According to the literature, [1,2,4]triazolo[1,5-a]pyrimidines are typically synthesized using three main precursors: 1H-1,2,4-triazol-3-amine, 2-hydrazineylnicotinic acid, substituted 2,3-diaminopyrimidin-4(3H)-one, among which 1H-1,2,4-triazol-3-amine is the most frequently employed. This nucleophilic adduct went through the condensation with appropriate bifunctional electrophilic systems such as diethyl 2-(ethoxymethylene) malonate, α, β-unsaturated carbonyls and nitriles, 1,3-dicarbonyl compounds, ethyl 3-aryloxirane-2-carboxylate, as well as condensation with aldehydes and active methylenes, to prepare the desired triazolopyrimidine scaffold.
In present study, we described an innovative, straightforward route involving Michael addition-cyclization between α-azidochalcones 7 and 1H-1,2,4-triazol-3-amine 8 to yield novel derivatives of [1,2,4]triazolo[1,5-a]pyrimidines 9. Notably, vinyl azides have recently emerged as powerful and versatile synthons in organic synthesis to afford a wide range of biologically active nitrogen-containing heterocycles.
As outlined in Scheme 1, the synthetic approach toward the target 5,7-diaryl-[1,2,4]triazolo[1,5-a]pyrimidin-6-amines 9 began with the preparation of various α-azidochalcone derivatives 7. Initially, condensation reactions between different arylaldehydes 1 and 1-(aryl)ethan-1-ones 2 were carried out to construct the chalcone backbone, which was subsequently subjected to olefin bromination. The resulting intermediates then underwent nucleophilic substitution with sodium azide, followed by elimination of hydrogen bromide (HBr) to yield the corresponding α-azidochalcones 7. Finally, the reaction between compounds 7 and 1H-1,2,4-triazol-3-amine 8 occurred smoothly to afford the desired multi-substituted [1,2,4]triazolo[1,5-a]pyrimidine derivatives 9.
To prove the generality of this protocol and expand the substrate scope, various arylaldehydes 1 and 1-(aryl)ethan-1-ones 2 were used, including thiophene, unsubstituted phenyl, and phenyl rings bearing either electron-donating groups (EDGs) (like methyl or methoxy) or electron-withdrawing groups (EWGs) (like chlorine or bromine). As a result, a broad series of triazolopyrimidines 9a-9t was synthesized. The structures of the obtained compounds were fully characterized using IR, H and C NMR spectroscopy, high-resolution mass spectrometry (HRMS), and elemental analysis. Partial assignments of these resonances are given in the Experimental Part.
On accordance with previous literature, α-azidochalcones 7 went through the Michael addition-cyclization in the presence of bi-nucleophilic adducts, followed by removal of a nitrogen and a water, to produce the desirable nitrogen-containing compounds. The similar route took place in present study between α-azidochalcones 7 and 1H-1,2,4-triazol-3-amine 8, confirming by characterization analysis. For instance, the data for compound 9s is representative:
Its IR spectrum showed two absorption peaks at 3368 and 3235 cm for NH group. The HRMS spectrum of this compound exhibited the molecular ion peak in positive mode ([M + H]) at m/z 356.0459, which was 46.0167 mass units less than the sum of the masses of corresponding α-azidochalcones 7 and 1H-1,2,4-triazol-3-amine 8 in the ratio of 1 : 1, which is consistent with the loss of one molecule of nitrogen (N, -28.0061) and one molecule of water (HO, -18.0106) during the reaction, thereby confirming the proposed pathway. The H NMR spectrum of triazolopyrimidine 9s displayed a broad signal around δ 3.79 ppm for NH group and a sharp singlet band at δ 8.33 ppm assigned to the proton on the C-2 position of [1,2,4]triazolo[1,5-a]pyrimidine backbone. Moreover, four doublet peaks with the appropriate chemical shifts and coupling constants were appeared for eight aromatic H-atoms of the two aryl rings in the range of 7.55 to 7.88 of spectrum. The C NMR spectrum of this compound exhibited 13 distinct peaks in the suitable chemical shifts, which was in complete agreement with the desirable molecular symmetry of the proposed structure. The spectral data for other newly synthesized compounds 9 consistently supported their structures.
