D interaction study of CID 138508665 with 6B2E; (B1), 3D interaction study of CID 138508964 with 6B2E; (A2), 2D interaction study of CID 138508665 with 6B2E; (B2), 3D interaction study of CID 138508964 with 6B2E; (C1), 3D interaction study of MK-8722 with 6B2E; (D) 3D interaction study with 6B2E; (C2), 2D interaction study of MK-8722 with 6B2E; (D2) 2D interaction study with 6B2E. Images were generated and presented using BIOVIA-Discovery Studio Visualizer 2021 and ChemDraw 2021 software, respectively. [Click here to view] |
Table 2. Binding affinity scores (in decreasing order) of the molecules of standard and query datasets.
| S. No | PubChem Compound CID | Binding Affinity Score (in Kcal/mol) |
|---|---|---|
| 1 | MK-8722 | −9.6 |
| 2 | Berberine | −8.8 |
| 3 | Arctigenin | −8.2 |
| 4 | Canagliflozin | −8.2 |
| 5 | A-769662 | −8 |
| 6 | Quercetin | −7.9 |
| 7 | Curcumin | −7.7 |
| 8 | trans- Resveratrol | −7.5 |
| 9 | Rosiglitazone | −7.3 |
| 10 | 2,4-dinitrophenol | −5.9 |
| 11 | 5-aminoimidazole-4-carboxamide riboside (AICAR) | −5.8 |
| 12 | Salicylic acid | −5.2 |
| 13 | Metformin | −4.5 |
However, they significantly vary in their distance with respect to the ADaM site, where Compound PubChem CID 138508665 and Compound PubChem CID 138508964 are literally proximal at 2.003 and 2.421 A0, respectively, overlapping with each other to a feasible extent. On the other hand, having a bound at 9.470 A0, MK 8722 is explicitly more distal to the ADaM site than berberine (which is bound at a distance of 7.728 A0) and the two top-ranked analogues. Above all, the binding poses of the two analogues also appear to be relatively divergent. Based on the above discussion, the hierarchy of proximity (from nearest to farthest) of the four top-ranked molecules with respect to the ADaM site can be denoted as: CID 138508665
 | Figure 2. 2D illustrative maps indicating the interactions of (A) MK-8722, (B) Berberine, (C) CID 138508665, and (D) CID 138508964 with the amino-acid residues at the binding site of AMPK (PDB ID: 6B2E). The ligands are presented in ball and stick model. [Click here to view] |
Table 3. (a)Molecular interaction profile of the four top-ranked molecules (6B2E_MK-8722, 6B2E_Berberine, 6B2E_138508665, and 6B2E_138508964) (b) Values indicating EHOMO, ELUMO, and Energy gap.
| (a) Molecular interaction profile of the four top-ranked molecules | ||||||||
| Complex Name | Binding Energy/ Affinity (Kcal/mol) | Total Number of H-bonds | Total Number of Salt Bridges | Total Number of Pi interactions | Ligand Contact Surface Area (in Ao2) | |||
| 6B2E_MK-8722 | −9.6 | 8 | 0 | 12 | 321.19 | |||
| 6B2E_Berberine | −8.8 | 5 | 0 | 7 | 208.93 | |||
| 6B2E_138508665 | −10.3 | 4 | 0 | 11 | 289.14 | |||
| 6B2E_138508964 | −10.2 | 3 | 0 | 11 | 283.77 | |||
| (b) Values indicating EHOMO, ELUMO, and Energy gap | ||||||||
| Molecules/ Ligands | EHOMO (eV) | ELUMO (eV) | Energy gap ΔG (eV) | |||||
| MK- 8722 | −4.41 | −0.01 | 4.4 | |||||
| Berberine | −3.32 | −1.62 | 1.7 | |||||
| CID 138508665 | −5.63 | −3.58 | 2.05 | |||||
| CID 138508964 | −2.90 | −2.39 | 0.51 | |||||
Perhaps, this is due to slight distortion of the two analogues with respect to MK−8722 so as to feasibly interact with the binding region, which is evident from Figure 1. It has been observed that all the four molecules and CG 7 established active contacts principally with the amino-acid residues of A and B chains of the protein target. To begin with, the CG 7 has four contacts with the residues GLY 19, ARG 82, and ASP 88, among which three are conventional hydrogen bonds and one is an electrostatic bond. Second, MK−8722 established eight contacts comprising conventional hydrogen bonds and carbon–hydrogen bonds with ARG 49, LEU 152, HIS 154, and GLU139 residues of A and B chains, unlike berberine, which has five contacts only with the A-chain residues, including LYS 45, GLY 98, GLU 94, PHE 27, and ASP 157, respectively, out of which two are conventional hydrogen bonds and three are carbon–hydrogen bonds. The detailed results are summarised in (Figure 2A, B, and Table S4). Furthermore, Compound PubChem CID 138508665 Figure 2C and Compound PubChem CID 138508964 Figure 2D interact particularly with the B chain residues that are common for both the analogues, except one residue (SER 93). Wherein, the former interacts with three residues, including ARG 82, SER 93, and ASP 136 via two conventional hydrogen bonds and two carbon–hydrogen bonds, while the latter interacts with only two residues, including ARG 82 and ASP 136 via two conventional hydrogen bonds and one carbon–hydrogen bond, respectively (Table S2). To sum up, the two analogues (Compound PubChem CID 138508665 and Compound PubChem CID 138508964) are considerably competent in terms of binding energy relative to the standard molecules.
3.2. Molecular dynamics simulations
When a respective ligand interacts with its protein target, it initiates conformational changes spanning all degrees of freedom and involving various torsion angles to achieve a stabilised overall conformation, if possible. These dynamic atomic motions play a pivotal role in the functional behavior of the system and can be investigated through MD simulations Figure 3. Here, we simulated the protein-ligand complexes and the native protein over 100-nanosecond (ns) timeframe with an aim to assess and ascertain the dynamic behavior of the ligands while accommodating at the protein’s binding pocket and the molecular interactions concerned with the stability of the complexes by critically studying Rg, RMSD, RMSF, hydrogen bonding, and interaction energies (kJ/mol). The Rg score quantifies the average deviation between the rotating axis and the centre of mass and indicates the compactness of the protein structure. Our findings concerning standard and analogue complexes compared with the native protein consistently show convergence in the scores, indicating robust stability.
