Synthesis, QSAR analysis and molecular docking study of a new series of 3-hydroxypyridine-4-one derivatives as anti-tyrosinase agents

Abstract

Tyrosinase is a critical rate-limiting enzyme in the melanogenesis pathway. Consequently, its inhibition represents a rational therapeutic strategy for treating skin disorders associated with excessive melanin production. In the present study, a series of novel 3-hydroxypyridine-4-one derivatives (6a-i) were synthesized, and their chemical structures were confirmed using spectroscopic techniques. The inhibitory potency of these compounds against tyrosinase was predicted using quantitative structure–activity relationship (QSAR) analysis. QSAR modeling was conducted on twenty-four previously synthesized 3-hydroxypyridin-4-one derivatives with established anti-tyrosinase activity. The best-performing model was subsequently employed to predict the IC₅₀ values of the newly synthesized compounds. Among the evaluated statistical methods, the Multiple Linear Regression (MLR) model demonstrated the highest accuracy and precision, exhibiting the lowest data dispersion. Furthermore, its predictive performance for pIC₅₀ values was superior, with R² = 0.93 and Q² = 0.81. The MLR results indicated hyperchem descriptors, 2-D functional descriptors, and GETAWAY descriptors as the most influential parameters contributing to model performance. Finally, molecular docking simulations revealed favorable interactions between the new synthesized compounds and the active site of tyrosinase, supporting their potential as effective tyrosinase inhibitors.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13065-025-01709-6.

Introduction

Tyrosinase (E.C. 1.14.18.1) is a multifunctional metalloenzyme containing two copper ions in its active site, which play a crucial role in melanin biosynthesis in living organisms [1, 2]. Tyrosinase enzyme catalyzes the conversion of monophenolic compounds to diphenols, followed by the oxidation of o-diphenol compounds to o-quinones. Tyrosinase is recognized as the rate-limiting enzyme that catalyzes the hydroxylation of L-tyrosine to L-DOPA and 3,4-dihydroxyphenylalanine to dopaquinones, processes that can lead to abnormal melanin accumulation in the outermost layer of the skin [3–6]. Furthermore, due to its involvement in neuromelanin formation in the brain and neurodegeneration associated with Parkinson’s disease, tyrosinase inhibitors have attracted considerable interest in pharmaceutical industry [7–9].

Tyrosinase inhibitors, such as kojic acid, tropolone, hydroquinone, arbutin, and mercury are currently used to treat skin disorders related to melanin hyperpigmentation (Fig. 1). They are also used in cosmetics as skin-whitening agents and in agriculture as biological insecticides [1, 4]. These inhibitors typically function by chelating the Cu²⁺ ions at the active site of tyrosinase or by displacing them, as seen with mercury [1, 10]. Among tyrosinase inhibitors, kojic acid is a well-known inhibitor, that exerts its inhibitory effect by chelating the copper ion in the active site of tyrosinase [11]. Furthermore, rhodanine-3-propionic acid, lodoxamide, and cytidine 5′-(dihydrogen phosphate) were identified as promising tyrosinase inhibitors, exhibiting superior inhibitory activity compared to α-arbutin (Fig. 1) [12]. Despite their advantages, existing tyrosinase inhibitors have limitations such as limited efficacy, potential toxicity, and suboptimal binding affinity. These challenges underscore the importance of designing and synthesizing new tyrosinase inhibitory agents aimed at enhancing their inhibitory potency and selectivity [1, 13].

Fig. 1.

The structures of kojic acid and hydroxypyridinones

One class of compounds reported as effective tyrosinase inhibitors in recent decades is hydroxypyridinones. Hydroxypyridinones are structurally very similar to kojic acid and bind to the Cu²⁺ ion of tyrosinase enzyme with higher affinity than kojic acid, resulting in more potent inhibition of the enzyme’s activity. Numerous studies on hydroxypyridinone derivatives have demonstrated the strong inhibitory potential of this class of compounds against tyrosinase enzyme [7, 11, 14–17]. In recent studies conducted by our research group, several 3-hydroxypyridin-4-one derivatives demonstrated tyrosinase inhibitory activity with IC₅₀ values ranging from approximately 8.94 µM to 25.82 µM. These values are comparable to or better than that of kojic acid, a commonly used reference tyrosinase inhibitor [18–20]. Tao Zhou et al. synthesized a series of hydroxypyridinone derivatives, some of which exhibited potent inhibition of mushroom tyrosinase monophenolase activity with IC₅₀ values ranging from 1.60 to 2.04 µM and diphenolase activity with IC₅₀ values ranging from 7.99 to 13.89 µM [11]. These findings provide strong evidence for the significant biological efficacy of 3-hydroxypyridin-4-one derivatives and support the rationale of our current study.

Computational techniques provide useful and powerful tools for drug discovery and development. Quantitative structure-activity relationship (QSAR) analysis is a widely used computational method in drug design that builds statistical models to predict the biological activity of new compounds [21, 22].

In this study, a new series of 3-hydroxypyridine-4-one derivatives were synthesized, and their chemical structures were characterized using spectroscopic techniques, including IR, ¹H-NMR, and ¹³C-NMR. The tyrosinase inhibitory activity of these compounds was then predicted through a QSAR analysis based on twenty-four previously reported compounds with known anti-tyrosinase effects. Additionally, molecular docking studies were performed to predict the possible binding modes of the synthesized compounds within the active site of the tyrosinase enzyme.

