SGLT2 was selected as the target for docking due to its key role in glucose reabsorption and relevance to antidiabetic drugs. A high-quality PDB structure with an intact binding site was chosen. The .pdb file obtained from the RCSB PDB included atomic coordinates for every atom in the protein, and frequently, other molecules such as bound ligands, ions, cofactors, and water molecules. The cryo-EM structure of human SGLT2 was retrieved from the Protein Data Bank (PDB ID: 7VSI, resolution 3.29 Å), solved in complex with sodium ions and a bound inhibitor. The binding pocket was defined based on the co-crystallized ligand. Because SGLT2 is a membrane transporter, docking was performed without lipid bilayer simulation, which is acknowledged as a limitation.

After selecting SGLT2 from the RCSB PDB, the protein was prepared using AutoDock Tools by removing water molecules, ions, and ligands, adding polar hydrogens, assigning Kollman charges, checking structural completeness, minimizing geometry, and finally saving the optimized, docking-ready receptor structure in .pdbqt format for further simulations.

Ligand preparation involved designing a library of 30 dioxabicyclo[3.2.1]octane analogs based on the parent scaffold used in SGLT2 inhibitors like ertugliflozin. Using ChemDraw, structural modifications including functional group additions. A total of 30 analogs were finalized and catalogued, each with a unique structural feature designed to interrogate a specific aspect of receptor binding (Figures 1–30 and Table 1).

After constructing dioxabicyclo octane hybrids in ChemDraw 8.0, each ligand underwent energy minimization using Open Babel with the MMFF94 force field to obtain stable geometries. Minimized structures were converted via PyMOL into .pdb format, then prepared in AutoDock Tools by adding polar hydrogens, assigning Kollman charges, defining rotatable bonds, and finally saving each ligand as .pdbqt for docking.

To validate the docking protocol, the co-crystallized ligand was redocked into the extracted binding site. The best docked pose was compared with the experimental pose, and heavy-atom RMSD was calculated. An RMSD below 2.0 Å was considered acceptable. Additionally, enrichment analysis was performed using known SGLT2 inhibitors and property-matched decoys to assess the protocol’s discrimination ability. The docking procedure was repeated three times using different random seeds to verify reproducibility.

Molecular docking was performed using AutoDock Vina to predict binding poses and affinities of 30 dioxabicyclo octane hybrids within the SGLT2 active site. All receptor and ligand files (.pdbqt) were loaded into AutoDock Vina Tools for grid box setup, and configuration files were generated. Docking was executed through Windows Command Prompt using vina.exe, with individual config files for each ligand, producing output and log files for further analysis.

For each docking run, a separate configuration file was created corresponding to each compound. However, for batch processing, a template configuration was maintained and edited for each compound. The structure of a typical configuration file (config1.txt for Compound 1) was as follows:

The same structure was replicated for Compounds 2 to 30, adjusting the ligand, out, and log file names appropriately (e.g., compound2.pdbqt, compound2_out.pdbqt, log2.txt, etc.).

Docked complexes were visualized and analyzed using PyMOL to evaluate ligand orientation, binding poses, and interactions within the SGLT2 active site. PyMOL’s 3D tools enabled inspection of pocket fitting, hydrogen bonds, surface interactions, and ligand burial. Color schemes, transparency, surface views, labels, and clipping planes were applied to clearly examine atomic contacts and validate docking accuracy for all dioxabicyclo octane hybrids.

Thirty dioxabicyclo[3.2.1]octane hybrid ligands were energy-minimized using the MMFF94 force field in Open Babel to ensure stable, low-energy conformations before docking into SGLT2. Structures sketched in ChemDraw 8.0 were converted to .mol format and optimized under standard convergence criteria. Minimized energies ranged from –83.56 to –94.17 kcal/mol, reflecting varied conformational stabilities (Table 2).

In this phase of the study, molecular docking analysis was conducted to evaluate the binding efficiency and interaction patterns of Compound 1–30, the parent dioxabicyclo[3.2.1]octane scaffold, against the target protein Sodium-Glucose Cotransporter 2 (SGLT2). Table 3 includes the best binding affinity values (docking mode) and both lower and upper RMSD values for each mode of Compound 1–30.

