Computational study prioritizes two benzohydrazide derivatives for EGFR testing

Six newly synthesized benzohydrazide derivatives produced encouraging results in a computational study of the epidermal growth factor receptor, with compounds 3d and 3f maintaining recurring contacts inside modeled ATP-binding sites during 200-nanosecond molecular dynamics simulations. The compounds are proposed as candidates for experimental validation, not established EGFR inhibitors: the research included synthesis, spectroscopy and a suite of in silico tools, but no cell-based or animal experiments and no direct measurements of binding affinity, kinase inhibition or anticancer activity. The series, designated 3a–f, was built around a bis-azomethine framework that differs from the simple mono-imine benzohydrazide and hydrazone structures often examined in anticancer studies.

Each molecule combines a 4-aminobenzohydrazide core with a nabumetone-derived hydrophobic segment containing a methoxynaphthyl group and a para-substituted aromatic Schiff-base extension. The six terminal substitutions were hydrogen, chlorine, nitro, hydroxy, methoxy and dimethylamino groups. This design was intended to vary electronic properties, polarity and molecular shape while examining whether the series could be recognized by more than one EGFR conformation.

Synthesis proceeded in three steps. Benzocaine was first converted to 4-aminobenzohydrazide, which was then condensed with nabumetone to form hydrazone 2. Reaction of that intermediate with six para-substituted aromatic aldehydes produced compounds 3a–f.

Their structures were confirmed by Fourier-transform infrared spectroscopy and proton and carbon-13 nuclear magnetic resonance spectroscopy. Diagnostic features included the disappearance of the terminal amino group of compound 2, signals from the new azomethine proton at 8.26–8.66 parts per million, and hydrazide N–H signals at 11.17–11.88 parts per million. Substituent-specific evidence, including the aryl chlorine band in 3b, nitro and phenolic signals in 3c and 3d, and the dimethylamino resonances in 3f, further supported formation of the intended products.

The isolated yields of the six final compounds ranged from 47% for the chloro derivative 3b to 81.3% for the methoxy derivative 3e. Predicted absorption, distribution, metabolism and excretion properties revealed a mixed profile. All six compounds were predicted to have high human oral absorption, acceptable molecular weights and hydrogen-bonding counts, and three to four predicted metabolic reactions, compared with six for erlotinib.

However, their calculated octanol–water partition coefficients exceeded the study’s preferred limit of 5 and were higher than erlotinib’s 3.67, indicating potential liabilities for low solubility and nonspecific binding. Compounds 3c–f also had 11 rotatable bonds, above the preferred limit of 10. Despite these shared concerns, 3e had the most balanced molecular size and shape, with a predicted molecular volume of 1,458.98 cubic ångströms and a globularity value of 0.84.

Predicted permeability estimates separated the derivatives more clearly. Compounds 3e and 3f had the highest estimated Caco-2 and MDCK permeability values, at 2,261.14 and 1,194.91 nanometers per second for 3e and 2,476.48 and 1,318.38 nanometers per second for 3f, respectively. The nitro-substituted 3c was weakest, with corresponding estimates of 179.59 and 77.33 nanometers per second, consistent with its relatively high predicted polar surface area of 120.47 square ångströms.

Predicted blood–brain barrier penetration was low across the series, with QPlogBB values from −2.41 to −0.73. Estimated human serum protein binding was slightly above the preferred range for 3a–c but within the reported range for 3d–f. On the basis of predicted drug-likeness and permeability, the researchers prioritized 3e as the most balanced analog and 3f for its high estimated permeability.

Docking was performed against two crystallographic EGFR kinase structures: 1M17, described as active-like, and 4HJO, described as inactive-like. The native ligands were first redocked as a protocol check, yielding root-mean-square deviations of 1.25 ångström for 1M17 and 2.59 ångström for 4HJO. Erlotinib, gefitinib and lapatinib served as reference inhibitors, while five property-matched database decoys were included as negative controls.

The decoys achieved weaker docking and induced-fit results than the synthesized compounds and reference drugs, and none formed the expected interaction with the hinge residue MET769. By contrast, all six new compounds and all three reference inhibitors retained that interaction in the reported poses. Receptors were prepared with protonation states assigned at pH 7.4, missing side chains rebuilt, and selected bridging water molecules retained for grid generation and docking.