Results and discussion

Chemistry

The new 3-hydroxypyridin-4-one derivatives were synthesized through a four-step process [18, 19], as illustrated in Fig. 2. The intermediate containing a carboxylic acid group (3) was prepared by reacting maltol (1) with 3-aminobenzoic acid (2) under pH = 5 and HCl (6 M) conditions. In the second step, intermediate (3) was activated by reaction with carbonyldiimidazole (CDI), followed by treatment with dimethylaminopyridine (DMAP) and methanol, yielding the ester derivative (4). The third step involved the conversion of the ester derivative (4) to the corresponding acyl hydrazide derivative (5) using hydrazine hydrate (NH₂NH₂.H₂O). Finally, compound (5) underwent Schiff base condensation with various aldehydes to afford the target derivatives 6a-i.

Fig. 2.

Regents and conditions: a H₂O, EtOH, HCl (6.0 M), pH = 5.0, 100 ˚C, 72 h; b Aceton, CDI, CH₃OH, DMAP, r.t., 24 h; c NH₂NH₂.H₂O, CH₃OH, reflux, 24 h; d EtOH, CH₃COOH, aldehydes, reflux, 4 h

QSAR study

A set of twenty-four previously synthesized 3-hydroxypyridin-4-one derivatives served as the training set, with their experimentally measured pIC₅₀ values against tyrosinase [18–20]. QSAR modeling was performed using three statistical methods including MLR, FA-MLR, and GA-MLR. The structures of the studied compounds are shown in Fig. 3.

Fig. 3.

Structures of previously synthesized 3-hydroxypyridin-4-one derivatives that served as the training set [18–20]

To explore the relationship between the chemical structure and anti-tyrosinase activity of 3-hydroxypyridine-4-one derivatives, three chemometric methods including, stepwise multiple linear regression (MLR), factor analysis combined with MLR (FA-MLR), and genetic algorithm-based MLR (GA-MLR) were used. Given that the training set comprises structurally similar compounds based on a common 3-hydroxypyridin-4-one scaffold, rigorous external validation was deemed crucial to assess the model’s true predictive power for new analogues. Based on a comprehensive evaluation of statistical parameters (Table 1), the MLR model was selected as the optimal equation. While all models were statistically significant, the MLR model demonstrated superior overall performance, evidenced by its combination of the highest fitted correlation coefficient (R² = 0.93), the highest cross-validated predictive ability (Q² = 0.81), the lowest standard error (SE = 0.34), and, most importantly, the highest predictive power for the external test set (R²p = 0.799). This consistent superiority across multiple validation metrics confirms the MLR model as the most robust and reliable for predictive purposes. While a Q² value of 0.81 is below the ideal threshold of 0.9, it is considered ‘good to excellent’ according to established benchmarks in QSAR modeling [23] particularly for a congeneric series of moderate size. The model’s primary strength, however, is demonstrated by its external predictive power, as confirmed by the strong correlation between predicted and experimental activities for compounds6h and6i. The stepwise-MLR procedure resulted in a final model with eight descriptors, optimally balancing predictive power and model simplicity. The risk of multicollinearity, which can undermine the interpretability and stability of a regression model, was assessed. The Variance Inflation Factor (VIF) was calculated for each descriptor, and all values were found to be below 5 (Table S3), confirming that multicollinearity is not a significant concern in the presented mode. The comprehensive QSAR analysis of 3-hydroxypyridine-4-one derivatives revealed several critical structure-activity relationships influencing their tyrosinase inhibitory potential. As dedicated in Table 1, according to the MLR model, eight significant factors were obtained. Mass, PJI2, and nHAcc negatively affect the anti-tyrosinase activity of the 3-hydroxypyridine-4-one derivatives. The observed negative correlation between molecular mass (Mass) and inhibitory activity suggests that smaller molecular frameworks may facilitate better binding interactions within the enzyme’s active site. This finding is consistent with previous reports on other classes of tyrosinase inhibitors [4], where steric factors play a crucial role in determining inhibitory potency. The negative coefficient of the nHAcc descriptor, while seemingly counterintuitive, can be rationalized by considering the unique environment of the tyrosinase active site. The binuclear copper center requires specific, strong coordination bonds for effective inhibition. An increase in the total number of hydrogen bond acceptors may lead to suboptimal orientation or compete with the essential carbonyl and hydroxyl groups of the hydroxypyridinone core for coordinating the copper ions, thereby reducing binding efficiency. This suggests a nuanced balance where the strategic placement of hydrogen bond acceptors is more critical than their total number, a phenomenon supported by studies on metalloenzyme inhibition [24–26]. These findings align with the proposed mechanism of tyrosinase inhibition, where electron-rich substituents facilitate binding via both hydrogen bonding and metal coordination, thereby enhancing inhibitory potency. The other descriptor that effect the anti tyrosinase activity was PJI2 (second Petitjean shape index) which measures molecular shape eccentricity, suggests that more compact and symmetric molecular shapes are favorable for tyrosinase inhibition. This implies that bulky or highly elongated substituents may hinder optimal fitting into the enzyme’s active site. The positive contribution of benzene ring substitutions (nCb) to activity highlights the importance of aromatic interactions in tyrosinase inhibition. Multiple substitutions on the hydrazide phenyl moiety likely enhance binding through both π-π stacking interactions and improved hydrophobic contacts with the enzyme’s binding pocket. This observation is consistent with crystallographic studies of tyrosinase-inhibitor complexes, which highlight the significance of aromatic interactions in the binding process [26]. The spatial arrangement descriptor F03[C-Cl] provided valuable insights into the geometric requirements for optimal inhibition. Enhanced activity associated with ortho-positioned substituents suggests that specific spatial orientations are necessary for productive binding interactions, either through direct interactions with the enzyme or conformational effects on the inhibitor scaffold that promote favorable binding geometry. Additionally, connectivity and WHIM descriptors (Hypertense 50, SEigz, E2u, G2e) showed positive correlations, indicating their beneficial contribution to inhibitory activity. The second model (FA-MLR), with a fitness of 0.86, included five descriptors obtained from the final MLR modelling. The GA-MLR model (Eq. 3) indicated the positive effects of E2U, X3AV, SEigZ, and nCb, alongside negative effects of nHAcc and nArCOOH on the anti-tyrosinase activity of the compounds. As it was showed, about 68% of variances in the original data matrix could be explained by the selected six parameters.

