Synthesis, X-ray structures and biological evaluation of novel α-aminophosphonate derivatives as anti-inflammatory and antioxidant agents: an integrated in vitro and in silico study Authors: Zoubir Douib, Hacene K’tir, Haiet Douib, Zinelaabidine Cheraiet, Taha Khaldi, Latifa Khattabi, Brahim Harkati, Vincent Dorcet, Fabrice Pointillart Published: 2026 ABSTRACT In the field of medicinal chemistry, the design of small molecules is acknowledged as an effective strategy. This study may contribute to the global fight against inflammatory diseases by focusing on the synthesis of novel α-aminophosphonate derivatives incorporating para-aminobenzoic acid (PABA) and benzocaine moieties and evaluating their anti-inflammatory and antioxidant activities. Compounds 1a–1i, 2a–2g and 3a–3f were efficiently prepared via a one-pot Kabachnik–Fields reaction in moderate to high yields. Structural characterization was achieved using spectroscopic techniques and single-crystal X-ray diffraction. Antioxidant activity, assessed by ABTS (<0.1 mM) and phenanthroline (<0.5 mM) assays, revealed that several derivatives exhibited significant radical scavenging and reducing power, particularly those bearing electron-donating groups. In vitro anti-inflammatory evaluation using the bovine serum albumin denaturation assay identified compounds 1f, 3f, 2g and 1i as the most active derivatives, with compound 1f (97.63% ± 0.99%) showing activity comparable to that of diclofenac (98.70% ± 0.15%). Toxicity tests on Tenebrio molitor larvae indicated low toxicity at effective concentrations. Compound 1f was only moderately toxic (50% of mortality rate) at the highest concentration tested (4 mM), and no mortality was observed at lower doses. DFT calculations provided insights into the electronic and physicochemical properties of the synthesized compounds. Molecular docking and molecular dynamics simulations generated hypotheses regarding their possible interactions with the COX-2 binding site, while ADMET analyses predicted the potentially acceptable pharmacokinetic profiles of several derivatives. However, these computational findings do not demonstrate direct COX-2 inhibition, target engagement, or selectivity. When considered alongside the preliminary experimental results, they support prioritizing compound 1f for further experimental evaluation. Introduction Inflammation is a fundamental protective response of the immune system that plays a crucial role in maintaining tissue homeostasis under the conditions of injury and infection or exposure to other physiological stressors.1 While acute inflammation is essential for host defence, its dysregulation or failure to resolve can lead to chronic inflammatory states that contribute to the development of autoimmune and inflammatory diseases, including asthma, psoriasis, and rheumatoid arthritis.2 These disorders affect millions of individuals worldwide and represent a major global public health burden. Consequently, the management of inflammatory diseases relies heavily on the use of analgesic and anti-inflammatory agents.3 However, the long-term use of currently available anti-inflammatory drugs, particularly nonsteroidal anti-inflammatory drugs (NSAIDs), is often associated with significant adverse effects. This has driven ongoing efforts to develop safer and more effective anti-inflammatory agents with improved therapeutic profiles. Organophosphorus compounds have attracted considerable attention due to their broad range of biological and pharmacological activities.4,5 Among them, α-aminophosphonates represent an important class of bioactive molecules, recognised for their relatively low toxicity and diverse biological properties, including antifungal,6,7 antibacterial,8,9 antitumour,10–12 and antiviral activities.13 These compounds are the structural analogues of α-amino acids and have therefore emerged as promising candidates in medicinal chemistry research.14 Para-aminobenzoic acid (PABA) and its derivatives are also known to exhibit a wide spectrum of biological activities, such as antimicrobial,15 antioxidant,16 and cytoprotective effects.17 The presence of both amino and carboxyl functional groups in the PABA scaffold provides versatile sites for chemical modification, making it an attractive building block for the design of novel bioactive compounds.18 In particular, the ethyl ester of PABA, commonly known as benzocaine, is a widely used local anaesthetic in clinical practice.19 Beyond its anaesthetic application, benzocaine has been reported to possess antibacterial,20 antimicrobial,21 antifungal,22 and antitumour activities,23,24 further highlighting its pharmacological relevance (Fig. 1). α-Aminophosphonates and PABA with potent biological activities. Based on these evidence, we hypothesised that the molecular hybridisation of α-aminophosphonate structures with PABA or benzocaine moieties could lead to new compounds with enhanced biological properties. Such a strategy may combine the favourable pharmacological features of both scaffolds and provide access to novel anti-inflammatory agents with additional antioxidant potential. In the present study, we report the synthesis of new α-aminophosphonate derivatives incorporating PABA and benzocaine moieties and the in vitro evaluation of their anti-inflammatory and antioxidant activities. To gain molecular-level insights into their biological behaviour, comprehensive in silico studies, including density functional theory (DFT) calculations, molecular docking, molecular dynamics simulations, and ADMET predictions, were also performed. Results and discussion Chemistry The designed α-aminophosphonate derivatives were synthesized via the Kabachnik–Fields reaction, involving the condensation of an amine, an aldehyde, and triethyl phosphite. This multicomponent protocol proved to be a practical and efficient approach for the rapid construction of structurally diverse α-aminophosphonate derivatives. All reactions were carried out under similar conditions (60 °C, 3 h), allowing for the direct comparison of reactivity and yields across different series. In the synthesis of series 1 and 2 α-aminophosphonate compounds (Scheme 1), reactions involving aromatic aldehydes bearing electron-donating substituents (e.g. methoxy and hydroxy groups) generally afforded higher yields (typically >90%). This behavior can be attributed to the formation of imine and increased reactivity of the intermediate toward nucleophilic attack facilitated by triethyl phosphite. Synthesis of compounds 1a–1i and 2a–2f. In contrast, aldehydes containing electron-withdrawing substituents, such as halogens, afforded slightly reduced yields (81–87%), likely due to the decreased electrophilicity of the imine carbon or steric hindrance affecting phosphite addition. Nevertheless, the yields remained satisfactory, demonstrating the robustness of the protocol. In the synthesis of series 3 α-aminophosphonate compounds (Scheme 2), where ethyl 4-formylbenzoate was employed as the aldehyde component, the reaction proceeded efficiently across a range of aromatic amines. Minor yield variations were observed depending on the electronic nature of the amine, with electron-donating anilines providing marginally higher yields than their halogen-substituted analogues. The presence of ester functionalities did not interfere with the reaction, further highlighting the excellent functional group selectivity of the Kabachnik–Fields approach. Synthesis of compounds 3a–3f. All reactions were monitored by TLC analysis. The products were isolated as crystalline solids and purified by simple recrystallisation from ethanol, eliminating the need for chromatographic purification. The consistently high yields, sharp melting points, and clean spectroscopic data reflect the reproducibility and synthetic reliability of this methodology. The spectroscopic data for all the final compounds (IR, 1H NMR, and 13C NMR) match the structures we assigned to them (Table S1 and Fig. S1–S50 and experimental section in SI). Crystal structure analysis To gain detailed insights into the structural features of the synthesized compounds, single crystals suitable for X-ray diffraction analysis were obtained by slow evaporation from ethanol. The X-ray structures of ten α-aminophosphonate derivatives are presented in Fig. 2–5 and S51–S56. These compounds crystallize in different space groups: triclinic P1̄ for 2e, 3a, 3b, 3c, and 3d; monoclinic P21/c for 2b and 2g; chiral P21 for 1i; and orthorhombic Pbca and P212121 for 1a and 1d. The single-crystal X-ray diffraction analyses confirm the expected chemical structures of these α-aminophosphonate derivatives (Fig. 2–5, S51 and S56). All phosphorus atoms adopt a tetrahedral geometry, coordinated