The reference phenolic compounds used in this study (Table 1) as well as galantamine (reference drug) and scopolamine hydrobromide (S1875-5G) were acquired from Sigma-Aldrich (St. Louis, MO, United States); all compounds had purity values over 95%.

Inhibitory activities of electric eel AChE (Type VI-S, EC 3.1.1.7, Sigma, St. Louis, MO, United States) and horse serum BChE (EC 3.1.1.8, Sigma) were determined using Ellman’s method with slight modifications (Ellman et al., 1961). Briefly, 140 µL of 0.1 mM sodium phosphate buffer (pH 8), 20 µL of 5,5′-dithio-bis (2-nitrobenzoic) acid (DTNB, 0.4 mM, Sigma), 20 µL of the enzyme (either AChE or BChE, 5.32 × 10⁻³ U), and 20 µL of the test sample were added to the reaction mixture and incubated for 15 min at 25 °C. Next, approximately 10 µL of acetylthiocholine iodide (0.4 mM) or butyrylthiocholine chloride (0.4 mM) was added to the incubated sample as the substrate reacting with DTNB to form 5-thio-2-nitrobenzoate anion, which resulted in a yellow-colored complex. The absorbance was then measured at 417 nm using a 96-well ELISA microplate reader (VersaMax, Molecular Devices, San Jose, CA, United States). Here, galantamine hydrobromide was used as the reference drug.

The measurements and calculations were processed using Softmax PRO 4.3.2. LS software. The percentage inhibitions of AChE and BChE were determined by comparing the reaction rates of the test samples with those of blank samples. The extent of the enzymatic reaction was calculated using the equation I% = [(C–T)/C] × 100, where I% is the activity of the enzyme given as percentage inhibition, C is the absorbance of the control solvent (blank) in the presence of the enzyme, and T is the absorbance of the test sample or positive control/reference inhibitor (i.e., galantamine) in the solvent in the presence of the enzyme. This value expresses the effect of the test sample or positive control/reference inhibitor on AChE or BChE enzymatic activity in terms of the percentage of remaining activity in the presence of the test sample or positive control. The data were expressed as average inhibition ± standard deviation (SD), and the results were taken from at least three independent experiments performed in triplicate.

Male Swiss albino mice (20–25 g) were used in the experiments. The animal study was conducted in adherence to the Animal Research: Reporting of In Vivo Experiments guidelines, and all experimental procedures were approved by the Local Animal Ethics Committee of Gazi University (approval number: G.Ü.ET-19.057; 4 October 2019). The animals were housed under controlled laboratory conditions (temperature: 21–24 °C; 12-h light/12-h dark cycles) for at least 1 week prior to the start of the experiments. During testing, the background noise was minimized, and the experiment room was consistently illuminated according to standard protocols. Throughout the study, the animals were provided a standard pellet feed and water ad libitum. Following the 1-week acclimatization period, the mice were randomly divided into eight groups, with eight mice per group. This group size was based on previous studies employing similar behavioral models, where a minimum of eight animals per group was reported (Orhan and Aslan, 2009; Getova and Mihaylova, 2013; Barkur and Bairy, 2015; Budzynska et al., 2015). The eight groups were as follows: A: 0.9% NaCl (sham), B: scopolamine hydrobromide (control), C: oleuropein, D: 3-hydroxytyrosol, E: gallic acid, F: rosmarinic acid, G: EGCG (phenolic compound), and H: galantamine (reference). During the acquisition phase of the passive avoidance test, each mouse was first placed in a light compartment; the guillotine door was opened 10 s later, and the step-through latency of the animal to a dark compartment was recorded. Upon entry, the door to the compartment was closed automatically, and a foot shock of 0.1 mA/10 g of body weight was delivered for 3 s (Orhan and Aslan, 2009). Immediately after the acquisition phase, the animals received their respective treatments: group A (sham) received only 0.9% NaCl; groups C–G were administered oleuropein, 3-hydroxytyrosol, gallic acid, rosmarinic acid, and EGCG, respectively, where each compound was prepared in 0.9% NaCl at a concentration of 20 mg/kg and delivered via intraperitoneal (i.p.) injection; group H received the reference drug galantamine at a dose of 10 mg/kg via subcutaneous (s.c.) injection (De Bruin and Pouzet, 2006). Thirty minutes after treatment, groups B–H received scopolamine hydrobromide (1 mg/kg, dissolved in 0.9% NaCl) via i.p. injection to induce amnesia. Twenty-four hours after scopolamine administration, the mice were reintroduced to the light compartment, and their step-through latencies to the dark compartment were measured. Mice that did not enter the dark compartment within 9 min (cutoff time) were considered to have retained a passive avoidance memory. The behavioral assessments were conducted by a single trained investigator who was blinded to the treatment procedures during the testing phase. Statistical analyses were performed using one-way ANOVA with GraphPad Prism version 6.01.

