Currently, more than 55 million people worldwide are living with dementia, and this number is projected to rise to 78 million by 2030. Dementia has become the seventh leading cause of death globally (World Alzheimer Report 2021). Alzheimer’s disease (AD) is the most prevalent form of dementia, responsible for nearly 70% of all cases. Clinically, AD is marked by symptoms such as memory loss, apathy, depression, impaired judgment, confusion, disorientation, and other cognitive impairments.¹ Despite significant efforts from both academic institutions and the pharmaceutical industry, no approved drugs are currently available that effectively treat AD. However, the drug development pipeline remains highly active, with disease-modifying therapies (DMTs) targeting the underlying pathology of AD accounting for approximately 70% of the 105 compounds currently in clinical trials.² Although the mechanisms of action of current agents in Phases I–III are diverse, they can generally be categorized as targeting either the amyloid cascade or downstream pathological processes, such as the Tau pathway. Notably, over half of the Phase III agents are anti-amyloid therapies, whereas only about 4% are specifically aimed at targeting Tau.³ The majority of agents in the current drug development pipeline are either monoclonal antibodies or β-site amyloid precursor protein cleaving enzyme (BACE) inhibitors. Under normal conditions, tau is a highly soluble, natively unfolded protein, which differs significantly from its hyperphosphorylated form. As a microtubule-associated protein, tau is essential for maintaining neuronal structure, ensuring microtubule stability, and supporting intracellular transport within neurons.⁴ Normal tau protein becomes pathological due to the hyperactivation of phosphatases, leading to the formation of paired helical filaments (PHFs) and neurofibrillary tangles (NFTs) in the brains of individuals with AD.^5–7 Phosphorylated tau (p-tau) is negatively associated with synaptic damage in Alzheimer's disease neurons; however, the exact mechanisms underlying this damage remain incompletely understood.⁸

In the brains of AD patients, six tau protein isoforms (352–441 amino acids) aggregate into neurofibrillary tangles, forming PHFs via double-helical stacking of C-shaped subunits.⁹ The formation of PHFs is believed to be closely linked to the hyperphosphorylation of tau protein, which entails the attachment of phosphate groups to amino acids like serine, threonine, and tyrosine. This excessive phosphorylation impairs tau’s ability to bind to microtubules, leading to its detachment and subsequent aggregation into neurofibrillary tangles.^10,11

The development of drugs or drug-like molecules is a lengthy and complex process involving steps such as target identification, compound selection, and biological screening. To streamline this process and reduce costs and time, computational tools have been increasingly utilized. Molecular docking and dynamics simulations model interactions with target molecules, while crystallography and DFT provide insights into compounds' properties, reactivity, and biological compatibility. An earlier inverse molecular docking study identified potential curcumin targets among human proteins, highlighting some related to cancer and AD.¹² Despite its therapeutic potential, curcumin’s clinical use is limited by poor solubility and bioavailability. To address this, researchers synthesized six curcumin derivatives and used ab initio molecular simulations to study their interactions with tau protein.^13,14 Bhanukiran and Hemalatha conducted a comprehensive study on the design and evaluation of a novel 3-OH pyrrolidine derivative, VA10. They employed single-crystal X-ray crystallography, molecular dynamics simulations, and biological assays to explore the structural and functional characteristics of VA10. Additionally, density functional theory (DFT) calculations were utilized to analyze its electronic properties and assess its potential as a Beta-site Amyloid Precursor Protein Cleaving Enzyme 1 (BACE1) inhibitor.¹⁵

We here employed 10 quinoxaline derivative compounds as potential inhibitors of tau protein aggregation. Their chemical structures are presented in Fig. 1. The drug-likeness of these compounds, referring to their potential suitability as orally active drugs was assessed using Lipinski's Rule of Five, a well-established guideline used to predict the oral bioavailability based on key physicochemical properties. Molecular docking simulations were performed to predict the initial conformations of the ligands bound to the tau protein and to broadly explore their potential binding sites. Furthermore, quantum chemical calculations were used to identify which of these derivatives exhibit stronger binding to tau protein and effectively inhibit its aggregation.