Table 1.

The results of different QSAR model analysis

Eq no.	Equation	R 2	Q2	F	SE	R 2 p
MLR	pIC₅₀ = -0.004(± 0.001) M -1.304(± 0.660) PJI2 +0.552(± 0.091) SEigz +0.197(± 0.561) E2u +2.967(± 1.772) G2e +0.046(± 0.568) H50 -0.212(± 0.069) nHA +0.131(± 0.055) nCb +0.060(± 0.045) F03[C-Cl] +5.049(± 1.349)	0.937	0.810	8.44	0.34	0.799
FA-MLR	pIC₅₀ = -0.094(± 0.031) M -3.304(± 0.740) PJI2 + 2.552(± 0.191) SEigz + 7.416(± 0.532) X3Av + 4.354(± 0.109) nCb − 3.458(± 0.168) nHA − 1.503(± 1.274)	0.869	0.825	89.94	0.49	0.704
GA-MLR	pIC₅₀ = 11.352(± 2.01) E2u + 32.416(± 8.532) X3Av + 7.345(± 0.761) SEigz + 2.154(± 0.135) nCb − 4.571(± 0.152) nHA − 0.239(± 0.013) nArCOOH + 6.908(± 0.758)	0.687	0.59	79.53	0.157	0.650

M Molecular weight, nHA Number of hydrogen bond acceptors, nCb Number of benzene rings, PJI2 Second Petitjean shape index, SEigz, E2u, G2e WHIM descriptors (Weighted Holistic Invariant Molecular descriptors), H50 (Hypertens50) Hyperchem descriptor, F03[C-Cl] 2D frequency of C-Cl at topological distance 3, X3Av Average connectivity index of order 3, nArCOOH Number of aromatic carboxylic acids. R²: Regression Coefficient for Calibration set, Q²: Regression Coefficient for Leave one out Cross Validation, R²p: Regression Coefficient for prediction set

Overall, our comprehensive QSAR analysis of 3-hydroxypyridine-4-one derivatives successfully identified several key structural features governing tyrosinase inhibition. The developed MLR model (R² = 0.93, Q² = 0.81) demonstrated superior predictive capability, revealing that optimal inhibitory activity depends on: (1) low molecular bulk (negative Mass, PJI2 correlations), (2) strategic hydrogen bond acceptors (negative nHAcc effect), particularly hydroxyl and halogen substituents capable of copper coordination, (3) aromatic systems (positive nCb) for π-π stacking and hydrophobic stabilization, and (4) specific spatial orientations (F03[C-Cl]). The positive contributions of WHIM descriptors (SEigz, E2u) further underscore the importance of molecular shape and electron distribution. These findings provide a robust structural framework for rational design of potent tyrosinase inhibitors, with the MLR model serving as a valuable predictive tool. Future work should focus on synthesizing and evaluating compounds incorporating these optimized features to validate our computational predictions. To assess the applicability domain (AD) and reliability of the model, the leverage values (h) were calculated for each compound. The warning leverage threshold (h*) was determined by the equation h*= 3k/n, where n is the number of training compounds and k represents the number of model parameters. Compounds with leverage values exceeding h^∗ are considered outside the AD. As shown in Table S1, all compounds exhibited leverage values below this threshold, confirming their placement within the model’s applicability domain. Additionally, a Y-randomization test was performed to evaluate the risk of chance correlation. In this test, the dependent variable (pIC₅₀) was randomly shuffled ten times, and new models were generated. The resulting statistical parameters (R² and R²LOOCV) were significantly lower than those of the original QSAR models (Table S2), confirming that the models were not obtained by random chance. The robustness and predictive power of the developed MLR model were rigorously assessed through internal and external validation techniques. The high Q² value (0.81) obtained from Leave-One-Out cross-validation indicates the model’s stability and low risk of overfitting. Furthermore, the Y-randomization test confirmed that the model was not based on chance correlation. It is worth noting that the model adheres to the common statistical rule of thumb (Topliss’s rule), which requires a minimum number of compounds for a reliable model. Our dataset of 24 compounds for a model with 8 descriptors well exceeds this minimum requirement, providing further confidence in the model’s statistical significance.

The applicability domain (AD) of the final MLR model was characterized using the leverage approach. The warning leverage (h*) was determined to be 1.25, calculated as 3p/n, where p is the number of model descriptors plus one, and n is the number of training compounds. All training set compounds were confirmed to reside within the AD (h < h*), as detailed in Table S1. For the newly designed compounds, the leverage analysis indicated that compound 6 g, characterized by its bulky anthracene-based substituent, fell outside the model’s AD (h = 1.32 > h*). Consequently, the predicted activity for 6 g should be interpreted with caution. All other newly synthesized compounds (6a-f, 6 h, 6i) were found to be within the applicability domain, and their predictions are therefore considered reliable.

The anti-tyrosinase activity

The predicted anti-tyrosinase activities of the newly synthesized compounds, calculated by the MLR method, are summarized in Table 2. Additionally, the in vitro anti-tyrosinase activities of compounds 7 h and 7i are also presented in Table 2.

Table 2.