 | Figure 3. MD analysis of wild type (WT) or native crystal structure of AMPK (PDB ID: 6B2E) and its MK-8722, berberine, CID 138508964, and CID 138508665 complexes indicating (A) radius of gyration, (B) RMSD, (C) RMSF, (D) number of hydrogen bonds, and (E) internal energy scores (in KJ/mol). [Click here to view] |
While the 6B2E-MK−8722 and 6B2E-Berberine complexes show average (Rg) scores of 3.065 ± 0.019 nm and 3.077 ± 0.022 nm, respectively, the corresponding scores for the analogue complexes 6B2E- Compound PubChem CID 138508665 and 6B2E- Compound PubChem CID 138508964 are orderly found to be 3.103 ± 0.027 nm and 3.109 ± 0.025 nm which is almost indistinct in relation to the two standard molecules. Contrarily, the score of the native protein is found to be 2.969 ± 0.031 nm and the simulation reveals that the native structure and its four complexes converge to a lower bound of 2.891, 2.977, 2.984, 2.980, and 2.983, indicating that the molecules compactly pack the 6B2E structure Figure 3A. Another useful metric for assessing overall conformational divergence and structural stability at a specific temperature is the RMSD which is an indicator of conformational stability of a protein based on the deviations of the atoms or groups and generally, smaller RMSD corresponds to a greater stability. The RMSD scores for the simulations of the two analogue complexes 6B2E- Compound PubChem CID 138508665 and 6B2E- Compound PubChem CID 138508964 are found to be 1.091 ± 0.170 and 0.843 ± 0.079 Å. However, the average RMSD scores for the standard complexes 6B2E-MK-8722 and 6B2E-Berberine are 0.78 ± 0.074 and 1.109 ± 0.151 Å, but 0.937 ± 0.109 Å for the native protein. This unequivocally demonstrates that the stability of the two analogue complexes is fairly comparable to the standard complexes, and the protein seems to get a bit stabilised when complexed with Compound PubChem CID 138508964 and MK-8722 Figure 3B.
Undulations that score the RMSF can also be used to accurately measure macromolecular stability in terms of average fluctuations of the atoms or the residues over a period of time, wherein a lower score denotes a higher level of stability. It is used to ascertain the degree of flexibility of a protein over a period (Lindahl et al., 2010). Unlike the RMSD scores, the average RMSF scores of the two analogue complexes are observed to be 0.370 ± 0.219 and 0.305 ± 0.222 Å, while the native protein, along with its well-studied standard MK-8722 and Berberine complexes, relatively show average scores of 0.285 ± 0.192, 0.282 ± 0.210, and 0.292 ± 0.229 Å, respectively, Figure 3C. Moreover, this score ranges between 0.210 and 0.746 for Compound PubChem CID 138508964 complex, whereas, for MK−8722 complex it ranges between 0.21 and 1.061. This clearly demonstrates that the overall conformation of 6B2E tends to stabilise a bit in the presence of MK−822 and Compound PubChem CID 138508964. Furthermore, for tracking the count of hydrogen bonds, which play a substantial role in stabilising conformation, the variations in the number of hydrogen bonds across the trajectory are plotted (Lindahl et al., 2010). The MK−8722, Berberine, and Compound PubChem CID 138508665 complexes show the maximum number of hydrogen bonds across the simulation timeframe in contrary to the 6B2E- Compound PubChem CID 138508964 complex Figure 3D. The plot shows that the molecules and 6B2E interacted more steadily along the trajectory, and the interactions are primarily driven by hydrogen and hydrophobic bonding, which in turn might have contributed to the plausible anchoring of the molecules within the active site. Using the Parrinello-Rahman parameter, the free interaction energies with respect to the four complexes are calculated in order to understand further the nonbonding interaction energies, wherein, the average interaction energies of MK- 8722, berberine, Compound PubChem CID 138508665, and Compound PubChem CID 138508964 are found to be −70.744 ± 35.561, −43.648 ± 15.754, −30.225 ± 25.792, and −18.106 ± 13.452 kJ/mol, respectively, Figure 3E.
To sum up, though the two analogues differ from MK−8722 and berberine in the aspects of distance, affinity, and binding orientation corresponding to the active site (Figure 1B), they exhibit favourable and stable interactions without significantly affecting the overall stability of the complexes. In addition, both the analogue complexes seem to compete with the standard complexes, especially in terms of Rg, RMSD, and RMSF Figure 3A, B, and C.
3.3. Calculation of binding free energy by MM/GBSA method
The MM/GBSA binding free energy calculation methodology was used as a post-scoring strategy to validate molecular docking analysis. The binding free energy of selected leads in a complex with 6B2E was calculated. The obtained MM/GBSA scores varied between−28.88 kcal/mol and−37.69 kcal/mol. These findings are significantly related to the docking scores. Based on the MM/GBSA analysis of the four complexes shown in Table S5, the two analogue complexes: 6B2E_Compound PubChem CID 138508665 and 6B2E_ Compound PubChem CID 138508964 appear to have the most favorable binding energy with ΔG_bind values of−34.17 and−37.69 kcal/mol, respectively. It suggests that these complexes are relatively more stable (than the standard complexes), as a lower ΔG_bind value indicates stronger and better binding affinity. These findings could certainly be beneficial for designing effective and especially selective small molecule-based activators.
3.4. Density functional theory (DFT) calculations
A smaller HOMO-LUMO gap suggests higher chemical reactivity and polarizability, which can facilitate stronger charge-transfer interactions with the target protein, potentially correlating with enhanced binding affinity and inhibitory potency. All hypothetical computations regarding stability were performed using Gaussian09. The four top-ranked molecules (MK−8722, berberine, Compound PubChem CID 138508665, and Compound PubChem CID 138508964) were fully optimised without imposing any symmetry constraints employing the Becke three-parameter exchange functional combined with the Lee-Yang-Parr (B3LYP) correlation functional at the 6–311 + G(d,p) level. The optimised structures of the four molecules demonstrating nonplanarity are illustrated in Figure S1. Electronic properties, including EHOMO, ELUMO and energy gap (ΔG), were investigated. The HOMO and LUMO plots of the four molecules are depicted in Figure 4, where the HOMO acts as an electron donor while the LUMO serves as an electron acceptor. The energies of the HOMO and LUMO are fundamental quantum chemical descriptors, providing insights into the reactivity, shape, and binding behaviour of molecules, as well as molecular substituents and fragments. The calculated HOMO-LUMO energy gaps for MK−8722, berberine, CID 138508665, and CID 138508964 are 4.4, 1.7, 2.05, and 0.51 eV, respectively, as presented in Table 3B. In general, molecules with lower/smaller energy gaps are considered more polarisable and often categorised as “soft” molecules.
 | Figure 4. HOMO, LUMO, and energy gap of the four top-ranked molecules. [Click here to view] |
3.5. Prediction of pharmacokinetics and toxicity
Basically, the aspect of drug likeness is about qualitatively assessing and comparing the pharmacokinetic properties of a molecule or a group of molecules with the approved and marketed drugs. In the present study, MK−8722 (top-ranked) from the standard dataset and the best 100 analogues from the query dataset were subjected to in silico pharmacokinetic prediction using the Swiss ADME tool. A broad array of parameters was considered in order to predict the pharmacokinetic behaviour of the molecules reliably. Totally fifteen crucial parameters including molecular weight, number of hydrogen bond donors and acceptors, topological polar surface area (TPSA), lipophilicity based on consensus log Po/w, water solubility (Log S), gastrointestinal absorption influence on drug metabolising enzymes of the liver (especially, CYP 1A2, CYP3A4), tendency to comply with Lipinski’s and Veber’s rule, and bioavailability score were successfully predicted and reported Table S2. Besides, the bioavailability radar maps constituting a cluster of selective parameters such as lipophilicity, molecular size, polarity, water solubility, molecular flexibility, and saturation were evaluated (only for the best five analogues) by means of the Swiss-ADME tool to enable the prediction of possible drug-likeness Figure S2.