The molecular docking studies were carried out using induced-fit docking (IFD) protocols implemented in the Schrödinger Small-Molecule Drug Discovery Suite (Schrödinger, LLC, New York, NY, United States) (Zou et al., 2014). The compounds constructed using the builder panel in Maestro were subjected to ligand preparation with LigPrep (Schrödinger Release 2023-1: LigPrep, Schrödinger) under the default conditions. The x-ray crystal structures of hAChE (PDB: 4EY7) and hBChE (PDB: 4TPK) were retrieved from the Protein Data Bank (Cheung et al., 2012; Brus et al., 2014). The proteins were prepared using the Protein Preparation Wizard tool, where hydrogen atoms were added first, followed by the assignment of all atom charges and atom types. Finally, energy minimization and refinement of the structures were performed to 0.3 Å of root mean-squared deviation (RMSD) by applying the OPLS3e force field. The receptor grid was generated by defining the centroid of the cocrystallized ligand as the box center, and the option “Dock ligands similar in size to Workspace ligand” was selected to automatically define the box size. A van der Waals (vdW) radius scaling factor of 1.00 and a partial charge cutoff of 0.25 were also used. The OPLS3e force field was applied throughout the grid generation process. The compounds prepared using LigPrep were docked onto hAChE and hBChE using the IFD protocol, which considers flexibility for both compounds and receptors. For hAChE, the residues D74, W86, Y124, Y133, S203, W286, F295, F297, Y337, F338, and H447 lining the binding sites of AChE were retained as flexible. The initial docking protocol was set to employ a 0.50 vdW radius scaling factor, and the resulting top-20 orientations of each compound were taken. An extra-precision algorithm was employed to redock the compounds with the low-energy refined structures generated by the Prime molecular mechanics/generalized Born surface area (MM/GBSA) method.

The ligand-residue free energies of the binding calculations were obtained using the MM/GBSA method in the Schrödinger suite. The equation used to calculate the binding energy is as follows: Δ G bind = Δ E MM + Δ G solv + Δ G SA , where ΔE_MM is the difference in minimized energies given by Δ E MM = E complex   ‐   E ligand   ‐   E receptor

The difference in GBSA solvation energies between the complex and the sum of ligands and proteins is denoted by ΔGsolv. Furthermore, ΔGSA is the difference in the surface area energies between the complex and the sum of proteins and ligands. A script was used to calculate the average MM/GBSA binding energy, which also generates the Coulomb energy (Coulomb), covalent binding energy (Covalent), hydrogen-bonding energy (H-bond), lipophilic energy (Lipo), generalized Born electrostatic solvation energy (Solv_GB), and vdW energy.