by two ethoxy groups, one Cα atom, and a doubly bonded oxygen atom. (a) ORTEP representation of compound 1a with thermal ellipsoids drawn at the 30% probability level. (b) Crystal packing of 1a, illustrating the formation of a pseudo-one-dimensional chain through hydrogen bonding (light blue dashed lines). Hydrogen atoms are omitted for clarity, except for those involved in hydrogen bonds. (a) ORTEP representation of compound 2b with thermal ellipsoids drawn at the 30% probability level. (b) Crystal packing of 2b, highlighting the formation of pseudo-dimers through hydrogen bonding (light blue dashed lines). Hydrogen atoms are omitted for clarity, except for those involved in hydrogen bonds. (a) ORTEP representation of compound 3d with thermal ellipsoids drawn at the 30% probability level. (b) Crystal packing of 3d, highlighting the formation of pseudo-dimers through hydrogen bonding (light blue dashed lines) and the presence of π–π interactions. Hydrogen atoms are omitted for clarity, except for those involved in hydrogen bonds. (a) ORTEP representation of compound 3c with thermal ellipsoids drawn at the 30% probability level. (b) Crystal packing of 3c, highlighting the formation of pseudo-dimers through hydrogen bonding (light blue dashed lines) and the presence of Br⋯O short contacts (light green dashed lines). Hydrogen atoms are omitted for clarity, except for those involved in hydrogen bonds. The Cα atoms are the chiral centers, accounting for the optical activity of these compounds. Four distinct crystal packing arrangements are observed depending on the α-aminophosphonate, contributing to structural stabilization through hydrogen bonding, Br⋯O short contacts, and π–π stacking interactions. In particular, the presence of an aminobenzoic acid moiety, as in compounds 1a (Fig. 2), 1d (Fig. S51), and 1i (Fig. S52), leads to the formation of pseudo-one-dimensional molecular networks via hydrogen bonding (e.g., N1–H1N⋯O4 = 2.966(5) Å and O5–H5O⋯O1 = 2.612(5) Å for 1a). The esterification of aminobenzoic acid, as observed for compounds 2b (Fig. 3) and 2g (Fig. S53), promotes the formation of pseudo-dimers through hydrogen bonding (e.g., N1–H1N⋯O3 = 2.943(5) Å for 2b) rather than the formation of one-dimensional chains. When the ethyl benzoate moiety is not directly attached to the amine—as in compounds 3a (Fig. S54), 3b (Fig. S55), and 3d (Fig. 4)—both the formation of pseudo-dimers via hydrogen bonding (e.g., N1–H1N⋯O3 = 2.947(5) Å for 3d) and additional π–π interactions involving this chemical group are observed. Finally, in the presence of a bromine atom—as in compounds 2e (Fig. S56) and 3c (Fig. 5)—both the formation of pseudo-dimers via hydrogen bonding (e.g., N1–H1N⋯O3 = 2.906(5) Å for 3c) and Br⋯O short contacts (3.286 Å) are observed. Biological evaluation Antioxidant activity The antioxidant potential of the α-aminophosphonate derivatives synthesised in this study was evaluated using two complementary assays, namely, the ABTS radical scavenging assay and the phenanthroline-based Fe2+ reducing power assay. The results are summarised in Table 1 and reveal variable antioxidant behaviour depending on the molecular scaffold and substituent pattern. Table 1Antioxidant activitiesCompoundsAntioxidant assaysABTS (IC50 (mM))Phenanthroline (A 0.5 (mM))1a>0.8>0.21b>0.8>0.21c>0.8>0.21d>0.8>0.21e<0.1<0.0251f>0.8>0.21g>0.8<0.0251h<0.1>0.21iNA0.18 ± 0.012a>0.80.19 ± 0.032bNA>0.22cNA>0.22d<0.1<0.0252eNA>0.22fNA>0.22g<0.1>0.23a<0.1>0.23b>0.80.05 ± 0.013cNA>0.23d>0.8<0.0253eNA0.15 ± 0.023fNA<0.025 Antioxidant activities In the ABTS assay, only a limited number of compounds exhibited strong radical scavenging activities. Notably, compounds 1e, 1h, 2d, 2g, and 3a displayed excellent inhibitory potential, with IC50 values lower than 0.1 mM, indicating their high antioxidant efficiency. This enhanced activity was primarily associated with derivatives bearing electron-donating substituents such as –OH, –OMe groups. In contrast, most of the remaining compounds showed weak or negligible ABTS scavenging activity (IC50 > 0.8 mM or NA), suggesting that effective ABTS radical quenching is strongly dependent on specific structural features, particularly the presence and appropriate positioning of electron-donating groups capable of stabilising radical species. The phenanthroline assay revealed a broader antioxidant response across the compound library. Several derivatives demonstrated strong reducing power, as evidenced by their low absorbance values at 0.5 mM. Compounds 1e, 1g, 2d, 3d, and 3f exhibited particularly pronounced activity, with absorbance values lower than 0.025, indicating the efficient formation of the Fe2+-phenanthroline complex. In addition, moderate antioxidant activity was observed for compounds 1i, 2a, 3b, and 3e, with absorbance values ranging from 0.05 to 0.19, while the remaining compounds showed weaker responses. These results indicate that the electron-transfer capacity, rather than direct radical scavenging, predominantly governs the antioxidant behaviour observed in this assay. Overall, the phenanthroline assay identified a greater number of active compounds than the ABTS assay. The observed differences between the two methods highlight the contribution of distinct antioxidant mechanisms, with the substituent electronics playing a central role in determining the activity. While the ABTS assay primarily reflects the radical scavenging ability, the phenanthroline assay evaluates the reducing power and electron-donating capacity, which appear to be more prevalent among the synthesised α-aminophosphonate derivatives. Anti-inflammatory screening The in vitro anti-inflammatory activity of the synthesised α-aminophosphonate derivatives was evaluated using the bovine serum albumin (BSA) denaturation assay, and the results are summarised in Table 2. Data are expressed as mean ± SD (n = 3). Table 2 In vitro anti-inflammatory activity of the synthesized compoundsaSample (0.078 mM)Inhibition %Diclofenac98.70 ± 0.15***1a28.67 ± 0.411b46.14 ± 2.28***1cNA1d48.25 ± 3.83***1e40.58 ± 0.57***1f97.63 ± 0.991g20.85 ± 2.59***1h31.78 ± 0.70***1i70.85 ± 1.99***2a40.64 ± 3.81***2b41.60 ± 0.30***2c46.40 ± 1.63***2d16.81 ± 4.33***2e63.68 ± 0.99***2f39.27 ± 0.44***2g72.51 ± 4.66***3aNA3bNA3cNA3dNA3eNA3f83.04 ± 4.47**a**P < 0.01. ***P < 0.001 compared to the diclofenac group. NA: no activity. In vitro anti-inflammatory activity of the synthesized compoundsa **P < 0.01. ***P < 0.001 compared to the diclofenac group. NA: no activity. Among the tested compounds, compound 1f exhibited the highest inhibitory activity against protein denaturation (97.63% ± 0.99%), similar to the reference drug diclofenac (98.70% ± 0.15%), indicating its excellent anti-inflammatory potential. High levels of inhibition were also observed for compound 3f (83.04% ± 4.47%), 2g (72.51% ± 4.66%), 1i (70.85% ± 1.99%), and 2e (63.68% ± 0.99%). Several compounds showed moderate anti-inflammatory activity, including 1d (48.25% ± 3.83%), 2c (46.40% ± 1.63%), 1b (46.14% ± 2.28%), 2b (41.60% ± 0.30%), 2a (40.64% ± 3.81%), and 1e (40.58% ± 0.57%), indicating the partial inhibition of protein denaturation. Compounds 1a, 1g, and 1h exhibited weak activity, while most members of series 3 (3a–3e) were inactive (NA), suggesting that the anti-inflammatory efficacy is strongly influenced by structural features. Overall, the results demonstrate that several α-aminophosphonate derivatives possess significant in vitro anti-inflammatory activity, with 1f, 3f, 2g, and 1i emerging as the most promising candidates. The observed activity trends support the importance of appropriate aromatic substitution and scaffold selection in enhancing protein denaturation inhibition, consistent with the subsequent structure–activity relationship and molecular modelling analyses. In vivo toxicity assessment To assess the preliminary safety profile of the most efficient biologically active α-aminophosphonate derivatives in antioxidant and anti-inflammatory assays, their acute toxicity was evaluated using Tenebrio molitor larvae at concentrations ranging from 0.25 to 4 mM (Table 3). This in vivo model provides a rapid and informative screening tool for identifying potential toxic effects prior to more advanced pharmacological studies. Table 3Toxicity of each compound on Tenebrio molitor larvaeCompoundsMortality rate (%)4 mM1 mM0.5 mM0.25 mM1b00001d00001e1008060201f500001g00001i00002c33.3333.3316.6616.662d505033.3302e33.3333.33002g33.3316.6616.6603a33.3333.33003f33.3333.3300 Toxicity of each compound on Tenebrio molitor larvae Overall, most of the tested compounds exhibited low or negligible toxicity, particularly at concentrations close to or exceeding their effective anti-inflammatory and antioxidant