2.1. Protein ID mapping

Protein identifier (ID) mapping was performed using the UniProt database (Consortium, 2023), with Protein Data Bank (PDB) ID and protein names as input. A total of 32 proteins (Table 1) were selected for further analysis based on a targeted literature review and database screening. The inclusion criteria were: (1) relevance to neuroprotective mechanisms or neuropsychiatric disorders (e.g., major depressive disorder, anxiety, and mood disorders) and (2) availability of 3D (three-dimensional) structures or interaction data in UniProt and PDB.

2.2. Protein-protein interaction

The proteins obtained from the ID mapping process were de-duplicated. The resulting list was input into the STRING (search tool for the retrieval of interacting genes/proteins) database (https://string-db.org/) (Szklarczyk et al., 2022) to identify interactions among the proteins. The analysis was performed using the following parameters: organism (Homo sapiens), network type (full STRING network), and minimum required interaction score (medium confidence threshold of 0.400).

2.3. Network analysis

The Cytoscape application was used to analyze the protein interaction network (Shannon et al., 2003). Centrality analyses performed included Betweenness Centrality, Degree, and Closeness Centrality. The CytoHubba plugin was employed to identify key proteins (hub proteins). The centrality measure used was Maximal Clique Centrality (MCC), defined by Equation 1.

Where S(v) is the collection of maximal cliques containing v; (|C| – 1)! is the product of all positive integers less than |C|; and if there is no edge between the neighbors of node v, then MCC(v) is equal to its degree (Chin et al., 2014).

2.4. Molecular docking

The molecular docking method described by Cahya et al. (2024) was used with some modifications.

2.5. Absorption, distribution, metabolism, excretion, and toxicology and bioactivity

Ligands with a BSS percentage greater than 50% were analyzed for their physicochemical properties, druglikeness, and medicinal chemistry using the SwissADME tool available at http://swissadme.ch/. The structures of the compound were uploaded in the SMILES (Simplified Molecular Input Line Entry System) code. The analysis was performed based on parameters consistent with Lipinski’s Rules of Five, Ghose, Veber, Egan, Muegge, pan assay interference compounds (PAINS), and BRENK. Ligands that met all these parameters were used for further analysis. Absorption, distribution, metabolism, excretion, and toxicity (ADMET) predictions were performed using the pkCSM and admet-SAR websites. Meanwhile, bioactivity predictions were performed using the Way2Drug platform.

2.6. Molecular dynamics

The MD method was used as described on the website https://ambermd.org/tutorials/ and by Awaluddin et al. (2021) with some modifications.

3.1. Protein ID mapping

The goal of protein ID mapping is to verify and standardize the protein identifiers of 32 selected proteins to ensure consistency and enable subsequent functional annotation and data integration. Based on the 32 selected proteins, 27 were verified using the Uniprot’s IDmapping tool. The list is available in Supplementary Table S2.

3.2. Protein-protein interaction

The aim of the protein-protein interactions was to identify key hub proteins potentially involved in neuroprotective mechanisms. The PPIN generated by STRING analysis comprised 24 proteins and 57 interactions at a medium confidence level (0.400) encoded by the input gene set (Figure 1). Further analysis was performed using network analysis.

Figure 1

The network topology of protein-protein interactions (encoded by its gene) associated with neuroprotection and mental health

3.3. Network analysis

Centrality analysis of 24 proteins (Table 2) and MCC-ranked key proteins (Table 3, Figure 2) highlight the central role of the dopamine receptor D2 (DRD2) gene (D2). D2 shows the highest degree, centrality, and ranking scores, underscoring its key function in the dopaminergic system’s link to neuroprotection. The guanine nucleotide-binding protein G[i] subunit alpha-1 (GNAI1) gene supports the importance of 5-hydroxytryptamine 1A receptor (5-HT1A), which shows high centrality through its connections across genes and pathways, emphasizing its role in neurotransmitter signaling as a G-protein subunit.

Figure 2

The network topology of the key proteins (encoded by its gene) based on the MCC ranking system, with colors indicating MCC scores: Red nodes show high MCC scores; yellow node shows low MCC scores.

GABAb proteins, especially Gamma-aminobutyric acid type B receptor subunit 2 (GABBR2), exhibit high betweenness centrality, suggesting a bridging role in inhibitory neurotransmission; Gamma-aminobutyric acid type B receptor subunit 1 (GABBR1), ranked fifth by MCC, also plays a significant part. mGlu2 and mGlu3, encoded by glutamate metabotropic receptor 2 (GRM2) and glutamate metabotropic receptor 3 (GRM3), are key modulators of glutamate signaling. Among them, mGlu2 shows stronger connectivity and higher centrality, linking with major neuroprotective proteins (D2, GABAb, 5HT1A) and playing a central role in balancing excitatory and inhibitory transmission. Though mGlu3 is less connected, it remains important in glutamate regulation.