The selected compounds were comprehensively evaluated for their physicochemical properties, pharmacokinetic profiles, drug-likeness parameters, and medicinal chemistry characteristics through the SwissADME computational platform developed and maintained by the Molecular Modeling Group at the Swiss Institute of Bioinformatics (Daina et al., 2017). This web-accessible resource integrates numerous validated predictive algorithms to generate detailed absorption, distribution, metabolism, and excretion (ADME) profiles for the candidate molecules. The topological polar surface area (TPSA) was assessed by the fragment-based methodology established by Ertl et al. (2000) that allows efficient and precise estimates of the molecular polar surface area, which is a parameter widely utilized in predicting the transport characteristics of drug candidates. The lipophilicity assessment incorporated multiple predictive models for the octanol–water partition coefficient (log P_o/w) by combining physics-driven approaches such as iLOGP, which reduces the overfitting tendency, with data-centered models trained on extensive experimental measurements. To enhance the reliability of prediction, a consensus log P value was calculated by integrating the outputs from multiple individual predictive models.

The toxicokinetic evaluation employed SwissADME’s integrated screening systems to identify structural motifs associated with pan-assay interference compounds (PAINS), which are known to generate false-positive signals in biological screening assays. Furthermore, Brenk filters were applied to identify potentially hazardous or chemically reactive functional groups that could compromise the compound safety or stability profile. The pharmacokinetic parameters, including substrate affinity for P-glycoprotein (P-gp) and inhibitory potential against cytochrome P450 (CYP450) enzyme isoforms, were predicted using support-vector-machine-based binary classification models embedded within the SwissADME framework. These predictive models were validated through 10-fold cross-validation protocols and independent test datasets and can be used to assess a compound’s probability of interacting with transport proteins and metabolic enzymes based on molecular descriptors and structural similarity to established substrates or inhibitors.

The skin permeability coefficient (log Kp) was calculated using the linear regression model established by Potts and Guy (1992), which correlates transdermal permeability with the molecular dimensions and lipophilicity parameters; more negative log Kp values are indicative of reduced capacity for passive diffusion across the epidermal layers. Additionally, the gastrointestinal (GI) absorption efficiencies and blood–brain barrier (BBB) penetration capabilities were evaluated using the BOILED-Egg classification model, which categorizes the compounds according to their predicted passive absorption and central nervous system access based on the apparent polarity (TPSA) and lipophilicity (WLOGP) values (Daina and Zoete, 2016). Within this visualization framework, compounds predicted to exhibit high GI absorption are represented in the “white” region, while those likely to traverse the BBB appear in the “yellow” (yolk) region. The model further indicates the P-gp substrate potential through distinct color-coded markers.

As shown in Table 1, all compounds from various chemical groups were tested against 10 mM of the ChEs, of which 12 compounds were able to inhibit AChE, while 8 compounds achieved BChE inhibition over 50%. Among these, gallic acid, rosmarinic acid, 3-hydroxytyrosol, trans-resveratrol, EGCG, quercetin, and fisetin possessed the ability to inhibit both ChEs. The most active inhibitor against AChE and BChE was found to be quercetin (IC₅₀: 1.22 ± 0.79 mM and 2.51 ± 0.04 mM, respectively). In our screening, the isoflavone (e.g., daidzein, genistein, and glycitein) and methoxyflavone (e.g., tricin, tangeretin, and diosmetin) derivatives as well as flavone O- and C-glycosides (e.g., isoquercitrin, orientin, and vitexin) were either completely inactive or exerted low inhibition capacities against both ChEs. Amentoflavone, myricetin, ellagic acid, and luteolin showed no inhibition toward the ChEs. However, caffeic acid (IC₅₀: 3.51 ± 0.62 mM), apigenin (IC₅₀: 3.52 ± 0.08 mM), and taxifolin (IC₅₀: 7.18 ± 2.05 mM) were able to inhibit only AChE. Although several of the phenolic compounds were found to be promising ChE inhibitors, it is obvious that their inhibitory effects are lower than those of the reference drug galantamine.