The pharmacokinetic properties of the quinoxaline compounds (1–10) were evaluated in silico using the SwissADME tool (https://www.swissadme.ch/) by converting each molecule into its canonical SMILES (Simplified Molecular Input Line Entry System) format. As the initial structure of the tau protein, we used the chain A monomer from the tau decamer available in the PDB (PDB ID: 5O3L).¹⁶ Molecular docking was performed using AutoDock 4.2 with its standard operating procedure and default settings, employing an empirical free energy scoring function. Since the binding site was initially unknown, blind docking was conducted by encompassing the entire receptor within a grid box of specified dimensions 126 Å × 126 Å × 42 Å with 0.375 Å grid spacing. The grid box was centered at the geometric center of the tau protein, with a mutation rate of 0.02 and a crossover rate of 0.8 applied during the docking simulations. During docking, all residues of the tau protein were kept rigid, while the ligands (drug candidates) were treated as flexible within the grid to allow exploration of various binding conformations. The evaluation limit was set to 250,000 with a population size of 300, generating 50 conformations over 27,000 generations using the Lamarckian Genetic Algorithm. Discovery Studio Visualizer¹⁷ was utilized to assess the interactions between the ligands and the target receptor.

All the molecular structures were fully optimized by using the wb97xd/6-311 + g(d,p) method of the density functional theory calculation using Gaussian 09, Revision B.01.¹⁸ Harmonic frequency analyses were conducted at the same theoretical level to identify stationary points and to determine thermal and zero-point energy corrections. Single-point energy computations were also carried out using the wb97xd/6-311 + g(d,p) level of theory to obtain Molecular Electrostatic Potential (MESP) surfaces in the gas phase. These surfaces were visualized using the GaussView 4.1.2 software suite. The interaction energy (E_i) approach serves as an effective method for interpreting reactivity in bimolecular reactions. It is defined as E_i = E_adduct − (E_tau + E_ligand). The MESP was derived using Eq. (1), where Z_A ​ represents the nuclear charge of atom A at position R_A​, and ρ(r′) denotes the electron density.¹⁹

Typically, regions with high electron density exhibit strongly negative values in the Molecular Electrostatic Potential (MESP), while electron-deficient areas are associated with positive MESP values. The point with the most negative potential (V_min) within electron-rich zones can be identified through MESP topographical analysis.

In this study, ten quinoxaline based ligand molecules were evaluated for their drug-likeness based on Lipinski’s Rule of Five and related physicochemical criteria, including molecular weight (MW), partition coefficient (LogP), topological polar surface area (TPSA), hydrogen bond donors (HBD) or/and hydrogen bond acceptors (HBA), and rotatable bonds (RB). These parameters are widely recognized for their relevance to oral bioavailability and pharmacokinetic behavior in drug development.^21–25

As evident from Table 1, all ten ligands demonstrated favorable profiles with respect to the drug-likeness thresholds (MW < 500 g/mol, LogP < 5, TPSA < 140 Ų, HBD ≤ 5, HBA ≤ 10, RB ≤ 10). The molecular weights ranged from 160.17 to 322.31 g/mol, well within the acceptable limit. LogP values were modest, falling between 0.79 and 2.35, indicating balanced hydrophilic–lipophilic properties that support both membrane permeability and aqueous solubility. The TPSA values of the compounds ranged from 34.89 Ų to 78.26 Ų, which are below the 140 Ų threshold known to correlate with good intestinal absorption. Hydrogen bonding potential was also within permissible limits, with all molecules having no more than 1 HBD and 5 HBA. Importantly, none of the ligands contained more than 5 rotatable bonds, further suggesting a favorable conformational flexibility compatible with oral bioavailability.

The absence of any Lipinski violations across all compounds strongly suggests that these ligands are likely to be orally bioavailable. This is a promising indication for their further development as drug candidates. Overall, ligands 1–4 serve as small, balanced scaffolds with strong oral drug-likeness, while ligands 5–8, with moderate size and low polarity, appear particularly promising for central nervous system applications due to their enhanced permeability potential. Ligands 9 and 10, though larger and more polar, remain compliant with drug-likeness rules and may be better suited for systemic targets that require stronger binding specificity. Notably, the lower TPSA and LogP values of ligands 5–8 support efficient membrane permeability, whereas the slightly higher molecular weights of ligands 9 and 10 still maintain favorable physicochemical profiles without breaching critical thresholds. Moreover, the consistently low number of rotatable bonds across all compounds points toward structural rigidity, which can improve binding specificity and reduce entropy loss upon target binding. However, this must be balanced with sufficient flexibility to accommodate the dynamic nature of protein binding sites.