doses. Among the evaluated derivatives, compound 1e showed the highest toxicity, inducing 100% mortality at 4 mM, with a clear dose-dependent decrease in toxicity observed upon dilution. At lower concentrations, the mortality rate was reduced to 20% at 0.25 mM, indicating that toxic effects occur primarily at high doses. Compound 1f, which demonstrated the strongest anti-inflammatory activity, displayed moderate toxicity (50%) only at the highest tested concentration, while no mortality was observed at lower doses. Similarly, compounds 2c, 2d, 2e, and 2g exhibited moderate toxicity at elevated concentrations, which diminished or disappeared at lower doses. In contrast, compounds 1b, 1d, 1g, and 1i showed no detectable toxicity across all tested concentrations, highlighting their favourable safety profiles. Importantly, at the lowest tested concentration (0.25 mM), which exceeds the IC50 and A0.5 values determined for the most active compounds (0.18–0.025 mM), most derivatives caused no mortality. For this reason, we can consider them as not toxic when considering effective doses. Based on these findings, the synthesized α-aminophosphonate derivatives exhibiting the highest biological activity demonstrated satisfactory safety profiles at therapeutically relevant concentrations. These results highlight their potential as promising candidates for the development of anti-inflammatory and antioxidant agents. Theoretical studies DFT study Frontier molecular orbital (FMO) analysis Recently, frontier molecular orbitals analysis has been reported as a crucial study to establish the chemical reactivity and interaction profile of bioactive compounds.25,26 The HOMO is associated with the electron-donating ability and corresponds to nucleophilic character, whereas the LUMO is correlated with the electron-accepting ability and electrophilic character. Therefore, a full evaluation of HOMO/LUMO energies provides essential insights into reaction pathways involving charge-transfer processes. The distribution and energies of the FMOs for compounds 1a–1i, 2a–2g, and 3a–3f, computed at the B3LYP/6-31G(d,p) level, are described in Table 4. The model molecule 1a is shown in Fig. 6, and the other molecules are shown in Fig. S57. Table 4Calculated global reactivity descriptors for compounds 1a–i, 2a–g and 3a–faN° η (eV) S (eV−1) µ (eV) χ (eV) ω Polarisability (Å3)Dipole moment N index 1a2.4639920.405845−3.2411493.2411492.1317130.3158876.902946−15.2251421b2.4577330.406878−3.1622363.1622362.0343410.3014585.649741−15.1399701c2.3758270.420906−3.3916283.3916282.4208710.3587367.274471−15.2874561d2.3740580.421219−3.1479503.1479502.0870570.30927013.55164−15.042001e2.4675290.405263−3.3366613.3366612.2559630.3342996.708407−15.3241911f2.4609990.406339−3.1529843.1529842.0197720.2992997.277978−15.1339831g2.2661650.441274−3.0582893.0582892.0636470.3058018.020159−14.8444541h2.5536530.391595−3.2837353.2837352.1112720.3128584.274552−15.3573891i2.4129700.414426−3.3147563.3147562.2767800.3373849.032849−15.2477272a2.3905210.418318−2.9856342.9856341.8644490.2762824.270039−14.896152b1.4011140.713717−2.7205952.7205952.6413390.3914067.699500−13.6417102c2.4438550.409189−3.1078133.1078131.9760790.2928242.058039−15.0716692d2.5215440.396582−2.8516182.8516181.6124500.2389402.914403−14.8931632e2.4658970.405531−3.2460473.2460472.1365090.3165983.530725−15.2319442f2.5046730.399253−3.0777453.0777451.8909680.2802121.898388−15.1024182g2.4528350.407691−3.1483583.1483582.0205520.2994142.447415−15.1211943a1.9154100.522081−3.1320323.1320322.5607110.3794584.187258−14.5674423b2.0789500.481011−3.3701313.3701312.7316150.4047835.119471−14.9690823c1.7294200.578228−3.4927193.4927193.5269290.5226365.257899−14.7421393d1.6326840.612488−3.2675443.2675443.2697230.4845226.819060−14.4202283e1.72887620.578410−3.4856443.4856443.5137600.5206856.777910−14.7345203f2.05881450.485716−3.4057783.4057782.8169920.4174355.140807−14.984593aChemical hardness (η) = (ELUMO – EHOMO)/2, molecular softness (S) = 1/η, electronic chemical potential (µ) = (EHOMO + ELUMO)/2, electronegativity (χ) = −µ, electrophilicity index (ω) = µ2/2η. Calculated global reactivity descriptors for compounds 1a–i, 2a–g and 3a–fa Chemical hardness (η) = (ELUMO – EHOMO)/2, molecular softness (S) = 1/η, electronic chemical potential (µ) = (EHOMO + ELUMO)/2, electronegativity (χ) = −µ, electrophilicity index (ω) = µ2/2η. Optimized structure of compound 1a, and the 3D representation of its frontier molecular orbits. For most derivatives, the HOMO is predominantly localized over the π-conjugated aromatic backbone, with significant electron density residing on the heteroatom-containing fragments (N, O, and S atoms). This indicates that these moieties act as electron-donating regions, playing a crucial role in oxidation processes and intermolecular charge transfer. In the 1st series of α-aminophosphonate compounds (1a–1i), the HOMO density is mainly delocalized across the aromatic ring system and adjacent heterocycles, suggesting efficient π-electron conjugation, while 2a–2g and 3a–3f α-aminophosphonate compounds exhibit enhanced HOMO delocalization due to extended conjugation and substituent effects, suggesting improved electronic communication within the molecular skeleton. The molecules exhibiting the least negative HOMO energies show enhanced nucleophilic reactivity, which is attributed to the presence of electron-donating substituents that increase the HOMO density over key conjugated segments. Conversely, the compounds exhibiting the most negative HOMO values imply stronger orbital stabilization and reduced ease of electron donation. These molecules are expected to be less reactive in pathways requiring electron transfer to electrophiles. In contrast, the LUMO is predominantly localized on electron-withdrawing moieties, such as carbonyl groups, heteroatoms or substituted aromatic rings. This localization highlights these regions as the preferred sites for nucleophilic attack or electron acceptance. Several derivatives (2b, 3c, and 3e) show strong LUMO localization on substituted heterocyclic fragments, indicating their enhanced electron-acceptor character. The calculated HOMO–LUMO gaps range approximately from ∼2.8 to ∼5.5 eV, depending on the nature and position of the substituents. Compounds with smaller ΔEgap values (e.g., 2b, 3d, and 3c) exhibit higher electronic softness and are expected to show enhanced chemical reactivity, lower excitation energies, and improved charge-transfer properties. Larger gaps, observed in some 1st series derivatives (1a–1i), indicate greater kinetic stability and lower polarizability. These observations reveal a consistent electronic trend across the three series of compounds, where structural variations introduced by substituents directly modulate FMO energies and result in reactivity differences. The calculated global descriptors, such as chemical potential (µ), electronegativity (χ), electrophilicity index (ω), polarizability (α), hardness (η), and dipole moment (µ-dip), reveal a coherent electronic basis for the observed structure–activity relationships. All studied compounds present negative chemical potential values (µ = −2.72 to −3.49 eV), confirming their stability and their ability to accept electron density from nucleophilic biological environments. Notably, the most active compounds display more negative µ values and higher electronegativity (χ ≈ 3.15–3.49 eV) compared to the weakly active analogues. Compound 3f, which shows high anti-inflammatory inhibition (83.04%), exhibits µ = −3.41 eV and χ = 3.41 eV, while the weakly active compound 2d (16.81%) presents a lower χ value of 2.85 eV, indicating reduced electron-accepting capability. The electrophilicity index (ω) emerges as a decisive descriptor correlated with biological activity. Highly active derivatives such as 1f (97.63%), 1i (70.85%), 2g (72.51%), and 3f (83.04%) exhibit ω values in the range of 2.27–2.82 eV, reflecting a strong propensity to stabilize charge-transfer interactions with biological targets. In contrast, compound 2d, one of the least active derivatives, displays a significantly lower ω value of 1.61 eV, supporting its diminished binding efficiency. These results indicate that the enhanced electrophilic character supports favourable ligand-target electronic interactions relevant to anti-inflammatory activity. Chemical hardness (η) values are generally moderate across the derivatives (η ≈ 2.24–2.55 eV), suggesting a balance between electronic stability and reactivity. However, extreme softness does not guarantee higher activity. For example, compound 2b, characterized by very low hardness (η = 1.40 eV) and high softness (S = 0.71 eV−1), exhibits only moderate inhibition (41.60%). This indicates that excessive