The molar refractivity (M R) represents the size and polarity of a given molecule, for which the acceptable range is between 40 and 130, with an average value of 97 (Ghose et al., 1999). Most of the molecules are found to be in conformity except Compounds PubChem CIDs 132097620 (131.70), 145072656 (131.21), 146554000 (131.61), 162656940 (133.94), 138567683 (131.24), 15341467 (144.44), and 88093910 (133.16). The polar surface area or TPSA deals with the measure of polarity on the basis of the total number of polar atoms, especially oxygen and nitrogen. It is one of the helpful metrics to predict the penetrating ability of the molecules, where a polar surface area of >140 A°2 tends to decrease the penetrability, while an area of <97 A°2 is necessary for the molecules to cross the blood-brain-barrier successfully. Based on this, Compound PubChem CID 54682869, compound PubChem CID 123382950, Compound PubChem CID 145778325, and compound PubChem CID 54689919 with TPSA 145.58, 169.03, 158.22, and 145.58 A°2 have violated the above qualifying range in contrast to all the remaining molecules. In 1997, Christopher Lipinski postulated a set of rules with an objective to define the term “druglikeness qualitatively”, and these rules together are called as Lipinski’s rule of five Ro5 [34]. According to this, a molecule should have a molecular weight of ≤ 500 g/mol, ≤5 hydrogen bond donors, ≤ 10 hydrogen bond acceptors (HBA), and lipophilicity, which is represented as log p of ≤5, respectively. Hence, to categorise any molecule as having druglike qualities, it should strictly obey the Ro5. Here, excluding Compound PubChem CID 15341467 and Compound PubChem CID 91809445, all the other analogues comply with Ro5. However, the consensus Log Po/w that reflects the lipophilicity is slightly above the prescribed limits for the Compound PubChem CID 15341467 (5.77).
Compound ID 91809445 showed one violation related to the number of hydrogen bond acceptors (12 HBAs). Based on Log S values, the analogues were classified as poorly, moderately, or highly soluble. The BOILED-EGG model predicted that most analogues have a high probability of oral gastrointestinal absorption, except for several listed compounds, while brain permeability was generally low, excluding a few exceptions. Several analogues, including CIDs 138508665 and 138508964, were identified as substrates of P-glycoprotein, a key efflux pump involved in toxin removal and multidrug resistance. Cytochrome P450 enzymes, particularly CYP3A4, were more commonly affected by the analogues than CYP1A2 [35]. The top two analogues (CIDs 138508665 and 138508964) are predicted substrates for both enzymes, suggesting potential impact on drug metabolism [36]. All molecules adhered to Lipinski’s rule of five, although four violated Veber’s criteria [36].
Finally, the bioavailability status is semi-quantitatively scored on the basis of TPSA, total charge, and Lipinski’s rule, which enables the classification of compounds with probabilities including 11%, 17%, 56%, or 85%. The probability assigned by the scoring will have F>10% in the rat [37]. All the compounds show a uniform bioavailability score of 0.55, indicating that around 55% of the compound is supposed to reach unchanged circulation (Table S2). In connection to this, schematic bioavailability radar maps of the seven molecules comprising the best five analogues (Compounds PubChem CIDs 138508665, 138508964,132097535, 138508669, and 123812032), MK−8722, and berberine are presented below that enable theoretical prediction of their oral bioavailability based on the selected descriptors including lipophilicity, size, polarity, water solubility, flexibility (as number of rotable bonds), and saturation (as fraction of SP3 carbons) on each axis of the radar Figure S2A–G. The shaded region in the centre indicates the range within which a molecule’s radar plot should fall to pass the oral bioavailability check. Here, the radar plots of all seven molecules are found to lie within the range (inside the shaded region), showing no violations with regard to the requirements for oral bioavailability.
Unmanageable toxicity is one of the key attributes for around 30% failures during clinical trials. Most of the time, the drug optimisation programs fail to monitor the balance between efficacy and toxicity, which ultimately leads to the selection of ineligible candidate drugs [28]. Hence, it is immensely necessary to manage toxicity at various stages of drug development. ADMETlab 2.0 is an invaluable computational tool that facilitates the reliable prediction of toxicokinetic parameters. Here, we have successfully predicted various toxicokinetic parameters of MK−8722 and the best 20 analogues with the help of the above tool, Table S3. All the obtained scores or values for the selected parameters were compared and contrasted with the reference ranges prescribed by the tool.
All analogues demonstrated permeability values near the optimal range (>5.15 log units), though most exhibited higher plasma protein binding (91%–100%), except for PubChem CID 123858811. The volume of distribution (Vd) for all compounds ranged between 1 and 3.6 l/kg, falling within the acceptable range of 0.04–20 l/kg. The fraction unbound in plasma (Fu), critical for pharmacodynamic and pharmacokinetic effectiveness, was low (<5%) to moderate (5%–20%) for most molecules, with only CID 123858811 approaching the higher threshold at 19.7%. Several compounds, including CIDs 124005037 and 123360613, showed moderate unbound fractions. All analogues had a short biological half-life of less than 3 hours, indicating rapid elimination and likely necessitating frequent dosing for therapeutic efficacy [38,39].
Most analogues show no significant AMES toxicity, skin sensitization, carcinogenicity, or respiratory toxicity, with exceptions including PubChem CIDs 138508665, 145403788, 152849090 (AMES toxicity), and 124005037, 152849090 (carcinogenicity). Several compounds, such as 135397715, 132097535, and 145403788, Taxus brevifolia-derived compound exhibit varying degrees of respiratory toxicity [40]. Moderate HERG blocking was observed in CIDs 123858811 and 132097620, while others showed poor or moderate HERG blocking potential. The analogues generally present a risk for liver toxicity and rat acute oral toxicity; however, CIDs 148879356 and 162105203 are mild to moderate liver toxicity inducers, and CIDs 123812032, 132097498, 145403788, and 152849090 are poor to moderately poor inducers of rat acute oral toxicity.