Figure 2 presents radar plots illustrating the physicochemical profiles of ligands 1–10, with each red hexagon representing an individual compound and the pink shaded area indicating the optimal range for oral bioavailability. This optimal zone encompasses six key parameters: lipophilicity (LOGP between 0.79 and 3.01), molecular size (160–322 g/mol), polarity (TPSA between 20 and 130 Ų), aqueous solubility (logS < 6), unsaturation (fraction of sp³ >0.25), and flexibility (≤ 9 rotatable bonds). Compounds whose profiles fall largely within this region are considered drug-like and are more likely to exhibit favorable oral bioavailability. All ligands exhibit moderate lipophilicity, with ligands 6–10 showing slightly higher values, which may enhance membrane permeability, although excessive lipophilicity can compromise solubility. Molecular size increases progressively from ligands 1 to 10, with ligands 8–10 being the largest potentially reducing permeability and dermal absorption. Polarity is generally higher in ligands 6–10, which can improve aqueous solubility but may limit passive membrane diffusion. While all ligands display low insolubility, ligands 8 and 10 show slightly elevated values, likely due to increased size or polarity, though overall absorption remains favorable. Ligands 1–5 show greater unsaturation, indicative of rigid or aromatic frameworks, whereas ligands 6–10 are more saturated and flexible traits that can enhance receptor binding adaptability but may reduce metabolic stability and skin permeation. Ligand 9 is the most flexible among the series, with increased flexibility potentially impacting target selectivity. Overall, compounds falling largely within the pink region are considered drug-like and likely to exhibit good oral bioavailability.

The analysis of physicochemical parameters viz., molecular weight, lipophilicity, TPSA, hydrogen bonding capacity, and molecular flexibility demonstrates that all tested compounds exhibit favorable drug-likeness characteristics, highlighting their promise as orally bioavailable candidates for further pharmacological development. From Table 2, it very clear that these compounds exhibit high gastrointestinal (GI) absorption, indicating strong potential for oral bioavailability. This aligns with their drug-likeness profiles discussed earlier and is a critical factor in oral drug development. Ligands 1–9 are blood-brain barrier (BBB) permeant, suggesting they can potentially affect or target the central nervous system (CNS). Ligand 10 is not BBB permeant, making it potentially more suitable for non-CNS indications, minimizing CNS-related side effects. Furthermore, none of the ligands are predicted to be P-glycoprotein (Pgp) substrates. This is favorable, as Pgp can limit drug absorption and brain penetration by actively exporting drugs out of cells. Non-substrate status indicates better absorption, distribution, and reduced multidrug resistance risks. The inhibition of cytochrome P450 (CYP) enzymes is a critical factor in assessing the potential for drug–drug interactions (DDIs). CYP1A2 is inhibited by all ligands (1–10), indicating a possible risk of interactions with co-administered drugs metabolized by this enzyme.CYP2C19 and CYP2C9 are inhibited only by Ligands 9 and 10, which may increase the likelihood of DDIs involving drugs processed by these enzymes. Importantly, none of the ligands inhibit CYP2D6 or CYP3A4, the primary metabolic pathways for a wide range of pharmaceuticals, which is favorable from a metabolic safety perspective.

Ligands with greater polarity, larger molecular size, and higher flexibility such as compounds 6 and 10 tend to show reduced dermal absorption. In contrast, compounds with higher lipophilicity and structural rigidity, such as ligand 8, display enhanced skin permeability. Importantly, none of the compounds inhibit the cytochrome P450 enzymes CYP3A4 or CYP2D6, suggesting a lower potential for drug–drug interactions. Collectively, ligands 1–8 combine high GI absorption, BBB penetration, and minimal CYP inhibition, making them promising for CNS-active drug design, while ligands 9 and 10, due to their broader CYP inhibition profile and in the case of ligand 10, lack of BBB penetration, may be better directed toward systemic non-CNS therapeutic applications.