To determine the antiamnesic activities of the compounds, we selected oleuropein, rosmarinic acid, gallic acid, EGCG, and 3-hydroxytyrosol for their dual ChE inhibition owing to their adequate quantities and conducted further evaluations using the passive avoidance test in mice as the prominent in vivo experimental model for memory deficit (Hornedo-Ortega et al., 2018). Here again, the animals were divided into eight groups (n = 8 per group) as follows: A: 0.9% NaCl (sham), B: scopolamine (control), C: oleuropein, D: 3-hydroxytyrosol, E: gallic acid, F: rosmarinic acid, G: EGCG, and H: galantamine (reference). During the acquisition period, the mice were first placed in the light compartment, followed by opening of the guillotine door 10 s later, and measurement of their transition times to the dark compartment using a digital timer (Table 2). After entering the dark compartment, the guillotine door was closed, and the mouse was retained on its feet for 3 s. Then, an electric current of 0.1 mA/10 g of body weight was applied throughout. Immediately after the learning period, only 0.9% NaCl was given to group A (sham) while groups B–G were administered the respective drugs through the i.p. route; to the animals in the H group, the reference drug galantamine was applied via s.c. injection at a dose of 10 mg/kg (Jang et al., 2008). After 30 min, scopolamine (1 mg/kg) was dissolved in 0.9% NaCl and administered by i.p. injection to groups B–G to induce amnesia. Twenty-four hours after the application, the mice were placed back in the light compartment, and their latency times were measured after opening the guillotine door. It was concluded that the mice that did not enter the dark compartment within 9 min (cutoff time) remembered the passive avoidance action. Our findings indicate that the aforementioned five phenolic compounds are effective in mice, with 3-hydroxytyrosol displaying the strongest antiamnesic effect.

The molecular interactions of the selected compounds that displayed inhibitory effects against AChE were determined through the residues at its active site using the Glide module in Schrödinger suite. As seen from the results, the compounds reached the catalytic triad subunit of the enzyme active site and interacted mainly with the oxyanion region comprising residues (Figure 1). Relevantly, rosmarinic acid localized to the active site of AChE by occupying the oxyanion and PAS sites. The presence of hydrogen bonds was observed between the hydroxyl groups on the phenyl ring close to the carboxylic acid group and the residues G120, Y133, and E202. The carboxylic acid group interacted with Y124, a member of the PAS site, via hydrogen bonding. The other dihydroxyphenyl ring in the compound structure formed a π–π interaction with Y341 belonging to the PAS region. The binding energy for rosmarinic acid was calculated as −11.26 kcal/mol. Next, 3-hydroxytyrosol localized to the enzyme active site by interacting with the oxyanion site residues; the hydroxyl group at the third position of the phenyl ring was hydrogen bonded with Y133 and G120, while the hydroxyl group at the fourth position was hydrogen bonded with E202 and the hydroxyl group attached to the ethyl group was hydrogen bonded with S125 at the entrance of the active site. The binding energy for 3-hydroxytyrosol was calculated as −6.48 kcal/mol. As gallic acid is a small molecule, it penetrated the enzyme active site deeply and was located near the residues forming the catalytic triad; the docked complex was stabilized by hydrogen bonds formed between the hydroxyl groups and the residues G120, Y133, and E202. The AChE binding energy for gallic acid was calculated as −5.92 kcal/mol. Quercetin is oriented to the catalytic site of the enzyme with the coumarin ring; because of this orientation, hydrogen bonding and π–π interaction occurred between the coumarin ring and the catalytic triad members S203 and H447, respectively. Other π–π interactions stabilizing quercetin at the active site occurred between the coumarin ring and F338 belonging to the oxyanion site. In addition, the hydroxyl group at the third position of the phenyl ring attached to the coumarin ring interacted with the side chain of F295 in the acyl binding site via hydrogen bonding. The AChE binding energy for quercetin was calculated as −12.24 kcal/mol. While occupying the oxyanion site, oleuropein also interacts with the residues of the PAS site; the hydroxyl group at the fourth position on the tetrahydropyran ring in the structure interacted with the oxyanion site member W86 side chain via hydrogen bonding. The other interactions detected at the PAS site were the hydrogen bond between the carbonyl group and Y124 as well as the π–π interaction between the dihydroxyphenyl ring and W286. The binding energy for oleuropein was calculated as −11.58 kcal/mol. EGCG formed hydrogen bonds with the hydroxyl groups of the benzopyran ring E202 and the side chain of the catalytic triad member H447, and a π–π interaction was observed between the benzopyran ring and H447. The hydroxyl groups of the trihydroxyphenyl ring directly linked to the benzopyran ring formed hydrogen bonds with the PAS site member D74, whereas the gallate moiety showed a π–π interaction with Y341. The AChE binding energy for EGCG was calculated as −12.95 kcal/mol. The conformation of resveratrol was stabilized through hydrogen bonding with the side chains of F295 and H447 as well as π–π interactions with W86 and Y341. The binding energy for resveratrol was calculated as −9.14 kcal/mol. The binding mode of fisetin in the hAChE active site indicated that it formed π–π contacts with both the catalytic triad member H447 and oxyanion hole residue F338 through its chromen-4-one ring. The hydroxyl group in the seventh position on the same ring formed hydrogen bonds with S203 in the catalytic triad and G122 in the PAS region, while the hydroxyl group of the phenyl ring interacted with F295 in the acyl binding region via hydrogen bonding. The binding energy of fisetin at the AChE active site was calculated as −11.33 kcal/mol.