Our biological studies on a novel series of 5,7-diaryl-[1,2,4]triazolo[1,5-a]pyrimidin-6-amines 9 began with evaluating their in vitro inhibitory potential against Saccharomyces cerevisiae α-glucosidase. In this study, acarbose was used as standard reference drug. Our primary results revealed that all the triazolopyrimidines 9a-9t showed good to excellent activities (with IC values ranging from 24.32 ± 0.18 µM to 151.32 ± 0.56 µM in comparison with the reference drug (with an IC value of 750.08 ± 0.48 µM). The results are summarized in Table 1. To provide a comprehensive structure and activity relationship (SAR), the triazolopyrimidines were divided into four groups considering the substituents on the Ar: (1) unsubstituted; (2) 4-Me substituted; (3) 4-OMe substituted; (4) 4-Cl substituted phenyl ring. To obtain an optimized α-glucosidase inhibitory potency, the substituents on the Ar ring was changed in each category.
In the first series, compound 9a with an unsubstituted phenyl ring as Ar possessed a moderate inhibitory activity (IC = 97.02 ± 0.83 µM). Introducing either an EDG (OMe) or an EWG (like Cl or Br) at the C-4 position of Ar enhanced the inhibitory potency of this pharmacophore (as seen in compounds 9b (IC = 70.85 ± 0.27 µM), 9c (IC = 56.21 ± 0.68 µM), and 9f (IC = 65.09 ± 0.38 µM)). Despite the beneficial presence of 4-Cl on Ar, moving this atom to the C-3 or C-2 positions of this phenyl ring caused a detrimental effect, which might be related with steric hindrance. Comparing compound 9 g with compound 9a, replacing the phenyl ring with thiophene resulted in a significant reduction in inhibitory potency (IC = 139.78 ± 1.55 µM).
Across the second series, there was almost similar trend about the role of substituents on the Ar. Comparing compounds in this group with their corresponding analogue in the first series revealed that introducing 4-Me on Ar did not lead to any improvement on the α-glucosidase inhibitory potency. This can be mechanistically explained by the methyl group's pure electron-donating nature and its potential to introduce minor steric clashes. The active site pocket accommodating Ar seemed to be electrostatically tuned such that increased electron density from EDGs like methyl could not favor binding. Furthermore, the small but significant steric bulk of the methyl group might prevent the aryl ring from achieving an optimal deep fit within its sub-pocket.
The presence of a 4-OMe group on this ring in group 3 enhanced the triazolopyrimidine backbone potential to inhibit the enzymatic activity. For example, compound 9 m with an IC value of 72.36 ± 0.98 µM demonstrated superior potency than that of compounds 9a and 9 h with IC values of 97.02 ± 0.83 µM and 94.52 ± 0.72 µM, respectively. Moreover, introducing an EDG (4-OMe) or an EWG (4-Cl) on Ar led to further improvement on inhibition potency, as compounds 9n and 9o possessed IC values of 52.66 ± 0.30 µM and 37.18 ± 0.23 µM, respectively. A general comparison of the compounds in this group with their analogues in the previous groups revealed that the presence of OMe improved the inhibitory activities. Its oxygen atom could serve as a hydrogen bond acceptor in the enzyme's active site. This potential for a favorable polar interaction could offset the negative electronic effects, leading to an overall enhancement in the binding affinity compared to the methyl-substituted or unsubstituted analogues.
Following the modification of substituents on the Ar, introducing 4-Cl on this phenyl ring increased the inhibitory potency of triazolopyrimidines in group 4, as compound 9q showed better inhibition potency with IC value of 58.11 ± 0.34 µM in comparison with its analogues (compounds 9a (IC = 97.02 ± 0.83 µM), 9 h (IC = 94.52 ± 0.72 µM), and 9 m (IC = 72.36 ± 0.98 µM)). Same as previous series, the presence of 4-OMe and 4-Cl helped to enhance the triazolopyrimidine scaffold to inhibit α-glucosidase activity. As can be seen, compounds 9r and 9s exhibited the IC values of 41.48 ± 0.56 µM and 24.32 ± 0.18 µM, respectively.