3.6. Enrichment analysis of biological processes and genes
Use p-values or enrichment scores from the gene set enrichment results generated by ShinyGO as the input data for clustering. Utilise ShinyGO’s graphical interface to apply hierarchical clustering to the enriched pathways or gene ontology terms, resulting in a dendrogram/tree representation reflecting functional similarity. Furthermore, an enriched pathways network was generated, and the network visualisation feature creates a network graph that connects enriched pathways based on gene overlap or functional relationships (Figure 5A–B). Eventually, biological processes and gene association were illustrated in Figure 6, while the size of the circle at the end of each bar represents the number of genes involved in the respective process, with larger circles indicating more genes (ranging from 20 to 40). The colour of the bars and circles reflects the statistical significance of the enrichment, expressed as the negative logarithm of the false discovery rate (-log10 FDR). Colours range from purple, indicating lower significance (around 28), to red, indicating higher significance (around 36), as shown in the colour legend. Biological processes featured include glucose homeostasis, carbohydrate homeostasis, cellular response to insulin stimulus, and response to endogenous stimulus. Thereafter, the top 20 most frequently studied pathways involved in regulating type 2 diabetes mellitus. The enrichment FDR was expressed as a percentage based on the nominal p-value from the hypergeometric test. Only pathways falling within the specified size limits are included in the enrichment analysis. Following the analysis, pathways are filtered according to a user-defined FDR cutoff. The significant pathways are then sorted based on FDR, Fold Enrichment, or other selected metrics. Choosing “Sort by average ranks (FDR & Fold)” orders pathways by the average of their ranks in both FDR and Fold Enrichment. Selecting “Select by FDR, sort by Fold Enrichment” first filters pathways by FDR and then sorts them by Fold Enrichment. When “Remove redundancy” is enabled, pathways that share at least 95% of their genes and 50% of the words in their names are consolidated, with the most significant pathway representing the group [31]. Additionally, longer pathway names are truncated to the first 80 characters in this mode. (Data were analysed using open-source software “ShinyGO 0.76”) The details results are summarised in Table 4.
 | Figure 5. (A) Hierarchical clustering tree visually represents the relationships among significant pathways shown in the enrichment tab, grouping those with many shared genes together. Larger dots correspond to pathways with more significant p-values. The width of the display adjusts automatically according to your browser window size. (b) Similarly, the interactive plot illustrates connections between enriched pathways, where two pathways (nodes) link if they share 20% or more of their genes (default setting). Nodes can be repositioned by dragging, zoomed with the scroll wheel, and the entire network can be moved by clicking and dragging on an empty area. In this plot, darker nodes indicate pathways with higher significance, bigger nodes correspond to larger gene sets, and thicker connecting lines reflect greater gene overlap. [Click here to view] |
 | Figure 6. Enriched biological processes and gene association. This bar plot displays the fold enrichment of selected biological processes. The length of each horizontal bar corresponds to the fold enrichment value shown on the x-axis. The y-axis lists the biological processes analysed. [Click here to view] |
Table 4. The table below lists the top 20 most frequently studied pathways involved in the regulation of type 2 diabetes mellitus. The enrichment FDR was expressed as a percentage and based on the nominal p-value from the hypergeometric test. Fold Enrichment represents the ratio of the percentage of genes in your list that belong to a pathway to the percentage of genes in that pathway within the background set. It indicates the probability that the observed enrichment occurred by chance. Further, larger pathways often have smaller FDRs due to greater statistical power. However, fold enrichment serves as a key measure of effect size, reflecting the extent to which genes from a given pathway are overrepresented, making it a crucial but sometimes overlooked metric.
| Enrichment FDR | No. Genes | Pathway Genes | Fold Enrichment | Pathway | URL | Genes |
|---|---|---|---|---|---|---|
| 6.43E-28 | 22 | 259 | 39.52 | Glucose homeostasis | http://amigo.geneontology.org/amigo/term/GO:0042593 | ABCC8 HNF4A GCK G6PC1 PPARG PRKAA1 HNF1A IL6 PDX1 AKT1 PIK3R1 TCF7L2 FOXO1 NEUROD1 IRS1 LEP UCP2 ADIPOQ IRS2 KCNJ11 INS SLC2A4 |
| 6.54E-28 | 22 | 260 | 39.37 | Carbohydrate homeostasis | http://amigo.geneontology.org/amigo/term/GO:0033500 | ABCC8 HNF4A GCK G6PC1 PPARG PRKAA1 HNF1A IL6 PDX1 AKT1 PIK3R1 TCF7L2 FOXO1 NEUROD1 IRS1 LEP UCP2 ADIPOQ IRS2 KCNJ11 INS SLC2A4 |
| 5.50E-25 | 20 | 241 | 38.61 | Cellular response to insulin stimulus | http://amigo.geneontology.org/amigo/term/GO:0032869 | SREBF1 GSK3B GCK G6PC1 PPARG PRKAA1 AKT1 PIK3R1 FOXO1 IGF2 IRS1 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 PTPN1 INS SLC2A4 |
| 1.57E-26 | 22 | 301 | 34.00 | Response to insulin | http://amigo.geneontology.org/amigo/term/GO:0032868 | ABCC8 SREBF1 GSK3B GCK G6PC1 PPARG PRKAA1 AKT1 PIK3R1 FOXO1 IGF2 IRS1 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 PTPN1 MTOR INS SLC2A4 |
| 2.20E-28 | 24 | 352 | 31.72 | Cellular response to peptide hormone stimulus | http://amigo.geneontology.org/amigo/term/GO:0071375 | SREBF1 GSK3B JAK2 HNF4A CAV1 GCK NFKB1 G6PC1 PPARG PRKAA1 AKT1 PIK3R1 FOXO1 IGF2 IRS1 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 PTPN1 INS SLC2A4 |
| 7.82E-30 | 26 | 421 | 28.73 | Cellular response to peptide | http://amigo.geneontology.org/amigo/term/GO:1901653 | IGF1 SREBF1 GSK3B JAK2 HNF4A CAV1 GCK NFKB1 G6PC1 PPARG PRKAA1 CD36 AKT1 PIK3R1 FOXO1 IGF2 IRS1 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 PTPN1 INS SLC2A4 |