electronic flexibility may reduce selectivity or lead to suboptimal interaction geometries. Thus, η and S act as secondary descriptors that modulate, rather than control, biological activity. Polarizability (α) shows a clearer positive correlation with anti-inflammatory potency. The most active compounds, such as 1f (α = 0.299 Å3), 1i (α = 0.337). Å3), and 3f (α = 0.417). Å3), possess higher polarizability values, favouring dispersion forces and induced-fit stabilization within hydrophobic or aromatic-rich binding pockets. Conversely, the low-activity compound 2d exhibits the lowest polarizability (α = 0.239 Å3), limiting its ability to engage in stabilizing non-covalent interactions. Dipole moment values range widely (µ-dip = 1.90–13.55 D), and their influence on activity is non-linear. Compounds with moderate dipole moments (approximately 6–9 D) display superior activity, as exemplified by 1f (7.28 D) and 3f (5.14 D). In contrast, compound 1d, despite having the highest dipole moment (13.55 D), shows only moderate inhibition (48.25%), suggesting that excessive polarity may impose desolvation or permeability penalties. Overall, the DFT results demonstrate that the anti-inflammatory activity of α-aminophosphonate derivatives is governed by a synergistic balance of high electrophilicity (ω), increased electronegativity (χ), and enhanced polarizability (α), while hardness and dipole moment play secondary, fine-tuning roles. These findings provide a robust electronic rationale for the observed experimental trends and support further structural optimization of this scaffold for anti-inflammatory drug development. Molecular electrostatic potential To further support the DFT-derived structure–activity relationships, molecular electrostatic potential (MEP) maps were generated for all synthesized α-aminophosphonate derivatives and are provided in the SI (Fig. S58). The 3D map of the molecular electrostatic potential of the model molecule 1a is shown in Fig. 7. The MEP surfaces provide a three-dimensional visualization of charge distribution, highlighting electron-rich (negative potential, red/yellow) and electron-deficient (positive potential, blue) regions that are directly relevant to molecular recognition and binding with inflammatory targets. 3D map of the molecular electrostatic potential of 1a. Across the series, the most negative electrostatic potential regions are consistently localized around the phosphoryl oxygen atoms (P Created by potrace 1.16, written by Peter Selinger 2001-2019 O), ester/carboxyl oxygen atoms, and heteroatom-containing substituents, confirming their role as primary hydrogen-bond acceptor sites. Conversely, the positive electrostatic potential regions are mainly centred on protonated amine (NH) moieties, identifying them as potential hydrogen-bond donors involved in ligand-target anchoring. A clear qualitative correlation emerges between the MEP features and anti-inflammatory activity. The highly active compounds such as 1f (97.63%), 1i (70.85%), 2g (72.51%), and 3f (83.04%) display well-defined and spatially extended negative potential regions distributed over both the phosphonate group and the aromatic/heteroaromatic substituents. This balanced charge distribution facilitates simultaneous hydrogen bonding, electrostatic interactions, and dispersion contacts, consistent with their high electrophilicity (ω ≈ 2.27–2.82 eV) and polarizability values obtained from DFT calculations. In contrast, weakly active compounds, such as 2d (16.81%), exhibit more localized and less intense negative potential regions, primarily confined to the phosphonate group, with limited charge delocalization over the aromatic framework. This reduced electrostatic complementarity aligns with its lower electrophilicity (ω = 1.61 eV) and minimal polarizability (α = 0.239 Å3), explaining its diminished biological performance. Molecular docking Given the pivotal role of cyclooxygenase-2 (COX-2) in the arachidonic acid cascade and its upregulation during inflammatory and pain-related processes, the present study aims to identify novel therapeutic agents capable of selectively modulating this enzyme while minimizing the gastrointestinal toxicity commonly associated with classical nonsteroidal anti-inflammatory drugs (NSAIDs). The crystal structure of human COX-2 co-crystallized with the selective inhibitor rofecoxib was selected as the molecular target as it provides a well-defined representation of the COX-2 selectivity side pocket, which is critical for high-affinity and selective ligand binding. Molecular docking studies were conducted to explore the binding interactions between the COX-2 active site and the newly synthesized α-aminophosphonate derivatives. By elucidating key intermolecular interactions, particularly hydrogen bonding with catalytic residues, and hydrophobic accommodation within the selectivity channel, this study seeks to clarify the molecular basis of enzyme inhibition and gain insights into the binding modes that underpin the observed anti-inflammatory activity, thereby supporting the therapeutic potential of these compounds as next-generation COX-2 inhibitors. To assess the accuracy and reliability of the employed docking protocol, a self-docking (re-docking) validation was performed using the native co-crystallized ligand (Table 5). Rofecoxib (RCX) was extracted from the human COX-2 crystal structure (PDB ID: 5KIR) and subsequently re-docked into the enzyme's active site. The superposition of the top-ranked docked pose with the corresponding crystallographic conformation resulted in a root mean square deviation (RMSD) of 0.8334 Å. This RMSD value is well within the generally accepted threshold (<2.0 Å), thereby confirming the robustness and predictive accuracy of the docking protocol and its suitability for reliably modeling the binding modes of the newly synthesized α-aminophosphonate derivatives (Fig. 8). Table 5Docking results of all compounds for the potential inhibition of the human COX-2 crystal structure (PDB ID: 5KIR)TitleScore energy (kcal mol−1)Residue interactionsH-bondHydrophobicπ–π stackingCo-ligand (RCX)−9.306PHE 518, ARG 513ILE 517, ALA 516, THR 94, GLN 192, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, TYR 348, VAL 349, LEU 352, SER 353, GLY 354, TYR 355, LN 192, THR 94, HIE 90, LEU 359, ARG 120, LEU 384, TYR 385, TRP 387—1a−8.063GLN 192, PHE 518, TYR 355 ARG 513, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, TYR 348, VAL 349, LEU 352, SER 353, TYR 355, LEU 359, VAL116, MET 113, ARG 120, HIE 90, TRP 387—1b−8.928GLN 192, PHE 518, TYR 355 ARG 513, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, LEU 534, VAL116, MET 522, ILE 517, ALA516, LEU 93, TRP 387, TYR 348, VAL 349, LEU 352, SER 353, ARG 120, LEU 354, HIE 90—1c−4.315GLN 192, LEU 352TYR 355, GLY 354, SER 353, VAL 349, TYR 348, MET 522, VAL 523, GLY526, ALA 527, SER 530, LEU 531, LEU 359, VAL 116, PHE 518, ILE 517, ALA 516, ARG 513, HIE 90, TRP 387, TYR 385, LEU 384, PHE 381—1d−5.865GLN 192ARG 513, ALA 516, ILE 517, PHE 518, LEU 534, LEU 531, SER 530, ALA 527, GLY 526, VAL 523, MET 522, PHE 387, LEU 384, TRP 387, TYR 348, VAL 349, LEU 352, SER 353, GLY 354, TYR 355, HIE 90—1e−6.605GLN 192PHE 518, ILE 517, ALA516, ARG513, PHE 381, LEU 384, TYR 385, MET522, VAL523, GLY523, ALA527, SER530, LEU 531, TYR 355, GLY354, SER 353, LEU 352, VAL349, TYR348, ARG120, LEUC 399, HIE 90TRP 3871f−6.120PHE 518, TYR 365ARG 513, ALA 516, ILE 517, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, TYR 348, VAL 349, LEU 352, SER 353, LEU 93, HIE 90, VAL 116, MET 113, TRP387, TYR 385—1g−5.007PHE 518ILE 517, ALA 516, ARG 513, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, ARG 120, VAL 116, MET 113, LEU 359, HIE 90, GLN 192, TRP 387, TYR 385, LEU 384, PHE 381—1h−8.950GLN 192, PHE 518, TYR 355, ILE 517 ARG 513, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, LEU 534, VAL116, MET 113, ALA516, LEU 384, TRP 387, TYR 348, VAL 349, LEU 352, SER 353, ARG 120, LEU 359, HIE 90—1i−9.742GLN 192, PHE 518, TYR 355 ARG 513, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, LEU 534, VAL116, MET 113, ILE 517, ALA516, LEU 384, TRP 387, TYR 348, VAL 349, LEU 352, SER 353, ARG 120, LEU 359, HIE 90—2a−5.142ARG 513, TYR 355VAL 523, GLY 526, ALA 527, SER 530, LEU 531, TYR 348, VAL 349, LEU 352, SER 353, GLY 354, PHE 518, ILE 517, ALA 516, TYR 385, TRP 387, PHE 381, LEU 359, ARG 120, THR 94, LEU 93, HIE 90, GLN 192—2b−6.473ARG 120ARG 513, ALA 516, ILE 517, PHE 518, SER 353, LEU 352, VAL349, LEU 534, LEU 531, SER 530, ALA 527, GLY 526, VAL 523, MET 522, TRP 387, VAL 89, HIE 90, LEU 93, GLN192, LEU 359, VAL 116TYR 3552c−4.520ARG 513, ARG 120PHE 518, ILE 517, ALA516, MET522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, TYR 348, VAL 