In conclusion, the SAR analysis revealed that inhibitory potency was highly influenced by the electronic nature and precise position of substituents on both aryl rings. The presence of OMe as an EDG or Cl and Br as an EWG at the para position of both Ar and Ar improved the inhibitory activity in comparison with their unsubstituted analogues, suggesting that the binding pocket accommodating these phenyl rings was likely sensitive to the electron density of the ligand. This can be rationalized by the fact that EWGs helped to improve the activity by increasing the compound's electrophilicity and forming favorable halogen-bonding interactions with electron-rich regions of the enzyme (such as carbonyl backbones), while EDGs might enhance van der Waals interactions. The presence of chlorine atoms on both phenyl rings in compound 9s exerted the remarkable synergistic effect to maximize hydrophobic interactions and potentially form dual halogen bonds, leading to a significant increase in binding affinity and inhibitory potency.
Moreover, the presence of chlorine atom at the C-2 or C-3 positions had destructive enzymatic effects. It can be contributed with the binding pocket for the Ar ring is sterically constrained. The presence of substituents at meta or ortho positions caused a torsional twist or increased steric bulk, preventing this ring from to locate in the optimal coplanar conformation for binding and disrupting key interactions with the enzyme's active site.
Furthermore, the substitution of Ar with a thiophene moiety consistently yielded the least potent derivatives, as reflected by the highest IC values. This could be attributed to differences in electronic distribution, bond angles, and reduced planarity compared to the phenyl ring, which may disrupt essential π-π stacking or hydrophobic interactions within the binding site. Therefore, this heterocycle was not a suitable bioisostere for the phenyl ring to inhibit α-glucosidase.
Overall, 5,7-bis(4-chlorophenyl)-[1,2,4]triazolo[1,5-a]pyrimidin-6-amine 9s (with IC value of 24.32 ± 0.18 µM) emerged as the most potent inhibitory activity against α-glucosidase, which was 30.84 times superior than that of acarbose (with IC value of 750.08 ± 0.48 µM). Therefore, it was selected for further biological evaluations, including kinetic study to find its inhibition mode, α-amylase inhibition potency to investigate its selectivity, as well as CD and fluorescence spectroscopy measurements to study its potential to alter the secondary and tertiary α-glucosidase's structure.
The α-glucosidase activity in the presence of various concentrations of substrate p-nitrophenyl α-D-glucopyranoside (1-4 mM) along with various concentrations of triazolopyrimidine 9s (0, 6, 12, 24, and 48 µM). Subsequently, the Lineweaver-Burk plot (Fig. 3A) and the plot between the reciprocal of the substrate concentration (1/[S]) and reciprocal of enzyme rate (1/V) over various inhibitor concentrations (Fig. 3B) were outlined to determine the inhibition mode and the Michaelis-Menten constant (K) value, respectively. As shown in Fig. 3A, increasing the concentration of triazolopyrimidine 9s led to a gradual increase in K, while V remained unchanged, indicating a competitive mode of inhibition. This suggests that compound 9s competes with the substrate for binding to the active site of α-glucosidase. Furthermore, Fig. 3B shows that the inhibition constant (K) of triazolopyrimidine 9s is 24 µM.
Selective inhibition of α-glucosidase, without affecting α-amylase activity, is of significant therapeutic value in discovery and development of antidiabetic agents to treat T2DM. α-Glucosidase catalyzes the final step of carbohydrate digestion by hydrolyzing the oligosaccharides and disaccharides into absorbable glucose units at the intestinal brush border. However, α-amylase catalyzes the initial step of polysaccharide breakdown like starch in the upper gastrointestinal tract. Although inhibition of α-amylase may contribute to reduced glucose absorption, it often leads to gastrointestinal side effects, including bloating, cramping, and diarrhea, due to the accumulation of undigested carbohydrates in the colon and their subsequent fermentation by gut microbiota. Therefore, selective α-glucosidase inhibitors are generally more favorable for effective glycemic control and long-term T2DM therapy in clinical use.
The in vitro α-amylase inhibitory activity of the targeted [1,2,4]triazolo[1,5-a]pyrimidin-6-amines 9a-9t was evaluated and the results are summarized in Table 2. The assay was conducted at a concentration of 100 µM to provide a direct comparison with the standard drug acarbose, which exhibited 50.10 ± 0.5% inhibition under the same conditions.