In the crystal structure of human AChE complexed with donepezil (PDB: 4EY7), we identified several key interactions between the ligand and residues lining the active site gorge. Notably, aromatic residues such as W86, Y124, W286, Y337, F338, Y341, and H447 formed hydrophobic and π–π contacts to stabilize the ligand within the gorge. Among these, W86, W286, and Y341 had some of the largest contact surface areas, effectively anchoring the ligand. In addition, a hydrogen bond was formed between the backbone or side chain of F295 and a polar group on donepezil, reinforcing its orientation near the catalytic center. Polar and charged residues like D74 and R296 are also involved in polar or ionic interactions, further enhancing the binding specificity. The compounds analyzed in the molecular docking simulations exhibited similar key interactions with the active-site residues and reflected comparable binding modes to that of the cocrystallized ligand.

The interactions of the selected phenolic compounds that displayed inhibitory effects against BChE were investigated through the amino acids constituting the active site of the enzyme. Based on our outcomes from the molecular docking experiments, these compounds were patterned to reach the catalytic triad subunit of the active site and interacted predominantly with the residues belonging to the oxyanion hole (Figure 2). The conformation for rosmarinic acid at the BChE active site showed an interaction between the hydroxyl groups on the terminal phenyl ring and L286, a member of the acyl binding site, via hydrogen bonds; we also observed interactions between the hydroxyl groups on the other phenyl ring located close to the carboxylic acid group and A328 and Y440 via hydrogen bonds. In addition, the terminal phenyl ring showed a π–π interaction with F329. The BChE binding energy for rosmarinic acid was calculated as −10.39 kcal/mol. Next, 3-hydroxytyrosol was positioned in a region between the oxyanion site and the catalytic triad. The hydroxyl group at the third position of the phenyl ring formed a hydrogen bond with H438, a member of the catalytic triad, while the other hydroxyl group on the ring interacted with G78 and Y440 via hydrogen bonding. The ring interacted with W82, a member of the oxyanion site, via π–π interactions. The binding energy for 3-hydroxytyrosol was calculated as −6.14 kcal/mol. Gallic acid formed π–π interactions between its phenyl ring and W82 as well as a hydrogen bond with the hydroxyl group in the third position on the ring and H438. The docked complex was additionally stabilized by water-bridged hydrogen bonds with G117, E197, and S198 via its carboxylic acid group. The BChE binding energy for gallic acid was calculated as −6.03 kcal/mol. Quercetin formed a water-mediated hydrogen bond network between the hydroxyl group on the coumarin ring and G117 as well as its catalytic triad member S198. In addition to π–π interactions between the coumarin ring and F329 at the oxyanion site, the hydroxyl groups of the phenyl ring interacted with H438 and Y440 via hydrogen bonding. The BChE binding energy for quercetin was calculated as −8.60 kcal/mol. Oleuropein approached the catalytic triplet of the enzyme active site by forming direct hydrogen bonds between the hydroxyl groups on the tetrahydropyran ring and H438 and Y440. The