The results of molecular docking simulations obtained by AutoDock 4.2 are listed as Table 3. The ligands were ranked by binding energy (BE) as follows: Ligand 10 > 9 > 2 > 7 > 4 > 3 > 1 > 8 > 6 > 5, with BE values ranging from − 5.06 to -7.39 kcal/mol (Table 3 and Fig. 3). A more negative BE reflects stronger and potentially more favorable interactions. The number of conformations within each ligand cluster provides insight into the stability and prevalence of a particular binding pose. Ligands 1–4 consistently interact with residues Asn359, Leu357, Gly333, and Val337, with binding energies ranging from − 5.29 to − 5.78 kcal/mol and forming stable hydrogen bonds with Asn359 (bond lengths ~ 1.83–2.20 Å), suggesting a conserved interaction hotspot in this region. Ligand 5 shows interactions with Asp348, Phe346, and Val350 and hydrogen bonds with Asp348 (2.06 Å) and Val350 (2.12 Å), despite its weaker binding energy (-5.06 kcal/mol), had the highest number of conformations (36), suggesting a frequently occurring binding mode. In contrast, Ligand 7 appeared only once, pointing to a unique but potentially specific interaction. Ligand 6 engages Gly367, Ser320, and Cys322, with a binding energy of − 5.08 kcal/mol and hydrogen bonding to Gly367. Ligands 3 and 4, with 12 and 15 conformations respectively, showed moderate BE and good pose stability. Residues such as Asn359, Leu357, and Gly333 frequently appeared in ligands 1–4, indicating a potential binding hotspot on the Tau protein. Meanwhile, residues like Phe346, Gly367, and Cys322, recurrent in ligands 5–10, suggest the existence of multiple binding regions within residues 320–370. Interestingly, ligands 9 and 10 demonstrate the strongest binding affinities, with − 6.60 and − 7.39 kcal/mol, respectively. Ligand 9 interacts with Pro364, Gly367, and His362, while ligand 10 engages a broader set of residues (Pro364, Gly367, Asn368, Thr319, Val363, Gly366, Cys322, and Leu325). These findings suggest that ligands 9 and 10, owing to their lower binding energies, broader residue engagement, and stable hydrogen bonding, exhibit the most favorable interactions with Tau protein and stand out as the strongest candidates for further evaluation, while ligands 1–4 show moderate and consistent binding through conserved Asn359 interactions.

Hydrogen bonding further supported the binding stability. All tested ligands exhibited hydrogen bonding with at least one Tau protein residue, and up to two in some cases. Notably, the first four ligands consistently formed hydrogen bonds with Asn359, highlighting this residue as a key interaction site. Ligands 1–4 formed bonds with Asn359 at favorable distances (1.83–2.20 Å). Ligand 10 formed hydrogen bonds with both Gly367 (1.84 Å) and Asn368 (2.14 Å), reflecting synergistic multi-residue interactions. Other ligands, such as 5, 6, and 9, formed stable bonds with Asp348, Gly367, and His362, respectively. Ligand 10 stands out as the most promising candidate due to its strong binding energy, multiple hydrogen bonds, and interaction with eight different Tau residues, reflecting a stable and specific binding profile. The frequent involvement of Asn359, Gly367, and Phe346 across multiple ligands highlights these residues as key targets in Tau-ligand interactions. Ligands that exhibit strong binding, multiple hydrogen bonds, and high conformational prevalence (such as Ligands 2, 3, 4, and 10) merit further in silico and experimental validation.

While docking studies offer a rapid assessment of binding potential and pose prediction, they are based on empirical scoring functions and simplified representations of electronic interactions. Consequently, they may not fully capture the subtleties of charge distribution, polarization, and orbital-level interactions that govern molecular recognition and complex stability. To gain deeper insight into the electronic structure, charge transfer characteristics, and reactivity of the most promising ligand–protein complexes, Density Functional Theory (DFT) calculations are warranted. DFT enables the quantitative evaluation of molecular electrostatic potential (MESP) surfaces. Such descriptor can elucidate the intrinsic electronic factors contributing to binding strength, specificity, and potential inhibitory activity. Moreover, the docking results suggest that several ligands interact with different binding sites and residues, which may imply varied electronic environments. DFT calculations can help validate and refine these interactions at the quantum mechanical level, identify the most electrophilically or nucleophilically active regions of the ligands, and predict how subtle structural modifications may enhance binding.

Initially, ligand molecules were optimized at the wb97xd/6-311 + G(d,p) level to obtain stable conformers. The optimized stable conformers of the ligands were used for analyzing the molecular electrostatic potential (MESP). Subsequently, MESP surfaces were generated for all ten ligand molecules to identify potential interaction sites distributed across their structures. MESP is a key topographical property widely used to understand molecular reactivity, provide a rough estimation of intermolecular interactions, and analyze molecular recognition, electrophilic reactions, and substituent effects, etc.^26–30 MESP analysis identifies the point of minimum potential (Vmin) in electron-rich regions of a molecule, as determined through topographical calculations. (Fig. 4). It is evident that the ligand candidates are likely to exhibit Vmin values near their respective heteroatoms. Table 4 clearly shows that the Vmin of the carbonyl (> C = O) group is more negative than that of the nitrogen atom in the pyrazine ring. The total V_min due to overall electron rich regions situated present on each ligand is in the decreasing order of : 2 > 8 > 1 > 7 > 5 > 6 > 4 ≈ 3 > 10 > 9. Based on the overall MESP-Vmin trend, the ligands are expected to exhibit a similar relative interaction pattern. Consequently, the relative ability of ligands towards binding the tau moieties (ligand/tau) may be anticipated in the decreasing order of 2 > 8 > 1 > 7 > 5 > 6 > 4 ≈ 3 > 10 > 9. Such trend is not complementing the docking studies as the interacting ability of all the ligands with tau protein predicted is found to be different. It is well known that predicting the reactivity trend of ligands based solely on MESP-Vmin values is challenging, especially when intermolecular interactions between tau moieties and ligands are involved. To get more reliable data, further DFT-based analysis is essential to complement the docking data by providing a more rigorous and detailed understanding of ligand–Tau interactions. This will aid in identifying the most promising candidates for lead optimization and rational drug design targeting Tau-associated neurodegenerative pathways.