Notably, the majority of the synthesized triazolopyrimidines demonstrated negligible inhibition against α-amylase. Moreover, seven derivatives (9a, 9d, 9e, 9 g, 9 L, 9p) showed no detectable activity (0% inhibition), while a further eight compounds (9b, 9 h, 9i, 9 m, 9q, 9r, 9t, 9n) exhibited only minimal inhibition, ranging from 2.14% to 8.38%, indicating the noticeable contrast to their potent activity against α-glucosidase (IC values ranging from 24.32 to 151.32 µM, Table 1) and confirming their high selectivity for the α-amylase. Among all tested compounds, 9o and 9s -- which were also the most potent α-glucosidase inhibitors -- showed the highest α-amylase inhibition, at 26.74 ± 0.32% and 23.2 ± 0.9%, respectively. However, even for these top performers, the inhibition remained substantially lower than that of acarbose.
In vitro α-amylase inhibitory activity assessment demonstrated that at a concentration of 100 µM [1,2,4], triazolo[1,5-a]pyrimidin-6-amine 9s and acarbose inhibited α-amylase activity up to 23.2 ± 0.9% and 50.1 ± 0.5%, respectively. These results suggest that, despite the excellent α-glucosidase inhibitory potency of compound 9s, further structural modifications on this backbone may be necessary to enhance higher selectivity over α-amylase and optimize its potency.
The difference between the absorption of right and left circular polarized light is measured to identify the chiral environment around aminoacids residues and study the changes in the secondary enzyme structures induced by inhibitors. In this study, the CD spectrum was recorded in the far UV region ranged from 190 nm to 240 nm, and then analyzed using CDNN software to compare the conformations of α-glucosidase-triazolo[1,5-a]pyrimidine 9s complex with native enzyme. There are several conformations, including α-helix, β-sheet, β-turn, and random coils. As presented in Table 3, CD spectroscopy analysis revealed a significant conformational change in α-glucosidase upon interaction with compound 9s. The secondary structure transitioned from a predominantly disordered state (60% random coil, 0% α-helix) to a highly ordered structure (41.2% α-helix, 51.4% β-turn). The CD spectroscopy results indicate that triazolopyrimidine 9s stabilizes the secondary structure of α-glucosidase, likely by inducing α-helical folding and reducing the enzyme's conformational flexibility -- structural changes that may contribute to the inhibition of its catalytic activity.
Fluorescence spectroscopy measurements were applied to find better insights into the mechanism through which compounds affect the tertiary structures of enzyme, leading to inhibit its activity. Triazolo[1,5-a]pyrimidine 9s at different concentrations (0, 10, 15, 20, and 25 µM) was added to the 3.0 mL solution containing a fixed amount of α-glucosidase. The steady-state fluorescence emission spectra were recorded at five different temperatures (298 to 338 K) in the range from 300 nm to 500 nm at the excitation wavelength of 280 nm on a Synergy HTX multi-mode reader (Biotek Instruments, Winooski, VT, USA), equipped with a 1.0 cm quartz cell holder. During the folding or unfolding process of the enzyme, exposure and change of hydrophobic patches in any folded/unfolded protein can be determined. All of the mixtures were held for 10 min to equilibrate before measurements. The fluorescence spectra of the buffer containing triazolopyrimidine 9s without the enzyme were subtracted as the background fluorescence.
The α-glucosidase enzyme contains three primary intrinsic fluorophores, named tryptophan, tyrosine, and phenylalanine, located within or near the active site. The fluorescence spectroscopy in this study, as depicted in Fig. 4, showed a shift in the emission maximum and a significant increase in the intensity in concentration-dependent manner (0 to 25 µM) around 340 nm, both of which means triazolopyrimidine 9s may be positioned near the binding locations to make the pivotal interactions which cause conformational changes, mainly altering the local chemical environment around tryptophan residue(s). Such interactions probably induce conformational changes in the tertiary structure of α-glucosidase, subsequently contributing to the enzymatic activity inhibition.
There are various non-covalent forces between ligands and proteins, including hydrogen bonds, van der Waals forces, electrostatic attraction, and hydrophobic interaction. To identify this type in triazolopyrimidne 9s-α-glucosidase complex. To identify this type, thermodynamic parameters should be determined. To this aim, the stability of this complex was evaluated by monitoring the fluorescence intensity 340 nm to 350 nm at five temperatures (298, 308, 318, 328, and 338 k) using the two-state equilibrium model N ↔ U. The denatured fraction (FD) of protein was calculated from Eq. 1, assuming a two-state mechanism for protein denaturation.