349, LEU 352, SER 353, GLY 354, PHE 381, TYR 385, TRP 387, LEU 359, VALM 116, THR 94, GLN192, HIE 90TYR 3552d−6.094ARG 120, GLN 192ARG 513, ALA 516, ILE 517, PHE 518, ILE 112, MET 113, VAL 116, LEU 359, ¨PHE 357, SER 353, LEU 352, VAL 349, TYR 348, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, TRP 387, LEU 384, HIE 90, LEU 93TYR3552e−3.691—ARG 513, ALA 516, ILE 517, PHE 518, TYR 355, GLY 354, SER 353, LEU 352, VAL 349, TRY 348, ILE 345, MET 113, VAL 116, LEU 359, ARG 120, MET 522, VAL 523, GLY 526, AMA 527, SER 530, LEU 531, TYR 385, TRP 387, PHE 381, GLN 192, THR 94, HIE 90—2f−5.300PHE 518, TYR 355, ILE 517ARG 513, PRO514, ALA 516, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, LEU 534, VAL 116, MET 113, GLY 354, SER 353, LEU 352, VAL349, GLN 192, HIE 90, ARG 120, LEU 359, THR 94, TRP 387—2g−3.803—PHE 518, ILE 517, ALA 516, ASP 515, ARG 513, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, TYR 348, VAL 349, LEU 352, SER 353, GLY 354, VAL 116, LEU 359, PHE 381, TYR 385, TRP 387, VAL 344, HIE 90, GLN 192, THR 94, ARG 120TYR 3553a−5.047ARG 120ARG 513, ALA 516, ILE 517, PHE 518, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, LEU 534, VAL 349, LEU 352, SER 353, GLY354, LEU 359, VAL 116, HIE 90, LEU 93, THR 94, GLN 192, TRP 387TYR 3553b−8.058GLN 192, PHE 518, TYR 355ARG 513, ALA 516, ILE517, MET522, VAL523, GLY 526, ALA527, SER 530, LEU 531, TYR348, VAL349, LEU 352, TYR355, ARG 120, MET 113, VAL 116, TRP, HIE 90—3c−2.289ARG 120PHE 518, ILE 517, ALA 516, ARG 513, MET 522, VAL 523, GLY 526, ALA 527, SER 510, LEU 531, LEU 359, TYR 355, SER 353, LEU 352, VAL 349, TYR 348, TYR 385, TRP 387, PHE 381, VAL 116, LEU 93, HIE 90, VAL 89, GLN 192ARG 1203d−3.811—LEU 98, HIE 90, VAL 89, TYR 355, GLY 354, SER 353, LEU 352, VAL 349, TYR 348, MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, GLN 192, ARG 513, ALA 516, ILE 517, PHE 518, TRP 387, TYR 385, LEU 384, PHE 381, VAL 344, PHE 205ARG 1203e−4.506ARG 513MET 522, VAL 523, GLY 526, ALA 527, SER 530, LEU 531, TYR348, VAL 349, LEU 352, SER 353, GLY354, PHE 518, ILE 517, ALA 516, GLN 192, TRP 387, TYR 385, PHE 381, ARG 120, THR 94, LEU 93, HIE 90TYR 3553f−6.226ILE 517, PHE 518GLY 354, SER 353, LEU 352, VAL 349, TYR 348, LEU 351, SER 510, ALA 527, GLY 526, VAL 523, MET 522, HIE 90, GLN 192, THR 94, LEU 384, TYR 385, TRP 387, ARG 120, ARG 513, ALA 516TYR 355 Docking results of all compounds for the potential inhibition of the human COX-2 crystal structure (PDB ID: 5KIR) Self-docked co-crystallized ligand (ID: RCX) of the urease enzyme (ID: 5KIR). The biological evaluation identified compounds 1f (97.63% ± 0.99%), 3f (83.04% ± 4.47%), 1i (70.85% ± 1.99%), 2g (72.51% ± 4.66%), and 2e (63.68% ± 0.99%) as the most promising lead candidates due to their strong interactions dominated by extensive hydrophobic contacts and hydrogen bonding with key residues. Compound 1i is identified as the superior ligand in terms of binding enthalpy, with a score of −9.742 kcal mol−1, surpassing the co-crystallized selective inhibitor rofecoxib (RCX) (−9.306 kcal mol−1). Mechanistically, 1i achieves this stability by mimicking the binding mode of classic coxibs. It forms a stabilizing hydrogen bond with Tyr355, Phe518, and Gln192. Crucially, the hydrophobic tail of 1i is positioned to engage Val523 and Arg513. The interaction with Val523 is the essential interaction for COX-2 selectivity, confirming that 1i successfully occupies the selectivity side-pocket, which is sterically restricted in the COX-1 isoform. Compounds displaying moderate anti-inflammatory activity (e.g., 1b, 1d, 1e, 2a, 2b, and 2c) showed acceptable docking scores but fewer hydrogen bonds or suboptimal orientation within the binding pocket. This suggests that binding energy alone is insufficient to fully explain biological activity, and the interaction geometry plays a critical role. In contrast, the weakly active or inactive compounds, including 1a, 1c, 1g, and most members of the 3rd series (3a–3e), exhibited less favourable interaction profiles despite some of them having moderate docking scores (Fig. S59–S75). This discrepancy may be attributed to the absence of strong hydrogen bonding, limited π–π interactions, or unfavorable physicochemical properties, affecting the biological performance. The reference drug diclofenac showed excellent in vitro inhibition (98.70%), validating the experimental model. Notably, compound 1f demonstrated comparable activity, highlighting its potential as a promising lead compound. However, the observed inconsistencies emphasize the limitations of static docking approaches and suggest that further molecular dynamics simulations and ADME studies are required to better understand ligand behaviour in biological systems (Fig. 9 and 10). 2D and 3D representations of the best pose interactions of compound 1f. 2D and 3D representations of the best pose interactions of compound 1i. Molecular dynamics simulations To complement the static docking analysis and to evaluate the dynamic behaviour of the most relevant COX-2-ligand complexes under physiological-like conditions, molecular dynamics simulations were performed for selected compounds over 100 ns using the Desmond package. The simulations were conducted under NPT conditions at 300 K, and the stability of the complexes was assessed through protein-ligand RMSD, protein RMSF, ligand RMSF, secondary structure evolution, and protein–ligand contact analysis. Herein, the COX-2 structure was prepared from PDB ID: 5KIR, and the stability of the selected ligand-COX-2 complexes was evaluated using RMSD and interaction-profile analyses. Protein–ligand RMSD analysis The protein-ligand RMSD profiles were analysed to evaluate the stability of the selected COX-2 complexes during the 100 ns MD simulations. For all complexes, the protein Cα RMSD increased during the initial equilibration period and then remained within an acceptable range of approximately 1.6–2.4 Å, indicating that ligand binding did not induce major conformational destabilization of the COX-2 backbone. Compound 1f showed a stable protein RMSD profile after approximately 20 ns, with values mainly around 1.8–2.3 Å. Its ligand RMSD showed a marked increase during the early stage of the simulation and then stabilized at around 15–17 Å, suggesting that the ligand underwent significant reorientation from the initial docking pose before reaching a stable bound conformation. In contrast, compound 1i displayed lower ligand RMSD values, mostly within 0.9–2.4 Å, indicating a better preservation of the initial binding pose and stronger dynamic anchoring within the COX-2 active site. The behaviour of 1f suggests adaptive repositioning inside the binding pocket, while 1i exhibits more conserved binding modes during the simulation (Fig. 11). Protein-ligand RMSD profiles of the compounds 1i and 1f. Protein RMSF analysis Protein RMSF analysis was performed to evaluate the local flexibility of the COX-2 residues in the simulated complexes. For all systems, most residues displayed RMSF values below approximately 1.5 Å, indicating that the protein backbone remained locally stable during the 100 ns simulation. The highest fluctuation was observed in the region around residues 40–55, reaching approximately 3.2–3.3 Å for compound 1f and about 3.7–3.8 Å for compound 1i. This increased mobility is likely associated with a flexible loop region and does not indicate the global destabilization of the enzyme. Outside this region, the RMSF profiles remained relatively low, particularly in the ligand-contacting regions marked by the interaction bars (Fig. 12). Protein–ligand RMSF profiles of compounds 1i and 1f. Protein–ligand contact analysis Protein-ligand contact analysis was carried out to identify the residues responsible for stabilizing the selected compounds within the COX-2 binding pocket. Compound 1f displayed a diversified interaction network involving hydrogen bonds, hydrophobic contacts and water bridges. The most persistent interaction was observed with Cys47, with a total interaction fraction of higher than 1.0, indicating the presence of multiple simultaneous contacts during the trajectory. Additional important interactions were observed with Met48 and Ser49, with interaction fractions of approximately 0.75 and 0.70–0.75, respectively. Hydrophobic contacts with Cys136, Pro155 and Pro156, together with water-mediated interactions involving Gly128 and Asp157, further contributed to the stabilization of the 1f-COX-2 complex. Compound 1i showed a different interaction pattern, dominated mainly by hydrophobic contacts with key COX-2 binding-site residues. In both complexes, Tyr 355 represented the most persistent contact, with an interaction fraction close to 0.9. Phe518 and Val523 also showed strong