1 Introduction
Breast cancer is the most frequently diagnosed malignancy among women and ranks among the top three most common cancers globally, alongside lung and colorectal cancer.
1,2 It primarily affects women, with risk increasing substantially with age; more than 80% of cases occur in women aged 50 years and older.
3 In 2020 alone, breast cancer was responsible for approximately 685,000 deaths worldwide, making it the leading cause of cancer-related mortality in women.
4 Therapeutic regimens such as tamoxifen and doxorubicin have contributed significantly to clinical management but are associated with considerable side effects, including cardiotoxicity and endocrine dysfunction.
5 Alarmingly, Asia experienced the highest rise in breast cancer prevalence in 2022, accounting for 39.1% of all new global cases.
6 Current frontline pharmacological therapies include agents such as doxorubicin, trastuzumab, tamoxifen, paclitaxel, pembrolizumab, and everolimus.
7 Despite their efficacy, these treatments frequently result in adverse outcomes that compromise patient quality of life and limit long-term adherence.
8 Consequently, there is an urgent need to discover novel therapeutic strategies that offer both safety and efficacy. In this context, natural products derived from medicinal plants have emerged as promising alternatives due to their multitarget actions and lower toxicity profiles. The exploration of plant-derived compounds presents a valuable opportunity to address the shortcomings of existing cancer treatments.
Indonesia is home to a rich biodiversity of medicinal plants that have long been used in traditional herbal remedies. These indigenous botanicals hold untapped potential as sources of bioactive compounds that may complement or substitute existing cancer therapies.
9,10 With increasing interest in non-toxic, sustainable, and culturally integrated treatment options, the exploration of Indonesia's natural products has gained momentum in both local and international scientific communities.
11 Herbal medicines, in particular, are gaining popularity among patients seeking gentler therapeutic alternatives to conventional chemotherapy and targeted treatments. The development of phytome-dicines from native plants offers several advantages, including accessibility, cost-effectiveness, and ecological sustainability.
12,13 Furthermore, these approaches can support the growth of Indonesia's health and biopharmaceutical industries. As scientific validation becomes more integrated with ethnobotanical knowledge, many plant species traditionally used in folk medicine are being subjected to modern pharmacological scrutiny.
14 Identifying specific mechanisms of action through computational and molecular techniques enhances the credibility of such therapies. Accordingly, there is growing interest in investigating locally available herbs with reported anticancer effects, especially for high-burden diseases like breast cancer.
Citrus aurantium, also known as bitter orange, is a plant species from the Rutaceae family that has garnered attention for its diverse pharmacological activities.
15 Native to Southeast Asia, this plant is now widely cultivated in various regions, including North Africa, the Middle East, and Europe.
16 Its fruits are edible, but its peels, flowers, seeds, and leaves are particularly valued for their rich content of bioactive compounds such as flavonoids, alkaloids, and essential oils.
17 Numerous studies have reported the therapeutic potential of
Citrus aurantium derivatives, including their antibacterial, antifungal, anxiolytic, anti-obesity, hypoglycemic, and antioxidant properties. Importantly, several constituents, such as naringin, hesperidin, and nobiletin, have demonstrated cytotoxic activity against cancer cell lines and modulate key oncogenic pathways.
18,19 These phytochemicals are believed to exert effects on molecular targets related to proliferation, apoptosis, angiogenesis, and inflammation. Their multitarget nature makes them attractive candidates for diseases with complex pathologies such as cancer.
20 Moreover, the relatively low toxicity of these compounds enhances their potential for long-term therapeutic use. Despite promising preclinical data, the precise molecular mechanisms through which
Citrus aurantium metabolites act against breast cancer remain underexplored.
To uncover the potential mechanisms underlying the anticancer effects of natural compounds, researchers increasingly rely on systems-level computational approaches. Network pharmacology is one such method that integrates bioinformatics, pharmacology, and systems biology to study drug-target-pathway interactions holistically.