The predicted anti-tyrosinase activity of the 3-hydroxypyridin-4-one derivatives 6a-i

Compounds	Ar	Predicted IC50 (µM)*	Experimental IC50 (µM)
6a		> 100	ND
6b		> 100	ND
6c		> 100	ND
6d		> 100	ND
6e		89.4	ND
6f		43.8	ND
6 g		74.3	ND
6 h		> 100	> 100
6i		89.2	98.5
Kojic Acid	--	----	19.1

IC₅₀: 50% inhibitory concentration (mean ± SD of three independent experiments); * Predicted IC₅₀ by MLR model

The integration of QSAR modeling and experimental validation has provided valuable insights into the structural determinants governing the anti-tyrosinase activity of 3-hydroxypyridin-4-one derivatives. Our MLR-based QSAR model (R² = 0.93, Q² = 0.81) demonstrated excellent predictive performance, as evidenced by the strong correlation between predicted and experimental IC₅₀ values for compounds 6h and 6i. The model’s reliability was further confirmed through rigorous validation techniques, including applicability domain assessment and Y-randomization tests. The SAR analysis identified several critical structural features influencing inhibitory activity. Molecular size emerged as a key factor, with smaller derivatives generally exhibiting greater potency, supported by negative coefficients for the mass and PJI2 descriptors. This observation aligns well with the steric constraints of tyrosinase’s active site, where bulkier substituents may hinder optimal binding. The importance of hydrogen bonding interactions was underscored by the negative correlation of the nHAcc descriptor, suggesting that strategic placement of hydrogen bond acceptors can significantly impact inhibitory activity. Notably, aromatic interactions played a vital role in enhancing potency, as indicated by the positive coefficient of the nCb descriptor. This finding is consistent with crystallographic studies of tyrosinase-inhibitor complexes, which emphasize the importance of π-π stacking interactions in the binding process. Additionally, the spatial arrangement of substituents, particularly at ortho positions (F03[C-Cl]), contributed to activity, emphasizing the significance of molecular geometry in inhibitor design. The strong agreement between predicted and experimental values for compound 6i (89.2 µM predicted vs. 98.5 µM experimental) validates the accuracy of our modeling approach. However, the superior activity of kojic acid (19.1 µM) suggests that additional factors, especially metal-chelating capacity, should be incorporated in future models. The predicted potency of compound 6f (43.8 µM), although not yet experimentally verified, offers a promising candidate for further study, as its structural features appear to optimally balance the activity-governing parameters revealed by our model. These findings have important implications for rational drug design.

Molecular docking study

The molecular docking study of the newly synthesized compounds was performed to investigate their binding modes within the active site of the tyrosinase enzyme with PDB code: 2Y9X. Before docking the synthesized compounds into the active site of tyrosinase, a redocking procedure was performed using tropolone to validate the docking protocol. The reliability of this protocol was confirmed by obtaining a root-mean-square deviation (RMSD) value of less than 2 Å, as shown in Fig. 4.

Fig. 4.

The crystal orientation (pink) and the re-docked model (yellow) of original ligand (tropolone)

The binding free energies of compounds 6e, 6f, and 6g, which were predicted as the most potent inhibitors, were − 6.2, -7.14, and − 6.7 kcal.mol⁻¹, respectively, which are comparable to the binding free energy of kojic acid (-6.2 kcal.mol⁻¹). According to the previously reported literature, the critical residues of tyrosinase include His244, His85, His263, Val283, His296, Asn260, Val248, His261, and Phe264. The active site of the tyrosinase enzyme is composed of both hydrophobic and hydrophilic amino acids, as well as two copper ions. The hydrophobic region consists of Val 283, Met280, Ala246, Phe90, Ala286, Val 248, Phe246, and His256 residues, whereas the hydrophilic region contains His244, His279, Gly86, His61, Ser282, His 85, His 263, Gly281, Glu256, Asn260, His256, Glu322, Gly96, and Ser282 residues [27]. The best docked poses of compounds 6e, 6f, and 6g in the active site of tyrosinase enzyme, with the PDB code of 2Y9X, are shown in Fig. 5.

Fig. 5.

The interaction of compounds 6e, 6f, and 6 g in the active site of tyrosinase. (green: Vander waals, light green: carbon hydrogen bond, dark green: hydrogen bond, dark pink: π-π, light pink: alkyl & π-alkyl, purple: π-sigma, orange: π-cation, yellow: π-sulfur, blue: halogen)

As shown in Fig. 5, the carbonyl and hydroxyl groups of compounds 6e and 6f formed hydrogen bonds with His85 residue. Additionally, the fluorine (F) substitution in compound 6e interacted with the Asn260 residue, and its phenyl rings formed π-sigma and π-sulfur interactions with the Val283, Val248, and Met257 residues. The amide carbonyl group and OCH₃ groups of compound 6f formed hydrogen bonds with Val283 and Arg268 in the hydrophobic region, respectively. The OCH₃ group of compound 6f contributed to the formation of hydrogen bond, which resulted in superior predicted inhibition activity of compound 6f. Furthermore, the phenyl ring of compound 6f formed a π-π stacking interaction with Phe264. Compound 6 g also interacted with key residues in the active site. The OH and C = O groups of compound 6g formed hydrogen bonds with Ser282 and His244, respectively. Additionally, the anthracene moiety of 6g established π-π stacking interactions with the aromatic side chain of Phe264. The interactions of other compounds with the target enzyme are provided in the supplementary file. The results of molecular docking were consistent with the predicted biological results, further confirming that the substitutions of fluorine and OCH₃ on the phenyl ring positively enhanced the inhibitory activity of these compounds.