hydrophobic interactions, with a fraction of approximately 0.70. The interaction with Val523 is particularly relevant because this residue is associated with the COX-2 selectivity pocket, suggesting that 1i is well accommodated within the hydrophobic selectivity region. Pro514 contributed mainly through water-mediated interactions, while Arg513 showed mixed hydrogen-bond and water-bridge contributions. Overall, the contact analysis indicates that 1f is stabilized mainly through a polar hydrogen-bond/water-bridge network, whereas 1i is stabilized predominantly by hydrophobic contacts within the COX-2 binding pocket. These results explain the RMSD and RMSF observations: 1i maintains a more conserved binding pose, while 1f undergoes conformational adaptation but remains stabilized by persistent polar contacts. Thus, the MD interaction profiles support the potential of these compounds as COX-2-targeted anti-inflammatory candidates (Fig. 13). Protein–ligand contacts of the compounds 1i and 1f. Among the two complexes, compound 1f showed the lowest maximum RMSF value, suggesting a slightly better local stabilization of the protein structure. Compound 1i exhibited similar RMSF patterns, with higher flexibility in the early residue region but stable behaviour across most of the protein chain. Importantly, the residues involved in ligand interactions were generally located in low-to-moderately fluctuating regions, supporting the stability of the ligand-binding environment. Overall, the RMSF analysis confirms that the simulated ligands do not induce major local destabilization of COX-2, and the observed flexibility is mainly restricted to the loop regions, while the binding-site architecture remains preserved throughout the simulation. Ligand–protein interactions Ligand–protein contact analysis revealed distinct stabilization patterns among the studied complexes. Compound 1f was mainly stabilized through polar and water-mediated interactions. The most persistent contact was observed with His39, with an interaction frequency of approximately 75%, followed by Ser49 at 54% and Cys47 at 50%. In addition, a water-mediated interaction involving the phosphonate region was maintained for approximately 55% of the trajectory. These results indicate that the phosphonate, amine and carboxylic acid functions of 1f contribute strongly to ligand anchoring through hydrogen bonding and water-bridge interactions. Hydrophobic contacts with Cys36 and Pro156 further support the stabilization of the aromatic scaffold. In contrast, compounds 1h and 1i showed a binding profile dominated by aromatic and hydrophobic interactions. Both ligands formed π–π stacking interactions with Tyr355 and Phe518, with contact frequencies of approximately 55% and 32%, respectively. The carboxylic acid group was further stabilized through a water-mediated interaction with Pro514, which was maintained for approximately 46% of the simulation. The presence of Val523 as a hydrophobic contact is particularly important because this residue is associated with the COX-2 selectivity pocket, suggesting favourable accommodation of these ligands in the hydrophobic binding channel. Overall, compound 1i is stabilized mainly by π–π stacking interactions and hydrophobic contacts, whereas compound 1f is stabilized predominantly by polar and water-mediated interactions. This difference explains the RMSD behaviour: 1i maintains a more conserved binding pose, while 1f undergoes conformational adaptation but remains strongly anchored through persistent polar contacts. Ligand torsion profiles Ligand torsion profiles were analysed to investigate the conformational behaviour of the selected compounds during the MD simulations. Compound 1f exhibited several preferred torsional states, but some dihedral angles showed broad and multimodal distributions, especially those associated with the phosphonate ethoxy groups and peripheral substituents. This indicates that 1f undergoes conformational adaptation within the binding pocket. Nevertheless, the central aminophosphonate-aromatic scaffold remained relatively organized around preferred angular regions, suggesting that the ligand preserves a stable pharmacophoric arrangement while allowing flexibility at terminal groups. In contrast, compound 1i showed more restricted torsion profiles. Most torsion angles were concentrated around defined conformational regions, mainly near −180°, 0°, +90° and +180°, indicating reduced internal rotation during the simulation. This behaviour can be attributed to the presence of the fused aromatic/indole-like scaffold, which decreases conformational freedom and improves ligand rigidity. The remaining flexibility was mainly localized at the terminal phosphonate ethoxy chains, which is expected for solvent-exposed substituents (Fig. 14). Protein–ligand Interactions of compounds 1i and 1f. Overall, the torsion profile analysis supports the RMSD, RMSF and contact results. Compound 1f displays greater torsional flexibility, allowing adaptive fitting and persistent polar/water-mediated interactions within the binding region. Compound 1i shows higher torsional stability, which explains its lower ligand RMSD values and its conserved binding modes stabilized by π–π stacking and hydrophobic contacts. Thus, the torsional behaviour confirms that the active ligands are stabilized through two complementary mechanisms: adaptive flexibility for 1f and conformational rigidity for 1i (Fig. 15). Torsional behavior of compounds 1i and 1f. Ligand properties Ligand-property analysis was performed to evaluate the conformational stability, compactness, solvent exposure and polar surface behaviour of the selected compounds during the MD simulation. The ligand RMSD profile showed that compound 1f underwent an initial conformational adjustment, followed by stabilization mainly around 1.0–1.4 Å, with short peaks reaching approximately 1.6–1.8 Å. In contrast, compounds 1h and 1i exhibited lower RMSD values, mostly between 0.5 and 1.0 Å, with occasional peaks close to 1.2 Å, indicating a better preservation of ligand conformation during the simulation. This confirms that 1h and 1i maintain more rigid and stable ligand poses, while 1f displays moderate adaptive flexibility. The radius of gyration remained stable for all ligands, with values around 4.35–4.60 Å, indicating that the compounds preserved their compactness and did not undergo major conformational expansion or collapse. No intramolecular hydrogen bonds were detected for any ligand, suggesting that ligand stability is mainly governed by intermolecular contacts with the protein and water molecules rather than by internal hydrogen-bond stabilization. The molecular surface area remained relatively constant, at approximately 352–370 Å2 for the three ligands, confirming the absence of major structural distortions during the simulation. More pronounced differences were observed in the solvent-accessible surface area. Compound 1f showed a marked decrease in SASA after approximately 20 ns, from around 350–430 Å2 to approximately 150–220 Å2, suggesting that the ligand became more buried in the protein environment after equilibration. In contrast, compound 1i displayed much lower SASA values, mostly at around 10–60 Å2, indicating that they remained deeply accommodated within the binding pocket throughout the trajectory. The polar surface area remained relatively stable for all systems, with values of approximately 145–160 Å2 for 1f and 155–170 Å2 for 1i. These values indicate that the polar groups of the ligands remain available for intermolecular interactions. Overall, the ligand-property analysis supports the previous RMSD, RMSF, torsion and contact results: 1f behaves as a moderately flexible ligand that becomes progressively stabilized and buried in the protein environment, while 1i shows greater conformational rigidity, lower solvent exposure and a more conserved binding behaviour (Fig. 16). The molecular surface area of compounds 1i and 1f. Ligand-based drug-likeness and ADMET analysis The physicochemical properties of the synthesized compounds (Table 6) were analyzed in relation to their in vitro anti-inflammatory activity evaluated at 0.078 mM (Table 2) to establish preliminary structure–property–activity relationships. Table 6Physicochemical properties of the synthesised small moleculesFormulaMWNHBDNHBANumber of rotatable bondsTPSAMR1aC18H22NO5P363.3425994.6797.261bC18H22NO6P379.34369114.9099.981cC18H21BrNO5P442.2425994.67104.961dC20H26NO7P423.402711113.13110.241eC19H24NO7P409.373710124.13105.771fC19H24NO6P393.372610103.90103.751gC20H27N2O5P406.41251097.91111.461hC18H22NO6P379.34369114.9099.281iC20H23N2O5P402.38359109.11110.462aC20H26NO6P407.402611103.90108.412bC22H30NO7P451.451713102.13119.372cC21H28NO6P421.42161292.90112.872dC21H28NO7P437.422712113.13114.902eC20H25BrNO5P470.291511114.0883.672fC20H26NO6P407.402611103.90108.412gC20H26NO5P391.40151183.67106.383aC21H28NO6P421.42161292.90112.873bC20H26NO5P391.40151183.67106.383cC20H25BrNO5P470.29151183.67114.083dC21H28NO5P405.42151183.67111.353eC20H25ClNO5P425.84151183.67111.393fC20H25FNO5P409.39161183.67106.34 Physicochemical properties of the synthesised small molecules All compounds fall within an acceptable drug-like chemical space, with molecular weights below 500 g mol−1, hydrogen bond donors between 1 and 3, hydrogen bond acceptors between 5 and 7, and TPSA values ranging from 83.67 to 124.13 Å2. Consequently, the observed differences in biological activity are not attributable to major violations of drug-likeness rules but rather to subtle variations in polarity, molecular flexibility, and interaction potential. The most active derivatives (1f, 3f, 2g, and 1i) exhibit inhibition percentages above 70%, with compound 1f showing activity comparable to that of the reference drug diclofenac. These compounds share a balanced physicochemical profile characterized by a moderate molecular weight (≈390–410 g mol−1), controlled polarity, and a limited number of rotatable bonds. 