| 1.17E-28 | 26 | 473 | 25.57 | Response to peptide hormone | http://amigo.geneontology.org/amigo/term/GO:0043434 | ABCC8 SREBF1 GSK3B JAK2 HNF4A CAV1 GCK NFKB1 G6PC1 PPARG PRKAA1 AKT1 PIK3R1 FOXO1 IGF2 IRS1 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 PTPN1 MTOR INS SLC2A4 |
| 6.87E-30 | 28 | 562 | 23.18 | Response to peptide | http://amigo.geneontology.org/amigo/term/GO:1901652 | ABCC8 IGF1 SREBF1 GSK3B JAK2 HNF4A CAV1 GCK NFKB1 G6PC1 PPARG PRKAA1 CD36 AKT1 PIK3R1 FOXO1 IGF2 IRS1 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 PTPN1 MTOR INS SLC2A4 |
| 4.47E-28 | 27 | 581 | 21.62 | Reg. of protein transport | http://amigo.geneontology.org/amigo/term/GO:0051223 | ABCC8 IGF1 NR1H3 SREBF1 GSK3B JAK2 MAPK1 HNF4A GCK CPT1A PRKAA1 CD36 IL6 PDX1 PIK3R1 TCF7L2 NEUROD1 SLC30A8 IRS1 BCL2 LEP UCP2 ADIPOQ IRS2 KCNJ11 PTPN1 INS |
| 1.20E-27 | 27 | 609 | 20.63 | Reg. of establishment of protein localization | http://amigo.geneontology.org/amigo/term/GO:0070201 | ABCC8 IGF1 NR1H3 SREBF1 GSK3B JAK2 MAPK1 HNF4A GCK CPT1A PRKAA1 CD36 IL6 PDX1 PIK3R1 TCF7L2 NEUROD1 SLC30A8 IRS1 BCL2 LEP UCP2 ADIPOQ IRS2 KCNJ11 PTPN1 INS |
| 3.68E-29 | 29 | 691 | 19.52 | Cellular response to organonitrogen compound | http://amigo.geneontology.org/amigo/term/GO:0071417 | IGF1 SREBF1 GSK3B JAK2 MAPK1 HNF4A MAPK3 CAV1 GCK NFKB1 G6PC1 PPARG PRKAA1 CD36 AKT1 PIK3R1 FOXO1 IGF2 IRS1 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 PTPN1 MTOR INS SLC2A4 |
| 1.05E-30 | 33 | 998 | 15.38 | Response to hormone | http://amigo.geneontology.org/amigo/term/GO:0009725 | ABCC8 NR1H3 SREBF1 GSK3B JAK2 MAPK1 HNF4A CAV1 GCK NFKB1 G6PC1 PPARG PRKAA1 IL6 PDX1 AKT1 HSPD1 PIK3R1 FOXO1 IGF2 IRS1 BCL2 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 PTPN1 MTOR INS TXNIP SLC2A4 |
| 4.00E-32 | 35 | 1121 | 14.53 | Response to organonitrogen compound | http://amigo.geneontology.org/amigo/term/GO:0010243 | ABCC8 IGF1 SREBF1 GSK3B JAK2 MAPK1 HNF4A MAPK3 CAV1 GCK NFKB1 CPT1A G6PC1 PPARG PRKAA1 CD36 IL6 PDX1 AKT1 HSPD1 PIK3R1 FOXO1 IGF2 IRS1 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 KCNJ11 PTPN1 MTOR INS SLC2A4 |
| 7.02E-33 | 37 | 1313 | 13.11 | Cellular response to oxygen-containing compound | http://amigo.geneontology.org/amigo/term/GO:1901701 | IGF1 NR1H3 SREBF1 GSK3B JAK2 MAPK1 HNF4A MAPK3 CAV1 GCK NFKB1 CPT1A SOD2 G6PC1 PPARG PRKAA1 CD36 IL6 PDX1 AKT1 PIK3R1 FOXO1 NEUROD1 IGF2 IRS1 CXCL8 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 KCNJ11 PTPN1 MTOR INS SLC2A4 |
| 2.24E-26 | 32 | 1264 | 11.78 | Chemical homeostasis | http://amigo.geneontology.org/amigo/term/GO:0048878 | ABCC8 NR1H3 JAK2 MAPK1 HNF4A MAPK3 CAV1 GCK SOD2 G6PC1 PPARG PRKAA1 HNF1A CD36 IL6 PDX1 AKT1 PIK3R1 TCF7L2 FOXO1 NEUROD1 SLC30A8 IRS1 FABP4 BCL2 LEP UCP2 ADIPOQ IRS2 KCNJ11 INS SLC2A4 |
| 1.87E-39 | 44 | 1832 | 11.17 | Response to oxygen-containing compound | http://amigo.geneontology.org/amigo/term/GO:1901700 | ABCC8 IGF1 NR1H3 SREBF1 GSK3B JAK2 MAPK1 HNF4A MAPK3 CAV1 GCK NFKB1 CPT1A SOD2 G6PC1 PPARG PRKAA1 HNF1A CD36 IL6 PDX1 AKT1 HSPD1 PIK3R1 TCF7L2 FOXO1 NEUROD1 SLC30A8 IGF2 IRS1 CXCL8 BCL2 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 KCNJ11 PTPN1 MTOR INS TXNIP SLC2A4 |
| 8.56E-29 | 37 | 1769 | 9.73 | Response to endogenous stimulus | http://amigo.geneontology.org/amigo/term/GO:0009719 | ABCC8 IGF1 NR1H3 SREBF1 GSK3B JAK2 MAPK1 HNF4A MAPK3 CAV1 GCK NFKB1 G6PC1 PPARG PRKAA1 CD36 IL6 PDX1 AKT1 HSPD1 PIK3R1 FOXO1 IGF2 IRS1 CXCL8 BCL2 RELA LEP UCP2 ADIPOQ SOCS3 IRS2 PTPN1 MTOR INS TXNIP SLC2A4 |
4. DISCUSSION
AMPK’s role extends to regulating mitochondrial biogenesis, redox homeostasis, cell growth, proliferation, and inflammation, all relevant to diabetic complications. AMPK inhibition mechanisms are critical for understanding its physiological roles and therapeutic modulation in metabolic disorders like diabetes. The enzyme’s activity is primarily controlled by allosteric AMP and ATP regulation, auto-inhibitory features, and phosphorylation of its catalytic (α) and regulatory (β and γ) subunits [41]. Key to inhibition is the dephosphorylation of Thr-172 on the α-subunit, which is the primary activation site. Compounds like BAY-3827 inhibit AMPK by binding to the ATP-binding pocket and inducing an αC helix-out conformation, stabilizing the activation loop through a disulfide bridge between Cys106 and Cys174, which ultimately disrupts the regulatory spine and promotes an inactive state [42]. Similarly, our top rank compounds, such as MK-8722, are potent, direct, allosteric activators targeting all 12 mammalian AMPK complexes [43]. In animal models, MK−8722 improves glycemia by inducing robust, durable, and insulin-independent glucose uptake and glycogen synthesis in skeletal muscle [43]. This makes it a potential therapeutic for diabetes by addressing glucose homeostasis [44].
Whereas, Berberine improves glucose tolerance, reduces body weight, lowers plasma triglycerides, and enhances insulin action in animal models of insulin resistance [45]. It increases GLUT4 translocation, reduces lipid accumulation, and downregulates lipogenic genes while upregulating those involved in energy expenditure. The molecular mechanism for activating AMPK by Thr172 phosphorylation and actively, often through mitochondrial-derived superoxide anions and peroxynitrite. Although it involves effects, such as direct inhibition of mitochondrial respiratory chain complex I and further ameliorates high glucose-induced cardiomyocyte injury by activating AMPK signalling to stimulate mitochondrial biogenesis and restore autophagic flux [46]. Seemingly, CID 138508665—these compounds, such as AICAR (5-aminoimidazole-4-carboxamide riboside), mimic AMP by binding to the γ-subunit, triggering phosphorylation of the catalytic α-subunit by upstream kinases like LKB1 and CaMKK-β [47]. This compound presents isobutyryloxymethyl groups, which are metabolized to formaldehyde and inhibit the function of mitochondria by increasing the AMP: ATP ratio, further leading to activating AMPK through canonical pathways [48]. Eventually, CID 138508964—key function for activation of AMPK by regulating energy homeostasis. This compound [49,50]. It plays a significant role in diabetes management as AMPK activation enhances glucose uptake in skeletal muscles and fatty acid oxidation and reduces hepatic glucose production. Thereby improving the insulin sensitivity [51].