21,22 Unlike traditional pharmacology, which focuses on single targets, network pharmacology considers the multidimensional effects of bioactive compounds across multiple signaling pathways. This is particularly relevant for diseases like breast cancer, where the pathogenesis involves a complex interplay of genetic, epigenetic, and environmental factors.
23 The approach enables researchers to predict therapeutic targets, identify synergistic compound interactions, and prioritize candidates for downstream validation. Moreover, it facilitates the rational design of combination therapies that can modulate several key nodes in oncogenic networks.
24 Combined with other in silico techniques such as molecular docking and dynamics simulation, network pharmacology can provide valuable insights into the pharmacodynamic profiles of natural products. Such integrative methodologies reduce the reliance on labor-intensive experimental trials, thereby accelerating drug discovery.
Given the urgent need for safer alternatives in breast cancer therapy and the therapeutic promise of plant-based compounds, Citrus aurantium represents a compelling candidate for further investigation. Its bioactive metabolites may act on crucial signaling pathways involved in tumor progression, hormone regulation, and immune modulation. A deeper understanding of its molecular interactions with cancer-associated proteins could inform the development of novel phytotherapeutic agents. By leveraging computational tools, researchers can map compound–target interactions and predict the structural compatibility of plant-derived ligands with relevant receptors. This approach offers a cost-effective and scientifically robust avenue for identifying new candidates for breast cancer treatment. In light of these considerations,Citrus aurantium's therapeutic relevance in oncology warrants further exploration, especially through network-based and molecular modeling techniques. Such investigations not only enhance our understanding of this medicinal plant but also contribute to the global effort of identifying effective and accessible therapies for breast cancer. As research progresses, Citrus aurantium may emerge as a valuable resource in the evolving landscape of cancer drug discovery.
2 Materials and methods
2.1 Software and database
Several computational tools and databases were employed in this study. The secondary metabolites of
Citrus aurantium were retrieved from the KNApSAcK Family database (www.knapsackfamily.com/).
25 Canonical SMILES of each compound were used in SwissADME (www.swissadme.ch/)
26 and ProTox 3.0 (tox.charite.de/protox3/#),
27 to predict pharmacokinetics and toxicity profiles. GeneCards (genecards.org/)
28 and SwissTargetPrediction (www.swisstargetprediction.ch/)were used to gather breast cancer-related targets and compound target predictions.
29 FunRich v3.1.3 software was used to generate a Venn diagram and identify overlapping targets. STRING (string-db.org/),
30 was used to generate protein–protein interaction networks, and Cytoscape v3.10.3 with CytoHubba was used for network visualization and hub gene selection.
31 Molecular docking was performed using AutoDock Tools 1.5.7
32 and AutoDock Vina,
33 and visualized using BIOVIA Discovery Studio 2021.
34 The 3D structures of the compounds were obtained from PubChem (pubchem.ncbi.nlm.nih.gov),
35 and the protein target ESR1 (PDB ID: 1XPC) was retrieved from the Protein Data Bank(www.rcsb.org/).
36 Molecular dynamics simulations were conducted using GROMACS 2016.3
37 with ligand topology prepared via ACPYPE.
38 MM/PBSA calculations were carried out using the g_mmpbsa tool integrated into GROMACS workflows.
392.2 Secondary metabolites of Citrus aurantium
The secondary metabolites of Citrus aurantium were obtained from the KNApSAcK Family database. This resource lists known phytochemical compounds derived from medicinal plants. All identified compounds were compiled into an Excel spreadsheet. For each compound, the corresponding canonical SMILES was extracted for use in downstream analyses. These SMILES notations served as input for pharmacokinetic and target prediction tools. Only those compounds with complete structural data were retained for further evaluation. The selection process aimed to identify diverse scaffolds with potential biological activity. The data set was also curated to remove redundant or structurally similar entries. This initial step provided a comprehensive library of bioactive compounds from Citrus aurantium. These compounds were then carried forward for ADME screening and molecular docking simulations.