1f exhibits a TPSA of 103.90 Å2 with only 10 rotatable bonds, suggesting reduced conformational entropy penalties and a favourable binding efficiency. Compounds 3f and 2g, which display the lowest TPSA values in the series (83.67 Å2), also show high inhibitory activities, indicating that reduced polar surface areas may enhance membrane permeability and decrease desolvation costs during target interactions. Similarly, compound 1i maintains high activity despite a slightly higher TPSA, likely due to its low molecular flexibility (9 rotatable bonds). In contrast, compounds with moderate activity generally exhibit higher TPSA values (>110 Å2) and/or increased molecular flexibility (≥11 rotatable bonds), which can negatively impact binding affinity through excessive solvation and increased entropic penalties. This effect is particularly evident for compound 2d, which combines high polarity and flexibility and shows the lowest inhibition. Another study for the ligand-based drug-likeness of the synthesised α-aminophosphonate compounds was realized using bioavailability radar plots (Fig. 17 and S76), which provide a visual assessment of crucial physicochemical properties, including lipophilicity (LIPO), molecular size (SIZE), polarity (POLAR), solubility (INSOLU), saturation (INSATU), and molecular flexibility (FLEX). The optimal physicochemical space for oral drugs is represented by the shaded area in each radar plot. As shown in Fig. 17, most compounds across the three series exhibit well-balanced physicochemical profiles, with most parameters falling within or close to the optimal region. Lipophilicity and molecular size are generally well controlled, indicating favourable membrane permeability without excessive hydrophobicity. The polarity values are also appropriately maintained, supporting a balance between solubility and permeability. Optimal physicochemical space of compounds 1f and 1i. The 1st and the 2nd series of α-aminophosphonate compounds (1a–1i) and (2a–2g) display particularly consistent profiles, with moderate flexibility and acceptable degrees of saturation, suggesting good conformational adaptability while avoiding excessive rotatable bonds. Minor deviations observed for some derivatives, mainly in the flexibility or polarity axes, are not pronounced and are unlikely to severely compromise their oral bioavailability. Notably, the most biologically active compounds (e.g.1f, 1i, 2g, and 3f) present radar profiles that closely match the optimal drug-like space, supporting the experimental findings and reinforcing their potential as orally viable candidates. Overall, the bioavailability radar analysis confirmed that the α-aminophosphonate scaffold provides a favourable balance of physicochemical properties, consistent with good oral drug-likeness. These results complement the Lipinski-based evaluation and support further optimization guided by ADMET analysis. In silico ADMET evaluation was performed to complement the in vitro anti-inflammatory results and support lead identification (Table S7). All compounds were predicted to exhibit acceptable human intestinal absorption (HIA), indicating favorable oral bioavailability. Although the Caco-2 permeability values were moderate, none of the compounds were predicted to be P-glycoprotein substrates, suggesting a limited risk of efflux-related absorption loss. Distribution analysis revealed that series 1 compounds were generally non-blood–brain barrier (BBB) permeable, while most derivatives from series 2 and 3 were predicted to cross the BBB, consistent with their lower TPSA values. The predicted volumes of distribution indicate moderate to high tissue penetration, particularly for the most active compounds. Metabolic predictions showed limited involvement of CYP2D6 for most compounds, while several derivatives were identified as substrates or inhibitors of CYP3A4, a common feature among drug-like molecules. Inhibition of CYP2C9 and CYP1A2 was observed for some compounds, in agreement with the frequently reported profiles for anti-inflammatory agents. From a toxicity perspective, all compounds were predicted to be non-mutagenic (AMES-negative) and exhibited acceptable acute oral toxicity. Notably, the hERG inhibition liability was mainly associated with series 2 and 3, while most series 1 compounds, including the most active derivative 1f, were predicted to be hERG-safe. Integrating the in vitro anti-inflammatory activity with the ADMET predictions identified compound 1f as the most promising lead candidate. This compound combines near-reference biological activity with favorable pharmacokinetic predictions, absence of hERG liability, and a limited risk of CYP-mediated metabolic interactions. Compounds 3f and 2g, despite their high inhibitory potency, present potential cardiotoxicity and metabolic liabilities, suggesting that further structural optimization is required to improve their safety profiles. Overall, these findings highlight 1f as a balanced lead for further preclinical development while providing a clear framework for the rational optimization of other active derivatives within the series. Experimental section (general procedures and materials) Chemistry: synthesis of α-aminophosphonate derivatives A three-component reaction was carried out among an amine (0.01 mol), an aldehyde (0.01 mol) and triethyl phosphite (0.01 mol). The mixture was dissolved in 30 mL of ethanol and heated at 60 °C with continuous stirring for 4 hours; the completion of the reaction was monitored by TLC in n-hexane/methanol. The solvent was evaporated under low pressure at a low temperature, and α-aminophosphonate derivatives were obtained in good to excellent yields. Biological evaluation Antioxidant test Chemicals 2,2′-Azinobis-3-ethylbenzothiazoline-6-sulfonic acid (ABTS), potassium persulfate (K2S2O8), iron chloride III (FeCl3), phenanthroline, methanol, bovine serum albumin (BSA) and Tris–HCl were purchased from Sigma-Aldrich. ABTS radical (ABTS˙+) inhibition process The modified method of Re et al.27 was used to measure the ABTS˙+ inhibiting capacity spectrophotometrically. ABTS˙+ was initially prepared by mixing 2.45 mM K2S2O8 and 2 mM ABTS, and then, the mixture was left undisturbed at room temperature for 16 hours. Thenceforth, 40 µL of the compound solution was mixed with 160 µL of diluted ABTS˙+ solution to produce an absorbance of 0.700 ± 0.025 at 734 nm. The microplate was then incubated for 10 more minutes before measuring the absorbance at 734 nm. The ABTS radical's inhibition was calculated using eqn (1), and the results were reported as IC50 (inhibition concentration at 50%) values.1I% = ((Ac − As)/Ac) × 100 (*)Ac: absorbance of the control (methanol).Ap: absorbance of the product. o-Phenanthroline assay The reaction mixture, as described by Szydlowska-Czerniaka et al.,28 contained 30 µL of o-phenanthroline (0.5% in methanol), 50 µL of FeCl3, 110 µL of methanol, and 10 µL of distilled water. 