The binding poses of potent AMPK activators such as MK-8722, Berberine, and the identified analogues 138508665 and 138508964 are hypothesized to mechanistically lead to AMPK activation by influencing the ADaM site, a critical regulatory region formed at the interface of the kinase domain (KD) of the α-subunit and the carbohydrate-binding module (CBM) of the ß-subunit [52]. MK−8722 is a pan-AMPK activator that directly and allosterically activates all 12 mammalian AMPK complexes, leading to robust glucose uptake and glycogen synthesis in skeletal muscle [53]. This suggests that MK−8722 likely binds within the ADaM site, or a similar allosteric pocket, to stabilize an active conformation of AMPK, possibly by reducing the conformational heterogeneity of the KD-CBM interface, as observed with other ADaM site ligands like Merck 991 [54]. Berberine, on the other hand, activates AMPK through several mechanisms, including the generation of ROS and activation of upstream kinases like Liver Kinase B1, and also by affecting lysosomal AMPK and maintaining cellular AMPK activity through inhibiting the dephosphorylation regulator UHRF1 [55]. While Berberine’s direct binding to the ADaM site has not been as extensively characterized as MK−8722, its effects suggest it might either interact with allosteric sites or indirectly influence the ADaM site’s conformational stability through upstream signaling pathways. For analogues like 138508665 and 138508964, their high potency would suggest a similar ADaM site interaction, where their specific chemical structures might induce a conformational shift that mimics the effect of AMP binding to the γ-subunit, thereby protecting the activation loop from dephosphorylation and stabilizing the active state of the kinase [40,56–61].
5. CONCLUSION
To sum up, the present study was devised to screen a library of selective small molecules that are structural analogues of some well-known natural and synthetic AMPK activators, against the crystal structure of AMPK (PDB ID: 6B2E) using a combination of reliable computational tools. The study has reasonably succeeded in revealing some valid in silico hits that could set a stage for the development of some interesting prospective prototype-kind of candidate α2β2γ1 AMPK. Based on this, MK-8722, berberine, compound PubChem CID 138508665, and compound PubChem CID 138508964 are found to be top-ranked molecules from the standard and query (analogue) datasets, respectively. Finding berberine as one among, is expected to broaden the scope for the development of diverse libraries of natural products-inspired activators with better efficacy and safety. Though some of the molecules are very likely to be carcinogenic, they show a likelihood of liver toxicity, which chemical manipulations may minimise. Apart from this, the study has also unveiled certain allosteric regions that are probably conducive to the binding of small molecules. Finally, our efforts towards screening and identification of selective β2-AMPK complex activators will continue to expand further by applying a combination of advanced chemical, biochemical, and computational approaches. Moreover, the enrichment analysis of biological processes and genes was performed successfully, and the cluster network was created.
6. THE LIMITATIONS
To validate the computationally identified AMPK activators, top-ranked small molecules will be synthesized or procured for in vitro evaluation. Their binding affinity and activation potential will be assessed using kinase activity assays with recombinant human α2β2γ1 AMPK complex. Selectivity will be tested against nonmuscle AMPK isoforms. Furthermore, cellular assays using C2C12 myotubes will evaluate glucose uptake, AMPK phosphorylation (Thr172), and downstream signalling pathways via Western blot. Cytotoxicity will be assessed using the MTT assay. Promising candidates will proceed to in vivo studies in a Type-2 diabetic mouse model to assess glucose tolerance, insulin sensitivity, and skeletal muscle AMPK activation. Additionally, we can develop a nano carrier drug delivery system for ameliorating T2DM.
7. ACKNOWLEDGMENTS
The authors would like to acknowledge the director of BITS-Pilani Hyderabad Campus. We also immensely thank Siksha O Anusandhan, Bhubaneswar, India, for facilitating the molecular dynamics (MD) simulations and density functional theory (DFT) studies.
8. LIST OF ABBREVIATIONS
AMPK: AMP-Activated protein kinase; ADMET: Pharmacokinetics and toxicity; CID: PubChem compound identifier; LGA: Lamarckian genetic algorithm; MD: Molecular dynamics; MM/GBSA: Molecular mechanics/generalised born surface area; PDB: Protein databank; T2DM: Type 2 diabetes mellitus.
9. AUTHOR CONTRIBUTION
All authors made substantial contributions to conception and design, acquisition of data, or analysis and interpretation of data; took part in drafting the article or revising it critically for important intellectual content; agreed to submit to the current journal; gave final approval of the version to be published; and agree to be accountable for all aspects of the work. All the authors are eligible to be an author as per the International Committee of Medical Journal Editors (ICMJE) requirements/guidelines.
10. FINANCIAL SUPPORT
There is no funding to report.
11. CONFLICTS OF INTEREST
The authors report no financial or any other conflicts of interest in this work.
12. ETHICAL APPROVALS
This study does not involve experiments on animals or human subjects.
13. AVAILABILITY OF DATA AND MATERIAL
All the data is available with the authors and shall be provided upon request.
14. USE OF ARTIFICIAL INTELLIGENCE (AI)-ASSISTED TECHNOLOGY
The authors declare that they have not used artificial intelligence (AI)-tools for writing and editing of the manuscript, and no images were manipulated using AI.