2.3 Pharmacokinetics, drug-likeness, and toxicity analysis
The pharmacokinetic profiles of each compound were evaluated using SwissADME. Canonical SMILES were submitted individually to assess parameters like Lipinski's Rule of Five and bioavailability scores. The analysis also included gastrointestinal absorption, water solubility, and synthetic accessibility. Compounds that violated more than one Lipinski criterion were excluded from further study. Next, ProTox 3.0 was used to predict oral toxicity, LD50 values, and toxicity classes. Toxicity mechanisms and structural alerts were also recorded when available. Compounds were categorized into classes 1 through 6 based on GHS toxicity criteria. All results were tabulated to compare safety profiles. This dual screening helped prioritize compounds that combined favorable pharmacokinetics with low predicted toxicity. The selected candidates were subjected to target prediction and docking analysis.
2.4 Target prediction and overlap identification
To predict protein targets relevant to breast cancer, two databases were utilized. First, GeneCards was queried using the keyword “breast cancer” to retrieve disease-related targets. Second, the SMILES of each compound were input into SwissTargetPrediction to identify compound-specific targets. Only Homo sapiens entries with probability scores greater than zero were retained. The resulting gene lists were compared to find overlapping targets. This intersection was visualized using a Venn diagram generated with FunRich v3.1.3. The overlapping targets were considered core targets for subsequent analysis. All identified genes were validated for relevance using STRING database annotations. These core targets were then subjected to network analysis. This approach ensured the selection of biologically meaningful interactions between compounds and disease.
2.5 Protein–protein interaction (PPI) network construction
The overlapping protein targets were uploaded into the STRING database (Homo sapiens organism) to construct the protein–protein interaction network. The interaction confidence score was set to high confidence (0.7) rather than the medium default (0.4), to ensure reliability of the predicted associations. The resulting network was exported in CSV format and imported into Cytoscape v3.10.3 for visualization and topological analysis. The CytoHubba plugin was used to rank and identify the top 10 hub genes based on degree centrality, representing the most highly connected and biologically influential nodes. Functional enrichment analysis was subsequently performed using STRING annotations, with an FDR-adjusted p-value≤0.05 as the significance threshold. Only Gene Ontology (GO) and KEGG categories were retained, and redundant terms were filtered using a similarity cutoff of 0.5. The top enriched pathways and biological processes were summarized with adjusted p-values and gene counts. This network-based analysis provided a robust systems-level understanding of how Citrus aurantium metabolites potentially modulate cancer-related molecular interactions.
2.6 Molecular docking
Molecular docking was performed using AutoDock Tools 1.5.7 and AutoDock Vina to predict the binding interactions between selected Citrus aurantium compounds and ESR1 (PDB ID: 1XPC), which was chosen due to its high connectivity in the PPI network. The protein structure was prepared by removing crystallographic water molecules, separating native ligands, adding polar hydrogens, and assigning Kollman charges using BIOVIA Discovery Studio 2021. Missing residues in the crystal structure were repaired using the Molecular Operating Environment (MOE) software. Ligand structures were retrieved from PubChem in SDF format and converted into PDB using BIOVIA Discovery Studio, then further converted into PDBQT format in AutoDock Tools. The docking grid was set to encompass the ESR1 active site with dimensions of 64 × 60 × 60 points (x, y, z), a grid spacing of 0.375 Å, and a grid center at (30.751,−1.346, 23.816). The docking parameters included an exhaustiveness corresponding to 100 genetic algorithm runs(ga_run = 100) and standard local search and GA-LS hybrid optimization parameters: ga_pop_size = 150, ga_num_evals = 250000, ga_num_generations = 27000, ga_mutation_rate = 0.02, and ga_crossover_rate = 0.8. Clustering was performed using an RMS tolerance of 2.0 Å with ranked cluster analysis. The redocking validation of the native ligand yielded an RMSD of 0.52 Å, confirming the reliability of the docking protocol. Known ESR1 modulators such as benzoxathiin-based SERMs were included as positive controls, while a decoy ligand set was used to estimate the false-positive rate. Binding affinities were recorded in kcal/mol, and docking poses were visualized in BIOVIA Discovery Studio 2021.