The integration of docking results with our QSAR predictions provides a molecular-level rationale for the inhibitory performance of the synthesized compounds relative to kojic acid. While the 3-hydroxypyridin-4-one scaffold serves as a competent bidentate chelator, the superior activity of kojic acid (IC₅₀ = 19.1 µM) over our best predicted compound, 6f (IC₅₀ = 43.8 µM), can be attributed to fundamental differences in metal-coordination geometry and binding entropy. Kojic acid, being a small, rigid molecule, positions its γ-pyrone carbonyl and hydroxyl groups in an ideal geometry for simultaneous coordination with both CuA and CuB ions in the tyrosinase active site, forming a stable, low-entropy complex.

In contrast, our derivatives, despite possessing the essential chelating motif, incorporate a more flexible hydrazide linker and bulkier substituents. While these features contribute to additional hydrophobic and π-π stacking interactions (as captured by the positive nCb descriptor in our QSAR model), they may also introduce conformational strain or suboptimal orientation of the hydroxypyridinone core, reducing the efficiency of copper chelation. This suggests a trade-off: the structural features that enhance binding through auxiliary interactions might simultaneously compromise the ideal geometry for primary metal coordination. Consequently, while our QSAR model successfully identifies key descriptors for binding affinity (e.g., molecular size, aromaticity), the ultimate inhibitory potency is governed by a delicate balance between metal-chelation efficiency and supplemental interactions—a nuance that may not be fully parameterized by the current descriptor set. This insight directs future work toward designing more constrained analogs that pre-organize the chelating motif for optimal metal coordination while strategically incorporating the favorable hydrophobic substituents identified in this study.

Materials and methods

Chemistry

All reagents of analytical grade were purchased from Sigma-Aldrich for the synthesis of 3-hydroxypyridin-4-one derivatives. All solvents were obtained from Merck. The ¹H-NMR and ¹³C-NMR spectra were recorded using Bruker 300 and 400 MHz spectrometers. For recording the NMR spectra, samples were dissolved in either DMSO-d₆ or CDCl₃, DMSO-d6 as solvents. Specra were acquired at a temperature of 25 °C using standard Bruker pulse programs. Tetramethylsilane (TMS) was used as the internal standard. The infrared (IR) spectra of the synthesized compounds were recorded using a PerkinElmer Vertex 70 FT-IR spectrometer with a spectral resolution of 4 cm⁻¹ and 16 scans per sample. The samples were dried in an oven and prepared as potassium bromide (KBr) pellets for analysis. Melting points were measured with an electrothermal IA 9100 device.

Synthesis of intermediate 3

In a flask, 98 mL of water and 10 mL of ethanol were added to a mixture of 50 mmol of maltol (1) and 50 mmol of 3-aminobenzoic acid (2). Then, the pH of the reaction was adjusted to 5 using hydrochloric acid (6 M). The resulting mixture was refluxed for 72 h, and the progress of the reaction was monitored by TLC. After completion of the reaction, the contents of the flask were filtered, and the obtained precipitate was washed with acetone. Finally, a cream-colored precipitate (3) was obtained with a yield of 86 ± 5.2%.

Synthesis of intermediate 4

5 mmol of intermediate 3, 9 mmol of carbonyldiimidazole (CDI), and 15 mL of dry acetone were placed in a two-necked flask and stirred at room temperature for one hour. Separately, 0.5 mmol of dimethylaminopyridine (DMAP) was dissolved in 7 mL of dry methanol and stirred for 30 min. The DMAP solution was then added to the flask containing compound (3) and CDI, and the reaction mixture was stirred at room temperature for 24 h. The progress of the reaction was monitored by TLC. Afterwards, the reaction mixture was extracted, dried over anhydrous sodium sulfate (Na₂SO₄), and concentrated by rotary evaporation. After drying, a white precipitate (4) was obtained with a yield of 20 ± 3.8%.

Synthesis of intermediate 5

A mixture of 0.5 mmol of intermediate (4), 1 mmol of hydrazine hydrate, and 5 mL of methanol was placed in a flask and refluxed at 90 °C for 24 h. The progress of the reaction was monitored by TLC. After the completion of the reaction, the resulting precipitate was separated by filtration and washed with cold methanol. After drying, a white precipitate (5) was obtained with a yield of 80 ± 2.4%.

Synthesis of final compounds 6a-i

0.5 mmol of aldehyde derivatives and 5 mL of dry ethanol, along with 4 drops of glycine acetic acid, were placed in a flask and stirred at room temperature for 15 min. Then, 0.5 mmol of intermediate 5 was dissolved in 5 mL of dry ethanol and slowly added to the reaction mixture. The mixture was refluxed at 90–100 °C for 4 h. After completion, the resulting precipitate was collected and dried. The obtained precipitate was further purified by recrystallization using ethyl acetate or ethanol to yield the final compounds 6a-i.

(E)-N’-butylidene-3-(3-hydroxy-2-methyl-4-oxopyridin-1(4 H)-yl) benzohydrazide (6a)

White solid, Yield: 69.8%, R_f= 0.6, m.p.: 151–154 ˚C, IR (KBr, ν/cm^− 1): 3225 (N-H, type two amine), 3066 (OH), 1654 (C = O, ketone), 1574 (C = O, amide), 1204 (C = N); ¹H-NMR (400 MHz, CDCl₃) δ (ppm): 12.42 (s, 1H, NH), 8.30 (d, J = 8.0 Hz, 1H), 8.10 (s, 1H), 7.91 (t, J = 5.6 Hz, 1H), 7.62 (t, J = 8.0 Hz, 1H), 7.49 (d, J = 6.8 Hz, 1H), 7.36 (d, J = 7.6 Hz, 1H), 6.24 (d, J = 6.8 Hz, 1H), 2.34–2.29 (q, J = 6.8 Hz, 2 H), 2.03 (s, 3 H), 1.57–1.48 (m, 2 H), 0.926 (t, J = 7.2 Hz, 3 H); ¹³C-NMR (100 MHz, CDCl₃): 169.25, 162.06, 154.34, 145.43, 141.01, 137.98, 134.92, 130.56, 130.46, 130.04, 129.31, 125.44, 111.01, 34.58, 20.02, 13.77, 13.54.