15. SUPPLEMENTARY MATERIAL
The supplementary material can be accessed at the link here: https://japsonline.com/admin/php/uploadss/4732_pdf.pdf
REFERENCES
1. Kumar A, Maurya P, Narayan KP. Fabrication of polymeric matrix type transdermal patch containing glimepiride for therapeutic management of type-2 diabetes. Indian J Pharm Educ Res. 2025;59(3):936–48. CrossRef
2. Diabetes facts and figures show the growing global diabetes burden for individuals, families and countries. Brussels, Belgium. Available from: https://idf.org/about-diabetes/diabetes-facts-figures/
3. Antar SA, Ashour NA, Sharaky M, Khattab M, Ashour NA, Zaid RT, et al. Diabetes mellitus: classification, mediators, and complications; a gate to identify potential targets for the development of new effective treatments. Biomed Pharmacother. 2023;168:115734. CrossRef
4. IDF: IDF Atlas 11th Edition 2025-1. In., 11th Edition - 2025 edn: International diabetes Federation 2025. Brussels, Belgium: International Diabetes Federation. Available from: https://diabetesatlas.org/media/uploads/sites/3/2025/04/IDF_Atlas_11th_Edition_2025.pdf
5. Scott LJ, Erdos MR, Huyghe JR, Welch RP, Beck AT, Wolford BN, et al. The genetic regulatory signature of type 2 diabetes in human skeletal muscle. Nat Commun. 2016;7:11764. CrossRef
6. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, et al. The Protein Data Bank. Nucleic Acids Res. 2000;28(1):235–42. CrossRef
7. Benziane B, Bjornholm M, Lantier L, Viollet B, Zierath JR, Chibalin AV. AMP-activated protein kinase activator A-769662 is an inhibitor of the Na(+)-K(+)-ATPase. Am J Physiol Cell Physiol. 2009;297(6):C1554–66. CrossRef
8. Kumar A, Sakhare K, Bhattacharya D, Chattopadhyay R, Parikh P, Narayan KP, et al. Communication in non-communicable diseases (NCDs) and role of immunomodulatory nutraceuticals in their management. Front Nutr. 2022;9:966152. CrossRef
9. Kumar A, Dhiman C, Kumar M, Kannappan N, Kumar D, Chourasia MK, et al. Development of a quality by design based hybrid RP-HPLC method for Glimepiride: bioanalytical and industrial applications. J Appl Pharm Sci. 2025 ;15(4). CrossRef
10. Lin C, Sun G, Li Y. Identification of AdipoRon analogues as novel activators of AMPK for the treatment of type 2 diabetes. RSC Med Chem. 2024;15(7):2413–21. CrossRef
11. Conejeros I, Velasquez ZD, Espinosa G, Rojas-Baron L, Grabbe M, Hermosilla C, et al. AMPK and CAMKK activation participate in early events of Toxoplasma gondii-triggered NET formation in bovine polymorphonuclear neutrophils. Front Vet Sci. 2025;12:1557509. CrossRef
12. Edmunds LR, Sharma L, Wang H, Kang A, d’Souza S, Lu J, et al. c-Myc and AMPK control cellular energy levels by cooperatively regulating mitochondrial structure and function. PLoS One. 2015;10(7):e0134049. CrossRef
13. Wang M, Wang B, Zhou S, Liu J, Lu H, Wu H, et al. Quercetin ameliorates chicken quality by activating the PI3K/PKB/AMPK signaling pathway in broilers. Front Vet Sci. 2022;9:951512. CrossRef
14. Khater SI, Almanaa TN, Fattah DMA, Khamis T, Seif MM, Dahran N, et al. Liposome-encapsulated berberine alleviates liver injury in type 2 diabetes via promoting AMPK/mTOR-mediated autophagy and reducing ER stress: morphometric and immunohistochemical scoring. Antioxidants (Basel). 2023;12(6):1220. CrossRef
15. Niimi T, Kitagawa Y. Distinct roles of mouse laminin beta1 long arm domains for alpha1beta1gamma1 trimer formation. FEBS Lett. 1997;400(1):71–4. CrossRef
16. Hubbard JA, Xiao B, Wilson JR. Production and crystallization of full-length human AMP-activated protein kinase (alpha1beta1gamma1). Methods Mol Biol. 2018;1732:1–14. CrossRef
17. Treebak JT, Birk JB, Hansen BF, Olsen GS, Wojtaszewski JF. A-769662 activates AMPK β1-containing complexes but induces glucose uptake through a PI3-kinase-dependent pathway in mouse skeletal muscle. Am J Physiol-Cell Physiol. 2009;297(4):C1041–52. CrossRef
18. Im E, Chung KC. Precise assembly and regulation of 26S proteasome and correlation between proteasome dysfunction and neurodegenerative diseases. BMB Rep. 2016;49(9):459. CrossRef
19. Scott JW, Ling N, Issa SM, Dite TA, O’Brien MT, Chen ZP, et al. Small molecule drug A-769662 and AMP synergistically activate naive AMPK independent of upstream kinase signaling. Chem Biol. 2014;21(5):619–27. CrossRef
20. Ottana R, Paoli P, Cappiello M, Nguyen TN, Adornato I, Del Corso A, et al. In search for multi-target ligands as potential agents for diabetes mellitus and its complications-a structure-activity relationship study on inhibitors of aldose reductase and protein tyrosine phosphatase 1B. Molecules. 2021;26(2):330. CrossRef
21. Schneider G, Böhm HJ. Virtual screening and fast automated docking methods. Drug Discov Today. 2002;7:64–70. CrossRef
22. Enejoh OA, Okonkwo CH, Nortey H, Kemiki OA, Moses A, Mbaoji FN, et al. Machine learning and molecular dynamics simulations predict potential TGR5 agonists for type 2 diabetes treatment. Front Chem. 2024;12:1503593. CrossRef
23. Daude RB, Bhadane R, Shah JS. Alpha-cyperone mitigates renal ischemic injury via modulation of HDAC-2 expression in diabetes: insights from molecular dynamics simulations and experimental evaluation. Eur J Pharmacol. 2024;975:176643. CrossRef
24. Hollingsworth SA, Dror RO. Molecular dynamics simulation for all. Neuron. 2018;99(6):1129–43. CrossRef
25. Wang E, Sun H, Wang J, Wang Z, Liu H, Zhang JZ, et al. End-point binding free energy calculation with MM/PBSA and MM/GBSA: strategies and applications in drug design. Chem Rev. 2019;119(16):9478–508. CrossRef
26. Hou T, Wang J, Li Y, Wang W. Assessing the performance of the MM/PBSA and MM/GBSA methods. 1. The accuracy of binding free energy calculations based on molecular dynamics simulations. J Chem Inform Modeling. 2011;51(1):69–82. CrossRef
27. Bartolotti LJ, Flurchick K. An introduction to density functional theory. Rev Comput Chem. 1996;7:187–216. CrossRef
28. Sun D, Gao W, Hu H, Zhou S. Why 90% of clinical drug development fails and how to improve it? Acta Pharm Sinica B. 2022;12(7):3049–62. CrossRef
29. Daina A, Michielin O, Zoete V. SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci Rep. 2017;7(1):42717. CrossRef
30. Raza K. Reconstruction, topological and gene ontology enrichment analysis of cancerous gene regulatory network modules. Curr Bioinform. 2016;11(2):243–58. CrossRef
31. Lu JM, Chen YC, Ao ZX, Shen J, Zeng CP, Lin X, et al. System network analysis of genomics and transcriptomics data identified type 1 diabetes-associated pathway and genes. Genes Immun. 2019;20(6):500–8. CrossRef