2.7 Molecular dynamics simulation
The best ligand–receptor complex obtained from molecular docking was subjected to molecular dynamics (MD) simulation using GROMACS 2016.3. The AMBER99SB-ILDN force field was applied for the protein, while ligand topologies were generated using ACPYPE based on the General AMBER Force Field (GAFF) with AM1-BCC charge model. The complex was solvated in a triclinic box with a minimum distance of 1.0 nm between the protein surface and the box edge, using the SPC/E water model. Counterions (Na+ and Cl−) were added to neutralize the system and maintain physiological ionic strength. Energy minimization was performed using the steepest descent algorithm for 5000 steps until the maximum force converged below 1000 kJ/mol⋅nm. The equilibrations were conducted in two phases: NVT (constant number of particles, volume, and temperature) for 1 ns and NPT (constant number of particles, pressure, and temperature) for another 1 ns, both maintained at 300 K using the V-rescale thermostat and 1 bar using the Parrinello–Rahman barostat. All covalent bonds involving hydrogen atoms were constrained using the LINCS algorithm. Long-range electrostatic interactions were treated with the Particle Mesh Ewald (PME) method, and a cutoff distance of 1.0 nm was applied for both Coulombic and van der Waals interactions. A time step of 2 fs was used, and coordinates were saved every 100 ps for analysis. A 300 ns production run was performed to evaluate the conformational stability and dynamic behavior of the complex. The SPC/E water model was selected for compatibility testing with AMBER99SB-ILDN, and its suitability was confirmed through prior benchmarks showing consistent hydrogen-bonding and dielectric properties. Post-simulation analyses included root mean square deviation (RMSD), root mean square fluctuation(RMSF), radius of gyration (Rg), solvent-accessible surface area (SASA), and hydrogen bond occupancy, which collectively assessed the complex's stability, compactness, and intermolecular interaction persistence throughout the simulation.
2.8 MM/PBSA free energy calculation
Binding free energy calculations were performed using the MM/PBSA method. The g_mmpbsa tool was used to analyze 30 snapshots from the 300 ns MD simulation. Energy terms included van der Waals, electrostatic, polar solvation, and non-polar solvation contributions. Each component was averaged across all frames. The total binding free energy provided quantitative insight into ligand affinity. Residue decomposition analysis was also conducted to identify key contributors to binding. The results were reported in kJ/mol. Compounds with the lowest binding free energy were considered to have the strongest interactions. This method validated docking results with more dynamic energy profiles. The MM/PBSA values complemented structural observations from MD simulation.
3 Results
3.1 Pharmacokinetics and toxicity of Citrus aurantium compounds
Pharmacokinetic evaluation of 111 metabolites derived from
Citrus aurantium revealed full compliance with Lipinski's Rule of Five, indicating favorable oral drug-likeness. Most compounds exhibited a bioavailability score of 0.55, suggesting moderate to good absorption potential
in vivo. Several metabolites, such as 3,4-dimethoxybenzoic acid and 4-Methoxybenzoic acid, stood out with higher scores of 0.85, enhancing their potential as drug candidates.
40 The physicochemical characteristics, including molecular weight, hydrogen bond acceptors, and lipophilicity, fell within acceptable ranges for all compounds. The majority of the metabolites were classified under toxicity class 5 based on LD
50 values, which implies low acute toxicity.