(E)-3-(3-hydroxy-2-methyl-4-oxopyridin-1(4 H)-yl)-N’-(3-methylbutylidene) benzo Hydrazide (6b)

Yellow solid, Yield: 62.7%, R_f= 0.6, m.p.: 210–213 ˚C, ¹H-NMR (300 MHz, DMSO) δ (ppm): 11.48 (s, 1H, NH), 8.00 (t, J = 3.0 Hz, 1H), 7.89 (s, 1H), 7.69–7.65 (m, 3 H), 7.61 (d, J = 7.5 Hz, 1H), 6.24 (d, J = 7.2 Hz, 1H), 1.98 (s, 3 H), 1.91 (br, 1H), 1.08–1.06 (m, 6 H), 0.91 − 0.82 (m, 2 H); ¹³C-NMR (75 MHz, DMSO) δ (ppm): 170.23, 157.82, 150.52, 138.36, 138.32, 135.37, 130.61, 130.57, 130.36, 129.10, 126.29, 111.48, 108.59, 31.56, 21.53, 20.13, 20.06, 13.88.

(E)-N’-hexylidene-3-(3-hydroxy-2-methyl-4-oxopyridin-1(4 H)-yl) benzohydrazide (6c)

White solid, Yield: 70.1%, R_f= 0.6, m.p.: 128–130 ˚C, IR (KBr, ν/cm^− 1): 3235 (N-H, type two amine), 3068 (OH), 1629 (C = O, ketone), 1574 (C = O, amide), 1204 (C = N); ¹H-NMR (400 MHz, DMSO) δ (ppm): 11.53 (s, 1H, NH), 8.03 (d, J = 4.0 Hz, 1H), 7.92 (d, J = 8.0 Hz, 1H), 7.74–7.68 (m, 3 H), 7.62 (d, J = 7.2 Hz, 1H), 6.26 (d, J = 7.2 Hz, 1H); 2.29–2.24 (q, J = 6.8 Hz, 1H); 1.99 (s, 3 H); 1.51–1.47 (m, 2 H); 1.30–1.25 (m, 5 H), 0.86 (t, J = 6.4 Hz, 3 H); ¹³C-NMR (100 MHz, CDCl₃) δ (ppm): 168.97, 162.04, 154.61, 154.40, 141.05, 140.98, 137.84, 134.97, 130.47, 130.06, 129.24, 125.43, 111.08, 31.80, 31.47, 26.36, 22.41, 14.08, 13.95.

(E)-N’-heptylidene-3-(3-hydroxy-2-methyl-4-oxopyridin-1(4 H)-yl) benzohydrazide (6d)

Yellow solid, Yield: 55.6%, R_f= 0.6, m.p.: 137–139 ˚C, IR (KBr, ν/cm^− 1): 3223 (N-H, type two amine), 3067 (OH), 1654 (C = O, ketone), 1575 (C = O, Amide), 1204 (C = N). ¹H-NMR (400 MHz, CDCl₃) δ (ppm): 12.34 (s, 1H, NH), 8.28 (d, J = 7.6 Hz, 1H), 8.05 (s, 1H), 7.90 (t, J = 5.2 Hz, 1H), 7.62 (t, J = 7.6 Hz, 1H), 7.45 (d, J = 5.2 Hz, 1H), 7.36 (d, J = 7.6 Hz,1H), 6.27 (d, J = 6.4 Hz, 1H), 2.35–2.30 (q, J = 6.8, 6.4 Hz, 2 H), 2.03 (s, 3 H), 1.51–1.44 (m, 2 H), 1.30–1.24 (m, 6 H), 0.85 (t, J = 6.8 Hz, 3 H). ¹³C-NMR (100 MHz, CDCl₃) δ (ppm): 172.52, 170.18, 161.74, 153.43, 145.59, 141.92, 138.36, 130.57, 130.34, 129.22, 128.87, 126.26, 111.54, 32.46, 31.53, 28.79, 26.46, 22.49, 14.38, 13.85.

(E)-N’-(2-fluorobenzylidene)-3-(3-hydroxy-2-methyl-4-oxopyridin-1(4 H)-yl) benzo Hydrazide (6e)

Yellow solid, Yield: 71.9%, R_f= 0.6, m.p.: 227–230 ˚C, IR (KBr, ν/cm^− 1): 3188 (N-H, type two amine), 3060 (OH), 1645 (C = O, ketone), 1588 (C = O, amide), 1202 (C = N). ¹H-NMR (300 MHz, DMSO) δ (ppm): 12.20 (s, 1H, NH), 8.80 (s, 1H), 8.10 (br, 1H), 8.03–8.01 (m, 2 H),7.75–7.64 (m, 4 H), 7.49 (t, J = 7.2 Hz,1H), 7.39 (t, J = 7.2 Hz, 1H), 6.26 (d, J = 7.2 Hz, 1H), 2.01 (s, 3 H, CH₃). ¹³C-NMR (75 MHz, DMSO) δ (ppm): 170.25, 162.24, 146.90, 145.63, 142.04, 138.41, 134.89, 134.10, 133.68, 133.35, 132.41, 131.04, 130.48, 129.08, 128.65, 127.80, 126.51, 124.13, 111.55, 13.92.