SERT (solute carrier family 6 member 4 [SLC64A]) holds a central position with high degree and closeness centrality, acting as a connector in neuroprotective pathways. CB1 (cannabinoid receptor 1 [CNR1]) and MORs (Mu-type opioid receptor [OPRM1]), while moderately connected, serve as key bridges in neuroprotective signaling. Last, MAOB shows moderate centrality, reflecting its role in monoamine metabolism. Together, these findings reveal a highly interconnected neuroprotective network, with D2, mGlu2, 5HT1A, GABAb, and mGlu3 as major regulatory hubs.

3.4. Molecular docking

Molecular docking was used to virtually screen Amaranthus-derived compounds for potential binding to target proteins. Redocking validated the docking protocol by reproducing known ligand binding modes, while blind docking identified potential unknown binding sites. Blind docking was performed instead of using cocrystallized ligands, as the available ligands did not match the desired pharmacological actions (e.g., agonist or antagonist). As shown in Table 4, all receptor-ligand complexes had RMSD values < 2 Å, confirming docking precision.

Binding affinities and BSS percentage values are summarized in Table S3 (Supplementary) and druglikeness and medicinal chemistry properties are summarized in Table S4 (Supplementary). Ligands with BSS percentage of > 50% and acceptable pharmacological profiles were shortlisted (Table 5). Several ligands showed better binding energy than control ligands (e.g., PDB: 2XCG, 5CNI, 5CNK, 4MR7), indicating potential for enhanced neuroprotective activity. Ligand-protein complexes (e.g., PDB: 4MR7, 5C1M, 5CNK, 5I6X, 5TGZ, 6CM4, 8PKM) were excluded due to violations of druglikeness or medicinal chemistry criteria (Table S4).

Figure 3 illustrates the binding interactions of selected ligands, showing whether they bind the same region as the control ligand, suggesting comparable biological activity. Figure 4 presents the 3D interaction models of the selected ligands with the receptor, offering a structural perspective on their binding orientations. Table 6 details key amino acid interactions.

Figure 3

The similarity amino acid residue interaction between selected ligand and control ligand. (A) MAOB-apigenin; (B) mGlu2-kaempferol; (C) mGlu2-dihydrokaempferol; (D) mGlu2-hydrangenol.

Figure 4

The visualization of 3D interaction between ligand and receptor. (A) MAOB-apigenin; (B) mGlu2-kaempferol; (C) mGlu2-dihydrokaempferol; (D) mGlu2-hydrangenol.

Notably, for MAOB, the control ligand binds mainly through hydrophobic interactions, forming hydrogen bonds with Tyr326 (tyrosine residue at position 326 in a protein) and Pro102. The selected ligand (e.g., apigenin), being more hydrophobic, shifts the interaction profile, possibly reducing polar contacts. However, its hydrogen bonding with Leu164, Thr201, and Glu84 suggests it may possess new polar groups, potentially enhancing MAOB inhibition and neuroprotective efficacy.

In mGlu2, residues Arg454 and Ala322 form hydrogen bonds with the control ligand but engage in hydrophobic interactions with kaempferol and hydrangenol, likely due to their aromatic structures. Dihydrokaempferol, in contrast, mirrors the control’s hydrogen bonding but with shorter bond distances, suggesting better alignment and stronger binding, potentially enhancing its functional activity.

3.5. ADMET and bioactivity

To identify the most promising neuroprotective drug candidates, selected ligands were evaluated based on their physicochemical properties (Table 7), ADMET profiles (Table 8), and bioactivity (Table 9). These evaluations ensured that each ligand possessed favorable characteristics for biological activity, pharmacokinetics, and safety. Physicochemical analysis showed that all compounds complied with standard druglikeness criteria, including Lipinski’s Rule of Five, Veber, Ghose, Egan, Muegge, PAINS, BRENK, and leadlikeness filters. The ligands exhibited total polar surface areas (TPSA) between 60 and 120 Ų, optimal for absorption and permeability, and all maintained bioavailability scores above 0.50 with good solubility, indicating their potential as drug candidates.

ADMET evaluation revealed moderate water solubility (approximately –3 log mol/L) across all ligands. Most compounds demonstrated high intestinal permeability (log Papp > 0.9), with hydrangenol and apigenin showing absorption rates above 80% while kaempferol showed poor Caco-2 permeability. For CNS targeting, hydrangenol showed the highest blood-brain barrier (BBB) permeability, although apigenin exhibited superior CNS penetration, making it more suitable for neuroprotective applications. In the metabolism analysis, all ligands except dihydrokaempferol were predicted as CYP1A2 (cytochrome P450, family 1, subfamily A, polypeptide 2) inhibitors, suggesting a potential risk for drug-drug interactions. Excretion profiling showed moderate clearance values (log mL/min/kg between –1 and 1), and none of the compounds were substrates for the renal OCT2 (organic cation transporter 2), indicating minimal renal excretion involvement.