41 Terpenoids like espatulenol, limonene, and nobiletin showed high LD
50 values above 3000 mg/kg, supporting their safety. Some metabolites including indole, quinoline, and malonaldehyde were categorized as class 4, with LD
50 values ranging from 331 to 606 mg/kg, indicating moderate toxicity. A few compounds such as neohesperidin, teresantanane, and linalyl acetate reached class 6 with LD
50 values up to 12,000 mg/kg, highlighting their excellent safety profile. These results reflect structural diversity among compounds, contributing to variability in toxicity levels. The summary of pharmacokinetic and toxicity characteristics for all compounds is provided in Table 1.
The analysis of chemical subclasses revealed consistent trends in both pharmacokinetics and toxicity outcomes. Flavonoids like hesperetin, diosmetin, and naringenin derivatives were mostly in toxicity class 4, with LD50 values around 2000 mg/kg. Polymethoxylated flavones such as nobiletin and tangeretin demonstrated superior safety, consistently classified in class 5 or higher. Monoterpene alcohols like geraniol and citronellol showed LD50 values above 4000 mg/kg, maintaining class 5 classification. Aldehyde derivatives including octanal and non-adienal showed slightly lower LD50 values, yet remained within the threshold for class 5. Borderline compounds such as methyl anthranilate and isopropyl alcohol hovered near class 4 cutoffs, warranting closer evaluation. Sesquiterpenes including beta-caryophyllene and alpha-farnesene exhibited high LD50 values and acceptable drug-likeness, suggesting their potential use. Synephrine and epoxybergamottin, key bioactives in Citrus aurantium, showed both favorable ADME profiles and low toxicity. Generally, the metabolite composition reflects a balance between efficacy and safety that is desirable in candidate compounds. The data supports the advancement of several compounds for further docking and simulation analysis. These profiles confirm that Citrus aurantium harbors a wide range of bioactive compounds with promising pharmacological properties.
3.2 Identification of overlapping targets between Citrus aurantium metabolites and breast cancer
To identify potential therapeutic targets, intersecting genes between Citrus aurantium metabolites and breast cancer were determined using a Venn diagram approach. The total number of predicted targets from Citrus aurantium metabolites was 683, while 218 breast cancer-related targets were obtained from GeneCards and DisGeNET databases. Through comparative analysis, 39 genes were found to be common between these two datasets. These overlapping genes represent potential molecular links through which the phytochemicals may exert anticancer effects. The remaining 644 genes were unique to Citrus aurantium metabolites, while 179 were specific to breast cancer. The presence of overlapping targets implies potential modulation of cancer-related biological pathways. These 39 common genes serve as candidate targets for downstream molecular docking and enrichment analysis. The use of intersecting target analysis is fundamental in network pharmacology as it narrows down the biological relevance of compounds. This overlap is visually summarized in the Venn diagram presented in Fig. 1. The integration of disease and compound targets provides a focused framework for elucidating bioactive mechanisms.
The identification of these 39 intersecting targets highlights the multi-target nature of
Citrus aurantium metabolites. Such multi-target potential is aligned with the pharmacological complexity of herbal medicine. Some of the shared targets are known to play central roles in cancer progression, including genes related to proliferation, apoptosis, and angiogenesis.
42 By focusing on this intersection, researchers can prioritize genes for molecular docking and validation experiments. Moreover, network visualization and PPI analysis of these targets can further reveal key hubs in breast cancer modulation. The absence of significant overlap would have limited the mechanistic interpretation of bioactivity. However, the existence of 39 shared targets justifies continued in silico and experimental exploration. These genes form the basis for enrichment analysis in GO and KEGG databases. Their functional roles can be studied in the context of cellular localization, binding activity, and pathway regulation. This strategy enhances the rational selection of candidate molecules with potential anti-breast cancer activity.