compound also occupied the PAS site with its u-shaped binding mode and attached to the oxyanion site through a π–π interaction between the terminal dihydroxyphenyl ring and F329. The BChE binding energy for oleuropein was calculated as −11.41 kcal/mol. EGCG formed hydrogen bonds between L286 and H438, with the trihydroxyphenyl ring directly attached to the benzopyran ring, as well as π–π interaction with F329. The terminal phenyl ring formed a π–π interaction with W82 as well as a water-mediated hydrogen bond with E197. Another interaction stabilizing the enzyme–inhibitor complex is the hydrogen bond between the hydroxyl group attached to the terminal phenyl ring and the side chain of H438. The BChE binding energy for EGCG was calculated as −13.52 kcal/mol. Resveratrol localized to the BChE active site by interacting with the oxyanion and acyl binding sites; the hydroxyphenyl ring showed strong π–π interactions with amino acids W231 and F329 as well as formed hydrogen bonds with L286 via the hydroxyl group. Resveratrol formed π–π interactions with W82 through the dihydroxyphenyl ring as well as water-bridged hydrogen bonds with S79 and D70 in the PAS through the hydroxyl group. The binding energy for resveratrol was calculated as −7.01 kcal/mol. Fisetin was accommodated in the BChE active site primarily through hydrogen bonding; the carbonyl oxygen of the chromen-4-one ring interacted with E197 and the catalytic triad member S198 via a water-mediated hydrogen bond network, while the hydroxyl group in the seventh position on the same ring formed a hydrogen bond with A328. Additionally, the hydroxyl group at the fourth position on the phenyl ring, which is attached to chromen-4-one, was anchored at the acyl binding site to form a hydrogen bond with S287. The binding energy for the fisetin–BChE complex was calculated as −8.18 kcal/mol. The cocrystallized ligand (3F9, PDB: 4TPK) was lodged deep within the active site gorge through π–π and π–cation interactions with F329, W231, and Y332; hydrogen bonds were also formed with D70 and E197. The interactions identified with the coligand were also observed for the tested compounds, along with additional contacts that further stabilized the binding.

To further validate the docking results, MM/GBSA binding free energy calculations were performed for all docked complexes using Prime from the Schrödinger suite. The calculated ΔG_bind values generally supported the docking-derived rankings. For AChE, donepezil (cocrystallized ligand, PDB: 4EY7) displayed the most favorable binding free energy of −103.49 kcal/mol. Among the phenolics, EGCG (−95.42 kcal/mol), fisetin (−59.01 kcal/mol), and quercetin (−56.49 kcal/mol) exhibited highly favorable binding energies, whereas gallic acid (−3.50 kcal/mol) and 3-hydroxytyrosol (−17.96 kcal/mol) showed comparatively weaker stabilization (Table 3). For BChE, the cocrystallized ligand 3FP (PDB: 4TPK) had a ΔG_bind value of −86.11 kcal/mol, similar to oleuropein (−83.89 kcal/mol) and EGCG (−75.68 kcal/mol); in contrast, gallic acid (−2.51 kcal/mol) and 3-hydroxytyrosol (−27.68 kcal/mol) exhibited notably weaker binding energies (Table 4). These data validate the docking results and quantitatively position the phenolic compounds relative to donepezil in AChE and with respect to the reference ligand in BChE.