32. Carbonetto P, Stephens M. Integrated enrichment analysis of variants and pathways in genome-wide association studies indicates central role for IL-2 signaling genes in type 1 diabetes, and cytokine signaling genes in Crohn’s disease. PLoS Genet. 2013;9(10):e1003770. CrossRef
33. Bhushan R, Rani A, Ali A, Singh VK, Dubey PK. Bioinformatics enrichment analysis of genes and pathways related to maternal type 1 diabetes associated with adverse fetal outcomes. J Diabetes Complications .2020;34(5):107556. CrossRef
34. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev. 2012;64:4–17. CrossRef
35. Hollenberg PF. Characteristics and common properties of inhibitors, inducers, and activators of CYP enzymes. Drug Metabol Rev. 2002;34(1-2):17–35. CrossRef
36. Muna UM, Hafiz F, Biswas S, Azim R. GBDTSVM: combined support vector machine and gradient boosting decision tree framework for efficient snoRNA-disease association prediction. Comput Biol Med. 2025;192(Pt A):110219. CrossRef
37. Martin YC. A bioavailability score. J Med Chem. 2005;48(9):3164–70. CrossRef
38. Matin MM, Chakraborty P, Alam MS, Islam MM, Hanee U. Novel mannopyranoside esters as sterol 14alpha-demethylase inhibitors: synthesis, PASS predication, molecular docking, and pharmacokinetic studies. Carbohydr Res. 2020;496:108130. CrossRef
39. Matin MM, Bhattacharjee SC, Chakraborty P, Alam MS. Synthesis, PASS predication, in vitro antimicrobial evaluation and pharmacokinetic study of novel n-octyl glucopyranoside esters. Carbohydr Res. 2019;485:107812. CrossRef
40. Singh R, Kumar A, Srivastava D, Narayan KP, Chaurasia M, Flora S. Nano-formulation technology for simultaneous paclitaxel and andrographolide delivery: pre-formulation insights. Essential Chem. 2025;2(1):2533990. CrossRef
41. Viollet B, Horman S, Leclerc J, Lantier L, Foretz M, Billaud M, et al. AMPK inhibition in health and disease. Crit Rev Biochem Mol Biol. 2010;45(4):276–95. CrossRef
42. Fraguas Bringas C, Ahangar MS, Cuenco J, Liu H, Addinsall AB, et al. Mechanism and cellular actions of the potent AMPK inhibitor BAY-3827. Sci Adv. 2025;11(34). CrossRef
43. Myers RW, Guan HP, Ehrhart J, Petrov A, Prahalada S, Tozzo E, et al. Systemic pan-AMPK activator MK-8722 improves glucose homeostasis but induces cardiac hypertrophy. Science. 2017;357(6350):507–11. CrossRef
44. Cottignies-Calamarte A, Marteau F, He F, Belouzard S, Dubuisson J, Tudor D, et al. Direct pharmacological AMPK activation inhibits mucosal SARS-CoV-2 infection by reducing lipid metabolism, restoring autophagy flux and the type I IFN response. J Virol 2025:e00394–25. CrossRef
45. Lee YS, Kim WS, Kim KH, Yoon MJ, Cho HJ, Shen Y, et al. Berberine, a natural plant product, activates AMP-activated protein Kinase with beneficial ,etabolic effects in diabetic and insulin-resistant states. Diabetes. 2006;55(8):2256–64. CrossRef
46. Hang W, He B, Chen J, Xia L, Wen B, Liang T, et al. Berberine ameliorates high glucose-induced cardiomyocyte injury via AMPK signaling activation to stimulate mitochondrial biogenesis and restore autophagic flux. Front Pharmacol. 2018;9:1121. CrossRef
47. Foretz M, Taleux N, Guigas B, Horman S, Beauloye C, Andreelli F, et al. Regulation of energy metabolism by AMPK: a novel therapeutic approach for the treatment of metabolic and cardiovascular diseases. Med Sci. 2006;22:381–8. CrossRef
48. Musi N. AMP-activated protein kinase and type 2 diabetes. Curr Med Chem. 2006;13(5):583–9. CrossRef
49. Sanz P. AMP-activated protein kinase: structure and regulation. Curr Protein Peptide Sci. 2008;9(5):478–92. CrossRef
50. Yanagi T, Kikuchi H, Takeuchi K, Susa K, Mori T, Takahashi N, et al. ULK1-regulated AMP sensing by AMPK and its application for the treatment of chronic kidney disease. 2023. CrossRef
51. Coughlan KA, Valentine RJ, Ruderman NB, Saha AK. AMPK activation: a therapeutic target for type 2 diabetes? Diabetes, Metabolic Syndrome and Obesity: Targets Therapy 2014;7:241–53. CrossRef
52. Yan Y, Zhou XE, Novick SJ, Shaw SJ, Li Y, Brunzelle JS, et al. Structures of AMP-activated protein kinase bound to novel pharmacological activators in phosphorylated, non-phosphorylated, and nucleotide-free states. J Biol Chem. 2019;294(3):953–67. CrossRef
53. Feng D, Biftu T, Romero FA, Kekec A, Dropinski J, Kassick A, et al. Discovery of MK-8722: a systemic, direct pan-activator of AMP-activated Protein Kinase. ACS Med Chem Lett. 2018;9(1):39–44. CrossRef
54. Gu X, Bridges MD, Yan Y, de Waal PW, Zhou XE, Suino-Powell KM, et al. Conformational heterogeneity of the allosteric drug and metabolite (ADaM) site in AMP-activated protein kinase (AMPK). J Biol Chem. 2018;293(44):16994–7007. CrossRef
55. Ren G, Ding Y, Wang LL, Jiang JD. Berberine stimulates lysosomal AMPK independent of PEN2 and maintains cellular AMPK activity through inhibiting the dephosphorylation regulator UHRF1. Front Pharmacol. 2023, 14. CrossRef
56. de Souza Almeida Matos AL, Oakhill JS, Moreira J, Loh K, Galic S, Scott JW. Allosteric regulation of AMP-activated protein kinase by adenylate nucleotides and small-molecule drugs. Biochem Soc Transactions. 2019;47(2):733–41. CrossRef
57. Kumar A, Rana R, Saklani R, Kumar M, Yadav PK, Tiwari A, et al. Technology transfer of a validated RP-HPLC method for the simultaneous estimation of andrographolide and paclitaxel in application to pharmaceutical nanoformulation. J Chromatogr Sci. 2024;62(4):356–63. CrossRef
58. Sharma N, Kumar A, Sambhakar S, Bhatia D, Hussain S, Mursal M, et al Recent advances in nanotherapeutics and theranostics for squamous cell carcinoma: a comprehensive review. Curr Drug Deliv. 2025. CrossRef
59. Kumar A, Dhiman C, Narayan KP, Kumar M, Sharma M. Nanotherapeutics: restoring gut-dysbiosis and inducing cancer apoptosis. In: Raghavan R, editor. Spatially variable genes in cancer: development, progression, and treatment response. Palmdale, PA: IGI Global Scientific Publishing; 2025. pp 177–204. CrossRef
60. Srivastava S, Kumar A, Yadav PK, Kumar M, Mathew J, Pandey AC, et al. Formulation and performance evaluation of polymeric mixed micelles encapsulated with baicalein for breast cancer treatment. Drug Dev Ind Pharm. 2021;47(9):1512–22. CrossRef\
61. Singh R, Kumar A, Srivastava D, Narayan KP, Chaurasia M, Flora SJ. Nano-formulation technology for simultaneous paclitaxel and andrographolide delivery: pre-formulation insights. Essential Chem. 2025 Dec 31;2(1):2533990. CrossRef