3.3 Construction and analysis of PPI network of overlapping targets
The protein–protein interaction (PPI) network analysis was conducted to explore the interconnectivity among the 39 overlapping targets of Citrus aurantium and breast cancer. Using the STRING database with a high-confidence interaction threshold (combined score≥0.7), we generated a PPI network that was subsequently visualized and analyzed in Cytoscape. Network topology was evaluated (degree and betweenness centrality, clustering coefficient) to prioritize putative hub nodes and information-flow bottlenecks. To contextualize network biology, functional enrichment was performed for Gene Ontology (biological process, molecular function, cellular component) and KEGG pathways with FDR-adjusted p-value≤0.05; redundant terms were minimized using a similarity cutoff of 0.5, and the top non-redundant terms are reported with adjusted p-values and gene counts. The network provides insight into how these proteins functionally interact to mediate biological processes relevant to cancer. As depicted in Fig. 2, most of the proteins form a highly interconnected network with a few peripheral nodes. Central hubs in the network such as ESR1, AKT1, BCL2, and CCND1 showed higher connectivity, indicating their critical roles in signaling cascades. The dense interactions suggest that these targets may participate in coordinated pathways governing cell cycle regulation, apoptosis, and hormonal signaling. The network highlights the multitarget potential of Citrus aurantium constituents in modulating complex biological responses. Notably, several of the proteins identified are known oncogenes or tumor suppressors in breast cancer. This comprehensive network representation enhances the understanding of molecular mechanisms involved in therapeutic effects.
A closer examination of the interaction map reveals functional modules and key regulatory nodes within the system. The presence of transcription factors such as ESR1 and AR suggests hormonal pathway involvement. Anti-apoptotic proteins like BCL2 and MCL1 reinforce the relevance of this network in cancer cell survival.
43 The inclusion of CDKs(e.g., CDK2, CDK6) and cyclins (e.g., CCND1) points to the modulation of cell cycle checkpoints.
44 Other notable nodes include PI3K/AKT signaling elements, such as PIK3CA and AKT1, known to regulate proliferation and metabolic activity.
45 The functional diversity of the proteins involved indicates that the anti-breast cancer potential of
Citrus aurantium is not limited to a single pathway. Instead, it reflects a systems pharmacology approach targeting multiple biological layers. The PPI network also provides the foundation for subsequent centrality and topology analysis. Hub genes identified through degree value and betweenness metrics will guide further docking and validation studies. Altogether, the PPI network validates the therapeutic plausibility of the 39 shared targets and uncovers potential molecular mechanisms for
Citrus aurantium's anticancer activity.
3.4 Identification and visualization of hub genes in the PPI network
To further refine the PPI network analysis, a hub gene identification step was performed to highlight the most central and influential targets. This was done using network topology parameters such as degree centrality, betweenness centrality, and closeness centrality. Fig. 3 illustrates the visual clustering of these hub genes, with nodes color-coded based on their centrality scores. Red nodes such as ESR1, HDAC1, MDM2, and CCND1 represent genes with the highest degree values, indicating their critical roles in protein–protein interactions. These hub genes are likely to be essential regulators in pathways associated with breast cancer. Nodes in orange (e.g., AKT1, AR, BCL2) and yellow (e.g., PIK3CA, KDM1A, HIF1A) show slightly lower, but still significant, centrality values. Blue nodes like HIF1A suggest peripheral yet functionally relevant positions in the network. The right panel further zooms in on the top 10 hub genes to clarify their interaction density. These include key genes implicated in cell proliferation, hormonal response, and apoptosis. Identifying such hub targets enables the prioritization of downstream validation through molecular docking and experimental studies.
The connectivity patterns among these core targets provide critical insights into their cooperative biological roles. ESR1, as the most connected node, reinforces the hormonal axis's importance in breast cancer treatment. The inclusion of AKT1 and PIK3CA emphasizes the involvement of the PI3K/AKT signaling pathway, often dysregulated in breast cancer. HDAC1 and MDM2 play roles in epigenetic regulation and p53 degradation, respectively, highlighting mechanisms of transcriptional control and apoptosis evasion.