Toxicity predictions indicated no liver toxicity for any ligand. However, hydrangenol tested positive in the AMES test, suggesting potential genotoxicity and classified as extremely toxic in oral acute toxicity. Kaempferol and dihydrokaempferol were rated as highly toxic and apigenin had the lowest toxicity, indicating a relatively safer profile.

Bioactivity analysis showed that dihydrokaempferol had the strongest antioxidant and anti-inflammatory activities while hydrangenol scored the lowest in these categories. Kaempferol showed the highest antiapoptotic activity while hydrangenol demonstrated the strongest predicted activity in neurotransmitter antagonism and dopamine release. Apigenin showed potential in neurotransmitter uptake inhibition and was the top predicted MAOB inhibitor. Although dihydrokaempferol exhibited the weakest dopamine-related activity, it was the only compound predicted to have potential activity against neurodegenerative diseases. These findings highlight the importance of integrating multiple pharmacological and safety parameters to determine the most viable neuroprotective candidates.

3.6. Molecular dynamics

MD helps confirm whether the ligand-protein interactions seen in docking are realistic, stable, and biologically meaningful under dynamic conditions, making it a critical step in validating potential neuroprotective drug candidates. The binding free energy are shown in Table 10, Figure 5, while RMSD and distance is shown in Figure 6 and Figure 7, respectively.

Figure 5

Time-dependent binding energy profiles of ligand-receptor complexes over 200 ns molecular dynamics simulations. (A) MAOB-apigenin; (B) mGlu2-kaempferol; (C) mGlu2-dihydrokaempferol; (D) mGlu2-hydrangenol.

Figure 6

RMSD of C-Alpha atoms for the protein, ligand, and protein-ligand complex during 200 ns molecular dynamics simulations. (A) MAOB-apigenin; (B) mGlu2-kaempferol; (C) mGlu2-dihydrokaempferol; (D) mGlu2-hydrangenol.

Figure 7

Distance analysis over 200 ns molecular dynamics simulations of the ligand-protein complexes. (A) MAOB-apigenin; (B) mGlu2-kaempferol; (C) mGlu2-dihydrokaempferol; (D) mGlu2-hydrangenol.

As shown in Table 10, all ligands exhibited negative van der Waals interaction energies (VDWAALS), indicating favorable steric interactions, with apigenin and dihydrokaempferol showing the most pronounced effects. Dihydrokaempferol also demonstrated the strongest electrostatic interaction energy (EEL), suggesting that charge-based interactions play a key role in stabilizing the complex. Dihydrokaempferol recorded the highest Generalized Born Solvation Energy (EGB), indicating reduced stability in aqueous environments compared to other ligands. In contrast, hydrangenol and kaempferol showed lower EGB values, reflecting more favorable polar solvation while apigenin exhibited moderate solvation energy. Nonpolar solvation energy (ESURF) was the most favorable for apigenin and dihydrokaempferol, suggesting better compatibility with nonpolar solvents.

Despite its gas-phase stability reflected by the most negative ΔG_gas, dihydrokaempferol exhibited the highest overall solvation energy (ΔG_solv), indicating lower solubility in water, whereas hydrangenol with the lowest ΔG_solv appeared to be the most readily solvated in aqueous conditions. All ligands displayed negative total binding free energies (ΔG_total), suggesting thermodynamically favorable interactions, with dihydrokaempferol showing the most favorable ΔG_total, followed by apigenin, identifying both as the most promising candidates in terms of binding affinity and complex stability. To further evaluate the stability of these interactions over time, binding energy fluctuations were monitored throughout the 200 ns simulation, as illustrated in Figure 5.

Based on Figure 5, dihydrokaempferol exhibited a lower average binding energy, approximately from –30 to –40 kcal/mol, with higher fluctuations observed between 75 ns and 150 ns. While dihydrokaempferol showed the lowest average binding energies, hydrangenol demonstrated the most consistent binding energy profile with minimal fluctuations over time. This indicates that hydrangenol forms a moderately strong yet highly stable complex with the receptor, suggesting it may be the most favorable ligand in terms of binding stability.

RMSD analysis of the protein, ligand, and the complex (Figure 6) aims to evaluate the dynamic stability of the protein-ligand complexes during the 200 ns MD simulation. The RMSD analysis highlights that hydrangenol formed the most stable protein-ligand complex, both in terms of protein structure and ligand retention. In contrast, dihydrokaempferol caused greater structural deviation in the complex despite maintaining its position within the binding site.

To further assess ligand stability within the binding pocket, distance analyses were performed over the 200 ns MD simulation (Figure 7). Distance analysis over 200 ns MD simulations revealed that dihydrokaempferol exhibited the most stable binding, with minimal pose deviation and consistent proximity to the binding site. Hydrangenol also maintained a stable interaction while kaempferol showed notable pose fluctuations, suggesting less stable binding. The protein-ligand distance remained steady across all complexes, although kaempferol displayed minor variation. Notably, apigenin exhibited slight fluctuations in the end to end distance, indicating potential system-level conformational changes.