3-(3-hydroxy-2-methyl-4-oxopyridin-1(4 H)-yl)-N’-(3-5-dimethoxy-4-hydroxy benzylidene) benzohydrazide (6f)

|White solid, Yield: 67.5%, R_f= 0.3, m.p.: 292–295 ˚C, ¹H-NMR (300 MHz, DMSO) δ (ppm): 11.84 (s, 1H), 8.34 (s, 1H), 7.96–8.06 (m, 2 H), 7.65–7.72 (m, 3 H), 6.99 (s, 2 H), 6.25 (d, J = 5.4 Hz, 1H), 3.82 (s, 6 H), 2.01 (s, 3 H); ¹³C-NMR (75 MHz, DMSO) δ (ppm): 170.24, 161.96, 149.49, 148.60, 145.61, 142.01, 138.59, 138.43, 135.40, 130.70, 130.43, 129.10, 128.87, 126.36, 124.79, 111.55, 105.16, 56.50, 13.91.

(E)-N’-(anthracen-9-ylmethylene)-3-(3-hydroxy-2-methyl-4-oxopyridin-1(4 H) yl) benzohydrazide (6 g)

Yellow solid, Yield: 53.4%, R_f= 0.6, m.p.: 277–280 ˚C, ¹H-NMR (300 MHz, DMSO) δ (ppm):12.26 (s, 1H), 9.68 (s, 1H), 8.19–8.13 (m, 5 H), 7.78–7.73 (m, 3 H), 7.69–7.57 (m, 6 H), 6.37 (d, 1H, J = 7.2 Hz), 2.07 (s, 3 H). ¹³C-NMR (75 MHz, DMSO) δ (ppm): 169.61, 166.63, 148.00, 145.50, 142.07, 138.52, 135.33, 135.19, 131.37, 130.93, 130.61, 130.29, 130.11, 130.06, 129.53, 127.75, 126.50, 126.08, 125.42, 125.31, 120.47, 111.56, 14.05.

(E)-3-(3-hydroxy-2-methyl-4-oxopyridin-1(4 H)-yl)-N’-(2-phenylpropylidene) benzo Hydrazide (6 h)

White solid, Yield: 66.5%, R_f= 0.6, m.p.:166–168 °C; IR (KBr, ν/cm^− 1): 3344 (OH), 3219 (N-H), 1658 (C = O, ketone), 1579 (C = O, amide). ¹H-NMR (300 MHz, DMSO) δ (ppm): 11.53 (s, 1H), 8.00 (s, 1H), 7.89 (s, 1H), 7.82 (d, J = 6.0 Hz, 1H), 7.69 (d, J = 4.2 Hz, 2 H), 7.60 (d, J = 7.5 Hz, 1H), 7.39–7.26 (m, 5 H), 6.23 (d, J = 7.5 Hz, 1H), 3.80–3.75 (m, 1H), 1.99 (s, 3 H), 1.43 (d, J = 6.9 Hz, 3 H).

3-(3-hydroxy-2-methyl-4-oxopyridin-1(4 H)-yl)-N’-((1E,2E)-3-phenylallylidene) benzo Hydrazide (6i)

White solid, Yield: 54.3%, R_f= 0.6, m.p.:170–172 °C, IR (KBr, ν/cm^− 1): 3376 (OH), 3037 (N-H), 1670 (C = O, ketone), 1557 (C = O, amide). ¹H-NMR (300 MHz, DMSO) δ (ppm): 11.81 (s, 1H), 8.22 (s, 1H), 8.06 (t, J = 4.2 Hz, 1H), 7.96 (s, 1H), 7.72 (d, J = 4.8 Hz, 2 H), 7.65 (d, J = 3.0 Hz, 1H), 7.63 (d, J = 3.6 Hz, 1H), 7.43–7.34 (m, 4 H), 7.10–7.08 (m, 2 H), 6.25 (d, J = 7.2 Hz, 1H), 2.00 (s, 3 H).

Tyrosinase inhibitory assay

The tyrosinase inhibitory activity of two compounds, 6i and 6h, was determined according to a previously reported method [19]. Briefly, serial dilutions of the tested compounds were prepared. In 96-well microplates, a mixture of tyrosinase and the tested compounds was incubated for 20 min. Subsequently, L-DOPA, as the substrate, was added to the mixture, and the absorbance was measured at 475 nm. The inhibition percentage of the tested compounds was calculated as follows:

% Inhibition = (A_control − A_sample/A_control) × 100.

QSAR study

Data set

For QSAR analysis, twenty-four 3-hydroxypyridin-4-one derivatives with similar chemical structures, which we previously synthesized and evaluated their anti-tyrosinase activities, were selected [18–20]. The compounds were chosen to form a congeneric series featuring systematic variations, primarily in the hydrazide side chain, to ensure a meaningful range of molecular descriptors while maintaining a consistent core structure for QSAR model development. These compounds were classified into calibration and test sets using the Kennard-Stone algorithm. The IC₅₀ values representing the anti-tyrosinase activities of these compounds were converted to logarithmic scale (pIC₅₀) and employed as the dependent variable in the QSAR analysis.

Descriptors

The chemical structures of the ligands were drawn using HyperChem 8.0 software and optimized using the AM1 and MM + algorithms. Subsequently, a range of molecular descriptors were calculated using Hyperchem v8.0, Gaussian98 software, and Dragon (version 5.5). Physicochemical properties such as molecular volume (V), hydration energy (HE), hydrophobicity (LogP), molecular surface area (SA), and molecular polarizability (MP) were obtained using Hyperchem v8.0. Gaussian98 software (Gaussian 09 Revision-A.02-SMP) was used to calculate the highest occupied molecular orbital (HOMO), the lowest unoccupied molecular orbital (LUMO) energies, the foremost positive, the most negative net atomic charges, the average absolute atomic charge, as well as the molecular dipole moment. Additionally, Dragon (version 5.5) was employed to compute different topological, geometrical, empirical, and constitutional descriptors; 2D autocorrelations; atom-centered fragments; and functional groups. Although the physicochemical and quantum-chemical descriptors (HOMO, LUMO, LogP, hydration energy, and polarizability) were calculated using HyperChem v8.0 and Gaussian software.