Compound 3d produced the strongest overall docking profile in both receptor models. In inactive-like 4HJO, its Glide XP score of −9.57 and induced-fit score of −11,755.16 were more favorable than those of the three reference inhibitors, while its modeled MM-GBSA binding estimate was −67.02 kilocalories per mole. Compound 3f achieved a slightly weaker XP score of −9.35 but the most favorable modeled binding estimate in this structure, −77.75 kilocalories per mole.

Compound 3e was also competitive, with an XP score of −9.21 and an MM-GBSA estimate of −67.63 kilocalories per mole. The same broad ranking appeared in active-like 1M17: 3d led with an XP score of −7.18, an induced-fit score of −12,912.62 and a modeled MM-GBSA estimate of −72.54 kilocalories per mole, followed by 3f and 3e. These values were explicitly treated as relative estimates for the modeled complexes, not absolute solution-phase binding free energies.

The researchers then subjected 3d and 3f to 200-nanosecond molecular dynamics simulations in both EGFR conformations. The complexes were solvated in orthorhombic boxes with a 13-ångström buffer, treated with the OPLS4 force field and TIP3P water, neutralized and placed in 0.15 molar sodium chloride. Restrained constant-volume heating to 300 kelvin was followed by constant-pressure equilibration at 1 atmosphere.

Energies and coordinates were recorded every 1.2 and 10 picoseconds, respectively. Analysis covered protein C-alpha root-mean-square deviation and fluctuation, ligand heavy-atom motion, secondary-structure content and persistence of key protein–ligand contacts. No major loss of secondary-structure content was observed.

Interaction analysis showed that both ligands repeatedly occupied ATP-site contact patterns over the simulated interval. In the 4HJO–3d complex, a MET769 hydrophobic contact occurred about 84% of the time, a LYS721 hydrogen bond about 64%, and a water-mediated contact near LEU694 about 41%. For 1M17–3d, hydrogen bonds involving MET769 and THR766 occupied about 93% and 92% of the trajectory, while contacts involving LYS721 reached about 96%.

In 4HJO–3f, PHE711 hydrogen-bond or hydrophobic interactions appeared in about 99% of frames, and a PRO770 water bridge persisted about 67% of the time. In 1M17–3f, a THR830 hydrogen bond was observed about 99% of the time. Ligand motion was generally greater than receptor motion, with the trajectories of 3f showing more movement than those of 3d.

These occupancies describe contacts within the 200-nanosecond window, but they do not establish affinity, residence time, inhibition or anticancer efficacy; slower dissociation or different behavior on longer timescales cannot be excluded. Overall, the study identifies 3d and 3f as the strongest candidates for experimental testing in the two modeled EGFR conformations, with 3e offering a more balanced predicted absorption and drug-likeness profile. The main concern across the series is high predicted lipophilicity, which could compromise solubility, while compounds 3c–f are relatively flexible.

The findings extend the computational characterization of benzohydrazide scaffolds beyond a mono-imine design by evaluating two EGFR states, three reference inhibitors and five decoy controls. They do not establish that any compound is a therapeutic EGFR inhibitor, and no direct biochemical binding, kinase, cellular or animal testing was performed. Consequently, the docking, MM-GBSA and molecular dynamics findings remain relative computational indicators for prioritizing further experimental evaluation.

Subject of Research: Computational evaluation of synthesized benzohydrazide derivatives against two EGFR conformations

Article Title: Synthesis, characterization, and in silico studies of new benzohydrazide derivatives targeting EGFR kinase

Article References: Abdulmonem, M. R., Mahdi, M. F., & Raauf, A. M. R. (2026). Synthesis, characterization, and in silico studies of new benzohydrazide derivatives targeting EGFR kinase. Pharmacia, 73, Article e186971. https://doi.org/10.3897/pharmacia.73.e186971

Image Credits: AI Generated

DOI: 10.3897/pharmacia.73.e186971

Keywords: EGFR, benzohydrazide, hydrazone chemistry, molecular docking, molecular dynamics, MM-GBSA, ADME prediction, drug discovery, kinase inhibitor leads, Synthesis, characterization, silico