In this Eq. 1, Y, Y, and Y are observed absorbance, the values of absorbance characteristics of a fully native and denatured conformation, respectively. After finding F, the apparent equilibrium constant (K) for a reversible denaturation process between native and denatured protein states and the standard Gibbs free energy change (ΔG°) for protein denaturation were calculated using Eqs. 2 and 4, respectively:
T and R are the absolute temperature and the universal gas constant, respectively. The Gibbs free energy (ΔG) is the most valuable standard of protein conformational stability in thermal denaturation. The integrated Gibbs-Helmholtz equation was utilized for measuring changes in the Gibbs energy of a system as a function of temperature as below:
where ΔC is the heat capacity of protein denaturation, the ΔC (11.6 kJ/mol K) of the α-glucosidase denaturation. In thermal denaturation, T is the temperature at which the protein is half-denatured. ΔHº and ΔSº are the standard enthalpy and entropy of denaturation. The standard entropy was calculated from a relation between the standard enthalpy (ΔS) and entropy (ΔH) of denaturation as below:
The results are summarized in Table 4:
Using the magnitude and sign of the thermodynamic parameters, type of interactions could be determined as follow: (1) hydrophobic interactions with ΔH° > 0, ΔS° > 0; (2) van der Waals forces with ΔH° < 0, ΔS° > 0; (3) hydrogen bond and van der Waals interactions with ΔH° < 0, ΔS° < 0; and (4) electrostatic interactions with ΔH° < 0, ΔS° > 0. Therefore, Table 3 revealed that triazolo[1,5-a]pyrimidine 9s formed the hydrogen bond and van der Waals interactions with aminoacids, particularly tryptophan, leading the enzyme to the unfolded state.
In recent years, computational techniques have played key role in accelerating the drug discovery and development process. Among them, machine learning (ML) and deep learning (DL) models have been widely used for virtual screening and molecular property prediction. Notably, language models such as BERT (Bidirectional Encoder Representations from Transformers), which originally developed for natural language processing, have been successfully adapted to interpret chemical SMILES strings as textual sequences. By pre-training BERT on large molecular databases, rich chemical representations can be learned and subsequently fine-tuned on task-specific datasets with limited size. To address the common issue of data scarcity in drug discovery, SMILES augmentation techniques -- generating multiple equivalent SMILES for each compound -- have proven highly effective. Recent studies show that combining BERT-based models with SMILES augmentation significantly enhances the prediction of biological activity and ADMET properties, offering a powerful framework for rational drug design.
In this study, a data-augmented BERT-based deep learning framework was employed to predict the potency of chemically accessible α-glucosidase inhibitors, which were subsequently synthesized and experimentally validated for inhibitory activity. Pre-trained chemical language models were fine-tuned with SMILES augmentation to predict α-glucosidase inhibitory activity, then molecular docking and dynamics (MD) studies were performed.
A deep learning pipeline was developed to evaluate ten pre-trained transformer models obtained from the Hugging Face repository, including chemical language models such as ChemBERT and the PC10M-series. Each model was fine-tuned using SMILES representations of over 1500 known α-glucosidase inhibitors. The dataset was partitioned for internal 4-fold cross-validation (75% training, 25% testing, repeated across 10 iterations) as well as external validation (in which entire publications were held out as test sets over 30 iterations) in order to minimize data leakage.
A comparison of model performance metrics before and after the application of SMILES augmentation revealed substantial improvements. The augmentation process increased the effective dataset size and contributed to a reduction in overfitting. For instance, under the PC10M-450k model, both overall accuracy (ACC) and Matthews correlation coefficient (MCC) improved notably following augmentation, indicating enhanced generalization capability. These effects are illustrated in Table 5, which compares ACC and MCC values for models trained with and without SMILES augmentation, showing that augmentation significantly improved all metrics.