Among the thirty compounds, significant variations in binding affinities were observed, ranging from moderate interactions (–5.1 kcal/mol) to very high binding affinity (–10.3 kcal/mol). The top five compounds showing the most promising docking performance in terms of lowest binding energy values and pose stability were Compound 24, Compound 25, Compound 26, Compound 23, and Compound 1.

The selection of the top five performing compounds was carried out strictly based on the best docking affinity obtained at the lowest RMSD values. After evaluating all the compounds (Compounds 1 through 30), the five best-performing molecules were identified as follows:

To understand the relative performance of the selected top five compounds, their docking modes were compared in terms of both affinity and pose convergence as described by RMSD values. The RMSD values reported by AutoDock Vina refer to pose clustering relative to the top-ranked pose rather than to an experimental reference. RMSD = 0 Å therefore indicates identical conformations between predicted poses and should not be interpreted as structural stability or binding accuracy. The following table presents a comparative snapshot of the top-ranked docking poses:

As shown in Table 4 and Figures 31–35, all five compounds exhibited high docking precision with zero RMSD values in their best poses, confirming consistent and reproducible conformational binding. The RMSD values reported by AutoDock Vina refer to pose clustering relative to the top-ranked pose rather than to an experimental reference. RMSD = 0 Å therefore indicates identical conformations between predicted poses and should not be interpreted as structural stability or binding accuracy. Compound 24 formed H-bonds with H80, Q457, π-stacking with F98, and hydrophobic contacts with L84, M146. The fused heterocycle enhances anchoring relative to parent scaffold. Compound 24, containing a fused nitrogen-heterocycle, achieved the strongest binding (–10.3 kcal/mol), indicative of robust electrostatic, hydrogen bonding, and possible aromatic stacking interactions within the SGLT2 binding pocket. This was closely followed by Compound 25 (–10.0 kcal/mol), which benefited from sulfur-containing fused thiophene, contributing to S-pocket complementarity and lipophilic contacts.

Compound 26 demonstrated a docking affinity of –9.7 kcal/mol and was also remarkable due to its fused imidazole ring—a known pharmacophore in several clinically active molecules targeting SGLT2. This hybrid may participate in multipoint interactions due to its nitrogen atoms and planar aromaticity. Compound 23, with a fused six-member aromatic ring, reached –8.2 kcal/mol, suggesting effective π–π stacking with key residues. While Compound 1 (–7.7 kcal/mol) appeared last in this elite group, it retained its significance as the control scaffold and provided a useful benchmark for interpreting the SAR trends among modified derivatives.

Following a detailed comparative assessment of docking performance, Compound 24 was identified as the best candidate among the thirty tested molecules. Its top binding affinity of –10.3 kcal/mol, coupled with a perfectly converged pose (RMSD = 0.000 Å), positioned it as the strongest ligand for SGLT2 inhibition among the screened series (Figure 31). The success of Compound 24 can be attributed to its unique fused five-membered nitrogen heterocycle, which likely enhances interaction diversity through hydrogen bonding, electrostatic contacts, and π-system complementarity within the active site.

Compound 25, though very close in performance, ranked second with a slightly lower binding energy (–10.0 kcal/mol). This compound, featuring a fused thiophene ring, also displayed a rigid and well-converged pose, signifying its promising potential. Compound 26 followed closely at third place, benefitting from its imidazole functionality, which is often found in bioactive molecules with efficient protein interaction profiles.

However, Compound 24 stands out due to its superior thermodynamic profile and likely higher probability of forming stable interactions with the binding residues of SGLT2. The presence of a nitrogen-rich fused heterocycle likely facilitates multiple donor–acceptor roles in the molecular recognition process, enhancing binding enthalpy and reducing entropic penalties. This balance makes Compound 24 the most favorable ligand in terms of binding strength, pose stability, and pharmacophore potential.