10 µL of the tested compound solution was added to this mixture. The absorbance at 510 nm was then measured after incubating for 20 min at 30 °C. The outcome was provided as a value of A0.5 (the concentration giving an absorbance of 0.5). In vitro anti-inflammatory activity The in vitro anti-inflammatory capacity was evaluated using the method of Karthik et al.29 with slight changes. Briefly, 100 µL of 78 mM dose of the synthesized compounds or 78 mM diclofenac was added to 1 mL of a solution of 0.2% bovine serum albumin (BSA) prepared in Tris–HCl (pH: 6.6), and the solutions were kept for 15 min in an oven at 37 °C. After that, the mixture was placed in a water bath for five minutes at 72 °C. Then, the cooling turbidity was determined at 660 nm using a cuvette spectrophotometer for each compound. A white blank was prepared in 1 mL of extract and 1 mL of Tris–HCl. Toxicity on Tenebrio molitor larvae Mealworms were selected because they are readily available, hardy and relatively small. The experimental procedure was carried out by injecting a volume of 2 µl of the compound solution at different concentrations (4, 2, 1, 0.5 mM) into the abdominal part using a GC-syringe (0–5 µm), and six larvae were tested for each concentration of the same compound. Injections were made caudally into the ventral side of the larvae, lateral to the midline, as this avoids the most essential organ systems. The larvae were then incubated at room temperature with ventilation for 5 to 7 days. Mortality was designated by the discoloration of the larvae; this experiment was conducted according to a previously reported method30–32 with minor changes. The toxicity effect on the worms was monitored for ten days according to the following equation:Mortality % = (number of dead larvae/6) × 100. Computational details Computational studies were performed to gain deeper insights into the physicochemical properties, reactivity, and binding behaviour of the synthesised α-aminophosphonate derivatives and support the experimental biological findings. An integrated in silico workflow combining density functional theory (DFT) calculations, molecular docking, molecular dynamics (MD) simulations, and ligand-based ADME/Toxicity predictions was employed. This approach enabled the correlation among molecular structure, reactivity, selectivity, and biological activity, evaluation of ligand–protein interactions and preliminary assessment of drug-development potential. Ligand preparation All three-dimensional structures of the synthesised α-aminophosphonate derivatives were constructed using ChemDraw 3D (version 2019) and saved in the MDL mol format. Initial geometry optimisation was carried out using the AM1 semi-empirical method implemented in Gaussian 09. The final geometry optimisation was then performed using DFT at the B3LYP/6-31G(d,p) level, yielding the most stable conformations. These optimised geometries were used for the calculation of global reactivity descriptors. Molecular electrostatic potential (MEP) maps were generated using GaussView 05 to visualise the charge distribution and identify potential reactive regions. The optimised ligands were subsequently imported into Maestro (Schrödinger Release 2021-3) and prepared using the LigPrep module. Energy minimisation was performed using the OPLS3 force field, and all relevant ionisation states, tautomers, ring conformations, and stereoisomers were generated (up to 32 per ligand). The lowest-energy conformation of each compound was selected for further studies. Protein preparation Molecular docking and MD simulations were performed using the crystal structure of human cyclooxygenase-2 (COX-2) retrieved from the Protein Data Bank (PDB ID: 5KIR). Protein preparation was conducted using the Protein Preparation Wizard in the Schrödinger Suite. The structure was processed by adding hydrogen atoms, assigning appropriate protonation states, removing crystallographic water molecules, and correcting missing residues using the Prime module. The prepared protein structure was subsequently energy-minimised prior to docking calculations. Molecular docking Molecular docking simulations were carried out using the Glide module within Schrödinger Release 2021-3 to predict the binding modes and interaction patterns of the α-aminophosphonate derivatives within the COX-2 active site. Docking was performed in the extra precision (XP) mode. The docking protocol was validated by re-docking the co-crystallised ligand, and the resulting RMSD value confirmed the reliability of the method. A receptor grid was generated using the Receptor Grid Generation module, centred on the active site of COX-2. Docked poses were ranked according to the Glide score, and the top-ranked conformations were selected for interaction analysis. Molecular dynamics simulations To evaluate the stability of the selected ligand-COX-2 complexes, 100 ns molecular dynamics simulations were performed using the Desmond package. Each complex was solvated in an SPC water model within a cubic box extending 10 Å from the solute. Simulations were conducted under NPT conditions at 300 K and 1.01 bar using the OPLS3 force field. System stability was assessed by monitoring the root-mean-square deviation (RMSD) values of the protein backbone and ligand throughout the simulation. Trajectory analyses and interaction profiles were generated using the Simulation Interaction Diagram tool. Ligand-based drug-likeness and ADME/Toxicity prediction The ligand-based drug-likeness and pharmacokinetic properties were evaluated to assess the suitability of the α-aminophosphonate derivatives as potential drug candidates. Key physicochemical descriptors were calculated in accordance with the Lipinski's rule of five using Swiss ADME. Additional ADME and toxicity parameters were predicted using the pkCSM server, while biological activity and pathway predictions were explored using the PASS (Way2Drug) platform. Potential molecular targets were investigated using Swiss Target Prediction. Global reactivity descriptors Global reactivity descriptors were calculated to evaluate the chemical reactivity and kinetic stability of the investigated compounds. These parameters were derived using the frontier molecular orbital (FMO) theory based on the energies of the highest occupied molecular orbital (EHOMO) and lowest unoccupied molecular orbital (ELUMO) obtained from DFT calculations. Descriptors including the energy gap (ΔE), chemical hardness (η), chemical softness (σ), chemical potential (µ), electronegativity (χ), ionisation energy (I), electron affinity (A), electrophilicity index (ω), and nucleophilicity index (ξ) were calculated using standard equations. These descriptors provide insights into charge-transfer capability, molecular stability, and interaction propensity with biological targets. X-ray structure analysis Single crystals suitable for X-ray diffraction studies were obtained via the slow evaporation of ethanol solution. The single crystals of the compounds 1i, 2g, 2e, 2b, 3b, 3d, 1a, 3c, 3a and 1d were mounted on a D8 VENTURE Bruker-AXS diffractometer (MoKα radiation source, λ = 0.71073 Å) at the Centre de Diffractométrie X (CDIFX), Université de Rennes, France, for data collection. Structures were solved via a direct method using the SHELXT program33 and refined with a full matrix least-squares method on F2 using the SHELXL-14/7 program.34 Crystallographic data are summarized in Tables S2 and S3. Complete crystal structure results (CCDC 2539916–2539924 and 2539973 for 1i, 2g, 2e, 2b, 3b, 3d, 1a, 3c, 3a and 1d), including bond lengths, angles, and atomic coordinates, are provided as CIF files in the SI. Conclusions In this work, a series of novel α-aminophosphonate derivatives of para-aminobenzoic acid and benzocaine were successively synthesized via an efficient Kabachnik–Fields reaction. The compounds were fully characterized by single-crystal X-ray diffraction and/or spectroscopies and evaluated for their antioxidant and anti-inflammatory activities. Several derivatives exhibited significant activity, with compounds 1f, 3f, 2g and 1i identified as the most promising derivatives, with 1f showing activity comparable to that of diclofenac. Toxicity studies confirmed low toxicity at tested concentrations. DFT calculations and molecular docking analyses provided mechanistic insights, highlighting favourable electronic properties and strong interactions with the COX-2 active site. ADMET predictions further supported their drug-likeness. Overall, these findings identify α-aminophosphonate derivatives as promising candidates for the development of new anti-inflammatory agents, warranting further in vivo investigation. Author contributions All authors have contributed significantly to this work. Z. D., Z. C., and H. D. performed the organic synthesis, crystallisation, characterisation. H. K. performed computational studies, including DFT and molecular docking. V. D. and F. P. performed the X-ray structure analysis. T. K. and L. K. performed the biological assays, including the antioxidant and anti-inflammatory evaluations. The manuscript was written by the primary author and reviewed and edited by all co-authors. Conflicts of interest The authors declare no conflicts of interest. Supplementary Material