The PC10M-450k model achieved the highest Matthews Correlation Coefficient (0.75) after augmentation, indicating a well-balanced and robust classification performance for both active and inactive compounds. With an ACC score of 0.90, this fine-tuned model was subsequently employed to predict the inhibitory activity of newly synthesized compounds 9a-9t. As shown in Table 6, the model accurately predicted the activity profiles of these compounds, demonstrating its practical utility in guiding the design and identification of further potential α-glucosidase inhibitors.
Statistical comparison via ANOVA test comparing all models revealed significant differences in accuracy scores (F = 163, p < 0.0001), indicating that the models did not perform equally. Post-hoc Tukey's HSD analysis showed no significant difference in accuracy between PC10M-450k and PC10M-396_250 (p > 0.05), while both significantly outperformed the other models (p < 0.01). Given its superior recall, PC10M-450k was selected for subsequent screening.
A one-way ANOVA test conducted across all transformer-based models yielded a highly significant result (F = 163, p < 0.0001), clearly indicating that model performance in terms of accuracy was not uniform. This statistically confirms that the choice of pre-trained model architecture significantly influences predictive outcomes.
Molecular docking studies were conducted using AutoDock4 (version 4.2.6) to analyze the interactions between the compounds. PDB ID 3A4A was selected for structural analysis due to its high sequence similarity (more than 80%) with α-glucosidase, finding valuable insights into the binding affinities and interaction profiles of the newly synthesized [1,2,4]triazolo[1,5-a]pyrimidin-6-amines 9a-9t. As summarized in Table 6, the docking scores -- showing the predicted binding affinity -- ranged from - 8.14 to -9.46 kcal/mol, revealing the strong binding potential of all triazolopyrimidines with the active site of enzyme. Notably, compound 9s (IC = 24.32 ± 0.18 µM), having the lowest IC value in the enzymatic assay, was also among those with the most favorable docking energies (-9.3 kcal/mol).
As depicted in Fig. 5, the docking analysis of triazolopyrimidine 9s within the α-glucosidase active site revealed a network of stabilizing interactions. This inhibitor occupied a predominantly hydrophobic cleft, with Phe303 engaging in π-π stacking with the triazolo[1,5-a]pyrimidine core. This residue is part of the active site's hydrophobic side chains, which means this interaction is important for securing the ligand, especially with 9s's hydrophobicity. Surrounding hydrophobic residues, including Phe314 (located at the bottom of the active site pocket), Tyr72, and Tyr158 (located at the entrance to the pocket), contributed the additional van der Waals contacts that consolidate the binding pose. A halogen bond was established between the chlorine atom of the chlorophenyl moiety and Arg315 at the charged surface, serving as a key anchoring interaction. Furthermore, His112 and Gln182 are positioned near the terminal aryl chloride, forming polar contacts and shape-complementary interactions which helped in maintaining the ligand's orientation.
Overall, ligand 9s was primarily stabilized by hydrophobic interactions, most notably with Phe303. Its orientation within the binding pocket was further guided by halogen bonds involving the chlorine substituent and residues Gln182 and Arg315. Additionally, His112, a valuable residue responsible for hydrogen bonding with the original ligand in 3A4A, contributed additional interactions that helped secure the ligand in place. Collectively, these interactions ensured a well-fitted and stable accommodation of triazolopyrimidine 9s within the α-glucosidase active site, consistent with its observed inhibitory potency.
Molecular dynamics (MD) simulations play a pivotal role in exploring the inhibitory mechanisms of α-glucosidase. This computational approach provides detailed observation of enzyme-inhibitor interactions, offering insights into both structural flexibility and functional behavior at the atomic scale. Through dynamic modeling, researchers can uncover the binding pathways and identify critical amino acid residues that govern the strength and specificity of ligand attachment. Moreover, MD simulations assess the temporal stability of the enzyme-inhibitor complex using metrics such as Root Mean Square Deviation (RMSD), which helps differentiate between competitive and non-competitive modes of inhibition.