The predictions revealed notable differences in the pharmacokinetic profiles of the five phenolic compounds analyzed (Figures 3, 4). Oleuropein exhibited suboptimal drug-likeness mainly owing to its high molecular weight (540.52 g/mol) and elevated TPSA (161.21 Å²), both of which exceeded the thresholds associated with good oral bioavailability; although its consensus log P value (2.41) indicated moderate lipophilicity, oleuropein was predicted to have low GI absorption with limited brain penetration and to be a substrate of P-gp, suggesting restricted cellular uptake. From the perspective of a structure–activity relationship (SAR), the absorption differences observed across the series can be rationalized by their physicochemical parameters. Oleuropein’s bulky glycosylated scaffold increases the polarity and hydrogen-bonding capacity, which together with its high molecular weight and flexibility reduces the passive permeability severely; in contrast, smaller phenolics like 3-hydroxytyrosol and gallic acid combine low TPSAs with limited rotatable bonds to enable efficient intestinal uptake. Rosmarinic acid represents an intermediate case, where favorable absorption is supported by its moderate lipophilicity despite the relatively high polarity. However, EGCG mirrors oleuropein in exhibiting excessive polarity and multiple hydrogen-bond donors, which correlate with its poor predicted absorption.

The SAR-based interpretations are aligned with the findings of the BOILED-Egg classification, where highly polar compounds cluster outside the white region of high GI absorption. In contrast, 3-hydroxytyrosol displayed favorable characteristics with a low molecular weight (140.14 g/mol), optimal TPSA (60.69 Å²), and balanced hydrophilic–lipophilic properties (log P: 0.60); it was predicted to be efficiently absorbed in the GI tract without crossing the BBB or interacting with P-gp. However, some potential CYP3A4 inhibition is suggested, indicating a possible risk of metabolic interactions. As the most hydrophilic compound in the series (log P: 0.21; TPSA: 97.99 Å²), gallic acid also shows high predicted GI absorption and no BBB permeability; similar to 3-hydroxytyrosol, it was not identified as a P-gp substrate but was predicted to inhibit CYP3A4. Rosmarinic acid presented intermediate properties with a molecular weight of 345.30 g/mol, TPSA close to the upper permeability limit (136.32 Å²), and moderate lipophilicity (log P: 1.17); despite its size and polarity, it was predicted to be well absorbed via the GI tract but not cross the BBB. No interactions with P-gp or the major CYP enzymes were anticipated. Lastly, EGCG had the highest TPSA (177.14 Å²) and a molecular weight of 394.33 g/mol, alongside a nearly neutral lipophilicity profile (log P: –0.03); these features correlated with poor predicted GI absorption and no BBB penetration. EGCG was also not expected to act as a P-gp substrate or CYP inhibitor.

In terms of water solubility, all compounds displayed favorable profiles ranging from soluble to very soluble according to the log S (ESOL) model; 3-hydroxytyrosol and gallic acid demonstrated the highest solubility with ESOL values of −1.48 and −1.64, respectively, which makes them very soluble. Oleuropein showed the lowest water solubility with an ESOL value of −4.76, making it “moderately soluble” while still being within an acceptable range for pharmaceutical formulations. The skin permeability coefficient (log Kp) values were negative for all compounds and ranged from −6.71 cm/s for 3-hydroxytyrosol to −8.39 cm/s for EGCG, indicating limited passive permeation across the skin. This suggests that topical formulations of these compounds would likely remain localized at the application site without significant systemic absorption.