QSAR models and model validation

Four different chemometrics methods including: (1) stepwise-multiple linear regression (MLR), (2) MLR with factor analysis (FA-MLR), (3) genetic algorithm- multiple linear regression (GA-MLR), and (4) principal component regression (PCR) were used to generate the QSAR equation. To validate the resulting equation, several statistical parameters were utilized including the correlation coefficient (R²), leave-one-out cross-validation correlation coefficient (Q²), cross-validation (Cvcv), variance ratio (F), standard error of regression (SE), and root mean square error for an external set of dataset (R²p) were applied. The predictive capability of the QSAR model was assessed using this external data set (R²p). The final MLR equation was developed using a stepwise variable selection method to ensure model parsimony and avoid overfitting. This approach iteratively includes only descriptors that make a statistically significant contribution (p < 0.05) to the inhibitory activity. To quantify and address potential multicollinearity among the selected descriptors, the Variance Inflation Factor (VIF) was calculated for each. A VIF value greater than 5 is typically considered indicative of significant multicollinearity. In our final model, all VIF values were well below this threshold (see Supplementary Material, Table S3), confirming that the selected descriptors are sufficiently independent and that the model is statistically stable.

Molecular docking study

The 3D crystal structure of the tyrosinase enzyme with PDB code 2Y9X was downloaded from the RCSB Protein Data Bank (https://www.rcsb.org/). Subsequently, water molecules and the co-crystallized ligand were removed from the structure. Polar hydrogens and Gasteiger charges were then added to the receptor structure using AutoDock Tools (v1.5.6). The prepared structure was subsequently saved in PDBQT format for use in molecular docking simulations. The structures of target ligand were drawn, optimized, and saved in the pdbqt format. The molecular docking study of the target ligands were conducted using Autodock Tools 1.5.6 software with the Lamarckian genetic algorithm. The key genetic algorithm parameters were set as follows: number of GA runs: 10, population size: 150, maximum number of energy evaluations: 2,500,000, maximum number of generations: 27,000, mutation rate: 0.02, crossover rate: 0.8, elitism value: 1, and the root-mean-square deviation (RMSD) tolerance for clustering was set to 2.0 Å. Validation of the molecular docking was performed by re-docking the native ligand of the 2Y9X crystal structure, which is tropolone. The grid box was set with dimensions of 40 × 40 × 40 Å, centered at coordinates x = -10.044, y = -28.706, and z = -43.443 Å. The grid spacing was 0.375 Å. Finally, the newly synthesized compounds were docked into the tyrosinase active site using optimized parameters, and Discovery Studio 2016 64-bit Client was used to visualize the interactions of the docked compounds.

Conclusion

The present study successfully synthesized a new series of 3-hydroxypyridin-4-one derivatives and evaluated their potential as tyrosinase inhibitors through combined experimental and computational approaches. A QSAR analysis was performed on 24 hydroxypyridinone derivatives with known anti-tyrosinase activity using three different methods: MLR, FA-MLR, and GA-MLR. Given the structural similarity of the training set compounds, which share a common 3-hydroxypyridin-4-one scaffold, external validation was considered particularly crucial. Despite the moderate size of the training set, the MLR model, selected as the optimal model, demonstrated strong predictive capability and statistical robustness, as evidenced by high R² (0.93) and Q² (0.81) values. The model’s reliability was further confirmed through rigorous validation techniques, including leave-one-out cross-validation and Y-randomization, which minimized the risk of chance correlation or overfitting. The MLR model highlighted the importance of Hyperchem descriptors, 2-D functional descriptors, and GETAWAY descriptors for designing new and potent compounds. The identification of key molecular descriptors influencing tyrosinase inhibition offers valuable insights for the rational design of future inhibitors. Moreover, molecular docking analyses confirmed stable and favorable interactions between the synthesized compounds and the active site of tyrosinase, supporting their potential efficacy. Most importantly, the strong correlation between the predicted and experimental IC₅₀ values for the newly synthesized, structurally similar compounds (6 h and 6i) provides crucial external validation and confirms the model’s practical utility for designing new analogs within this congeneric series. Although our series of 3-hydroxypyridin-4-one derivatives generally show lower potency compared to kojic acid, the study primarily aims to elucidate key structural determinants influencing tyrosinase inhibition. The insights gained provide a foundation for rational design and optimization of more potent inhibitors. Furthermore, several compounds in our series, such as compound 6f, exhibit promising predicted activity, highlighting potential for further optimization.

Author contributions

H. S. supervised the study, S. S. contributed to the synthesis of the compounds and prepared the manuscript, L. E. edited the manuscript and supervised the computational study, M. Kh. contributed to the biological assay, P. R. contributed to the QSAR study, Z. K. Gh. contributed to the prepared the manuscript, (A) M. contributed to the synthesis of compounds, A. (B) contributed to the synthesis of the compounds, R. S. supervised the study. All authors read and approved the final manuscript.

Funding

Financial assistance from the Shiraz University of Medical Sciences by way of grant numbers 26571 and 25951 is gratefully acknowledged.

Data availability

The data sets used and analyzed during the current study are available from the corresponding author upon reasonable request. We have presented all data in the form of Figures, Tables, and Supplementary information file. The PDB code (2Y9X) was retrieved from protein data bank (www. rcsb. org). https://www.rcsb.org/structure/2Y9X.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.