To evaluate the dynamic stability and interaction profile of the α-glucosidase-ligand complexes, 200 ns molecular dynamics (MD) simulations were performed for both acarbose and triazolopyrimidine 9s. The time evolution of RMSD values revealed distinct stability patterns. In the acarbose-bound complex (Fig. 6A), the protein backbone RMSD stabilized around 1.5-2.0 Å after ~ 50 ns, while the ligand RMSD gradually increased to ~ 8-9 Å, indicating significant ligand displacement. In contrast, the 9s-bound complex (Fig. 6B) exhibited slightly higher protein fluctuations (up to ~ 2.8 Å) but maintained a lower ligand RMSD (~ 4-5 Å), suggesting a more stable binding pose. These results revealed that acarbose, despite inducing minimal protein perturbation, underwent considerable reorientation or partial dissociation from the active site. Conversely, compound 9s remained tightly bound, even as the protein experienced moderate conformational changes, indicating a more stable and persistent interaction.
Although the ligand RMSD of 9s reached ~ 5.6 Å, this mainly reflects internal conformational flexibility and pose rearrangements within the open glycosidase pocket. Consistently, MM/GBSA analysis (Fig. 7) over the trajectory yielded a more favorable binding free energy for 9s (⟨ΔG_bind⟩ = -48.5 ± 8.2 kcal·mol⁻¹; range - 67.3 to - 24) compared with acarbose (⟨ΔG_bind⟩ = -33.8 ± 8.0 kcal·mol⁻¹; range - 65.3 to - 13), supporting a stably bound complex despite larger ligand RMSD.
Ligand-centric metrics further illustrated the differences in size and flexibility between compound 9s and acarbose. As shown in Fig. 8, acarbose exhibited a larger average RMSD (~ 2-3 Å) and radius of gyration (~ 5.4 Å) compared to 9s (~ 1 Å RMSD, ~ 4.3 Å rGyr), consistent with its bulky, polyhydroxylated structure. The number of intramolecular hydrogen bonds also differed: acarbose formed 0 to 2 internal hydrogen bonds intermittently, whereas compound 9s showed none, reflecting its simpler scaffold. Correspondingly, acarbose's molecular surface area (~ 480 Ų) and polar surface area (~ 500 Ų) were much greater than those of compound 9s (~ 300 Ų MolSA, ~ 100 Ų PSA). Its solvent-accessible surface area (~ 500 Ų) likewise far exceeded that of compound 9s (~ 200 Ų), indicating that acarbose remained highly solvated and extended throughout the simulation.
In MD terms, acarbose behaved like a flexible polar ligand: its RMSD, PSA, and rGyr showed larger fluctuations -- especially during the first ~ 50 ns -- before reaching equilibrium. This behavior mirrors previous reports where polar ligands exhibit significant early fluctuations in RMSD, PSA, and rGyr prior to stabilization. By contrast, compound 9s's compact metrics and lack of internal hydrogen bonding resembled those of a small, rigid ligand, which typically exhibited minimal fluctuations in radius of gyration or surface area. Overall, acarbose's larger size and polarity led to a more extended and internally hydrogen-bonded conformation that adjusted during the simulation, whereas these metrics suggest that compound 9s behaves as a compact, lipophilic ligand with stable conformational properties throughout the simulation.
Contact frequency analysis (Fig. 9) revealed that the compound 9s-α-glucosidase complex is primarily stabilized by hydrophobic interactions. Aromatic residues such as Tyr72, Tyr158, Phe159, Phe178, and Phe303 formed persistent contacts, supported by Val216, Arg315, Val410, and Arg442. Halogen bonding was highly localized, with Met70, Tyr72, and Gln182 maintaining interactions in ~ 60% of frames. Additional halogen contacts were observed with Asp69 and His280. Hydrogen bonding was less frequent, dominated by Asn414, with minor contributions from Tyr72, Gln279, and Arg315. Water-mediated bridges clustered around Tyr72 and Arg315, and to a lesser extent at Asp352, Gln353, and Leu313. Ionic interactions were negligible, consistent with the ligand's nonpolar character.
Overall, the triazolopyrimidine 9s-α-glucosidase complex was primarily stabilized by hydrophobic interactions, mediated by aromatic and nonpolar residues. These interactions were reinforced by frequent halogen bonds -- most notably with Gln182 -- and a peripheral hydrogen-bonding network centered around Asn414 (Fig. 10). The MD simulation results were consistent with the docking findings: Phe303 maintained persistent hydrophobic contacts, Arg315 interacted through both hydrophobic and water-mediated bridges, and His112 contributed weak or transient contacts. Among all residues, Gln182 emerged as a key polar anchor, forming stable halogen bonds throughout the trajectory.