SAR analyses were performed to evaluate the mutagenic effects of the selected phenolic compounds. In the PASS program, the probabilities are obtained as P_a (probability of being active) and P_i (probability of being inactive) values. When these values are related as “P_a > P_i” with the difference between them being large, the probability for the biological activity in question increases. In the SwissADME target prediction program, the probability for the biological activity in question is given directly as a percentage value. The screening results for the compounds that were active in the ChE inhibition assay as well as passive avoidance test are shown in Tables 3, 4. These compounds exhibited high probabilities of acting as Aβ aggregation inhibitors, which is another treatment strategy for AD. Apolipoprotein A1 (APOA1) is a structural protein from the group of high-density lipoproteins (Table 3); although the relationship between APOA1 and AD is complex and multifaceted, emerging evidence suggests that APOA1 may play a role in modulating the pathogenesis of AD. Given its involvement in lipid metabolism, inflammation, and neuroprotection, APOA1 has been proposed as a potential therapeutic target for AD (Endres, 2021). Therefore, it is notable that the compounds tested in our in silico analysis exhibited high probabilities for enhancing APOA1 expression. The in silico estimations of the compounds indicated the probabilities of common toxic effects like tremors, genotoxicity, vascular toxicity, hematemesis, and color changes to urine (Table 4).

To evaluate the potential metabolic interactions and off-target liabilities, the selected phenolic compounds were analyzed for their predicted inhibitions of the CYP450 enzymes as well as presence of structural motifs flagged by PAINS and Brenk filters. These computational alerts provide early indicators of potential assay interference, promiscuous binding, or chemical reactivity, which could compromise lead optimization or result in misleading biological outcomes. Among the compounds analyzed, 3-hydroxytyrosol and gallic acid were predicted to inhibit CYP3A4, one of the most clinically relevant isoforms involved in drug metabolism. This suggests the potential for drug–drug interactions when coadministered with substrates of this enzyme. In contrast, rosmarinic acid, EGCG, and oleuropein were not predicted to inhibit any of the major CYP enzymes evaluated, indicating a potentially lower risk of CYP-mediated interactions.

Despite the favorable CYP profiles for some of the compounds, all five molecules triggered at least one PAINS alert that was most associated with catechol_A substructures. According to Baell and Holloway (2010), such motifs are frequently linked to nonspecific activities across multiple biological targets, thereby increasing the risk of false-positive results in high-throughput screening assays. These findings underscore the importance of experimental validations to distinguish true pharmacological effects from assay artifacts. Additionally, Brenk alerts identifying chemically reactive or metabolically unstable functional groups were detected in several compounds. Specifically, 3-hydroxytyrosol, gallic acid, and EGCG triggered the catechol alert, while rosmarinic acid exhibited multiple alerts for catechol, hydroquinone, and michael_acceptor_1, indicating greater potentials for chemical reactivity or toxicity. Collectively, while the CYP inhibition profiles suggest a low risk of metabolic interactions for most of the compounds, the presence of PAINS and Brenk alerts highlights potential challenges in lead development. These findings emphasize the need for structural optimization and biological confirmation to ensure selective activity while reducing false-positive outcomes in early-stage drug discovery.

The main limitations of the present study are as follows. Although this study acknowledges the ChE inhibition potentials of several phenolic compounds, the inhibitory effects were generally lower than that of the reference drug galantamine. The in vivo experiments were conducted on scopolamine-induced amnesic mice, which is a widely used model for studying memory impairment that may not fully replicate the complex pathology of AD. Therefore, we recognize that the lack of dual-observer scoring is a limitation. The lack of detailed pharmacokinetic information and optimal dosing schedules for each molecule should be considered a limitation and warrants further investigations combining behavioral and pharmacokinetic assessments. The in silico toxicity assessment provides valuable predictive insights. However, further experimental validations are required to confirm the safety profiles of the active compounds. While the docking results offer valuable insights into the probable binding sites, a definitive characterization of the inhibition mechanism requires further investigation. We believe that a detailed mechanistic study, including kinetic analyses using methods like the Lineweaver–Burk plots, is a logical and important next step. Such future research would provide a more in-depth understanding of the precise interactions between these phenolic compounds and their target enzymes, further solidifying their therapeutic potentials.