Given the computational complexity associated with full protein–ligand systems, model compounds representing key interacting amino acid residues from the Tau protein were constructed. These were derived from the docking-optimized geometries, ensuring a realistic representation of the binding configuration. The selected ligand–Tau amino acid adducts, henceforth referred to as modeled ligand-bound Tau complexes, were optimized using the wB97XD/6-311 + G(d,p) level of theory. This hybrid functional, which incorporates empirical dispersion corrections, is particularly well-suited for accurately capturing non-covalent interactions such as hydrogen bonding and van der Waals forces, which are central to protein–ligand recognition.

The optimized geometries (Fig. 5) and computed interaction energies reveal notable features. In all complexes, the ligands consistently engaged the Tau moiety via N–H and C = O functional groups, with interactions typically occurring between two adjacent amino acids. These interactions stabilized the complex through well-aligned hydrogen bonds, as confirmed by the computed intermolecular distances and orientation of electron-dense regions. The electronic interaction energies (in kcal/mol) for the Tau–ligand complexes were computed as follows: − 34.9 (Ligand 1), − 35.9 (2), − 32.4 (3), − 32.9 (4), − 21.6 (5), − 21.6 (6), − 27.8 (7), − 27.2 (8), − 35.9 (9), and − 49.9 (10). Among all, Ligand 10 exhibited the most negative interaction energy (–49.9 kcal/mol), indicating a highly stable electronic interaction with Tau, consistent with its superior binding affinity observed during molecular docking. Ligands 2 and 9 also displayed strong interaction energies (–35.9 kcal/mol), reinforcing their potential as effective Tau binders. Interestingly, the interaction energies reflected not only hydrogen bonding but also subtle orbital-level interactions and polarizability effects. The cyclic N–H and C = O groups appeared to dominate the stabilization mechanism, forming directional interactions with backbone or side-chain functional groups of key Tau residues.

When the relative binding strengths were ranked based on electronic interaction energies and corrected Gibbs free energies, the order of interaction affinity was found to be:

10 > 2 ≈ 9 > 1 > 4 ≈ 3 > 7 > 8 > 5 ≈ 6. This trend was found to be in good agreement with the docking-derived binding energies, thereby validating the docking predictions through a more rigorous quantum mechanical framework. Notably, Ligand 10 consistently emerged as the top candidate, supported by both docking and DFT calculations. The correlation between these methods strengthens confidence in Ligand 10's potential as a strong Tau binder.

All ten ligands satisfy Lipinski's Rule of Five and related drug-likeness criteria, indicating their strong potential as orally active pharmaceutical candidates. These findings provide a solid foundation for further pharmacological evaluation, including in vitro and in vivo studies to assess absorption, distribution, metabolism, excretion, and toxicity (ADMET) profiles. Molecular docking studies also revealed that all the ten ligands could bind the tau protein with ligand 10 as the strongest binding agent to Tau protein (–7.39 kcal/mol), forming multiple hydrogen bonds and interacting with eight key residues, indicating a stable and specific binding profile. Asn359, Gly367, and Phe346 were identified as common interaction sites across several ligands, suggesting potential binding hotspots. While Ligand 5 had the most conformations, its weak binding energy highlights that pose frequency alone doesn't ensure strong interactions. Furthermore, DFT and MESP analyses were used to assess the interaction potential of ten ligands with tau protein, revealing key electronic factors and the role of specific functional groups in ligand–tau binding. These results confirm and extend the molecular docking findings, demonstrating that ligands forming strong hydrogen bonds and exhibiting favorable electronic characteristics for further drug development targeting Tau pathology. Overall, this combined theoretical analysis supports ligands 10, 2, and 9 as promising candidates for further development, with Ligand 10 consistently demonstrating the strongest interaction across all computational approaches. The findings provide a solid foundation for subsequent in vitro validation and structure-based lead optimization in the development of Tau-targeted therapeutics for neurodegenerative diseases such as Alzheimer’s.