46 The presence of BCL2 as a hub gene indicates a role in apoptosis inhibition, which may contribute to cancer cell survival. AR (androgen receptor) appears connected with both hormonal and cell cycle regulators, reflecting its emerging role in certain breast cancer subtypes.
47 CCND1 and KDM1A are linked to cell cycle progression and chromatin remodeling, both crucial for unchecked cellular proliferation. The interconnectedness of these nodes suggests coordinated functional modules rather than isolated signaling events. These observations support the systems biology perspective that
Citrus aurantium metabolites may target multiple converging points in oncogenic pathways. The extracted hub genes form a tightly integrated regulatory network with potential therapeutic implications.
3.5 Topological characteristics of key target genes in the PPI network
Table 2 summarizes the top 10 key targets selected based on topological parameters within the protein–protein interaction network. These parameters include neighborhood connectivity, closeness centrality, and betweenness centrality, each reflecting different aspects of network influence. ESR1 demonstrated the highest closeness centrality value of 0.9000 and the highest betweenness centrality of 0.2477, indicating its prominent role as a central hub and mediator of communication across the network. MDM2, ranked second, showed high connectivity and centrality values, suggesting a role in regulating multiple neighboring nodes and pathways. HDAC1, AKT1, and CCND1 also showed substantial neighborhood connectivity and closeness, reinforcing their potential regulatory significance in breast cancer. Despite having high neighborhood connectivity, genes like KDM1A and PIK3CA exhibited low or zero betweenness centrality, implying they might serve more localized or peripheral roles. BCL2, involved in apoptosis inhibition, had moderate values across all parameters, supporting its involvement in cancer cell survival. Interestingly, HIF1A and AR both shared identical closeness and betweenness scores, underscoring potential functional collaboration or co-regulation. The convergence of high centrality across multiple metrics in genes like ESR1 and MDM2 strengthens their candidacy for therapeutic targeting. These results provide a quantitative basis for selecting core targets in the subsequent docking and simulation stages.
The significance of these topological metrics lies in how they inform drug target prioritization. Closeness centrality quantifies how quickly a protein can influence others within the network, making ESR1 particularly important for rapid signal propagation. Betweenness centrality highlights the role of a protein as a bridge in communication paths, and high values like that of ESR1 and MDM2 indicate regulatory bottlenecks. High neighborhood connectivity, as observed in PIK3CA and CCND1, reflects a protein's extensive interaction with adjacent nodes, implying robust local influence. While KDM1A and PIK3CA have lower centrality scores, their high connectivity suggests supportive roles within subnetwork clusters. The presence of both hormonal (ESR1, AR) and signaling(AKT1, PIK3CA) targets illustrates the multidimensional nature of Citrus aurantium's mechanism in breast cancer. These targets span hormonal, epigenetic, and apoptotic processes, reinforcing the multi-targeted intervention model suggested by network pharmacology. The ranked and structured analysis of targets ensures a rational basis for selecting molecules for further docking studies. Such parameter-based screening enhances the reliability of hub gene identification beyond mere visual interpretation. This approach validates target importance through objective network analytics. These insights collectively support the hypothesis that Citrus aurantium metabolites act on a complex, interconnected signaling web involved in breast cancer.
3.6 Functional enrichment analysis of overlapping targets
Enrichment analysis presented in Fig. 4 identifies biological processes such as regulation of programmed cell death, apoptotic signaling, and metabolic control as dominant functions of the overlapping targets. The involvement in positive regulation of cellular processes and response to chemical or organic stimuli suggests active roles in cellular stress adaptation. Other terms, including macromolecule metabolic regulation and control of cell population proliferation, reflect potential influence on tumor maintenance or regression. Molecular function terms like protein kinase activity, transcription factor binding, and nuclear receptor signaling suggest relevance in hormone-regulated cancer pathways. The presence of enzyme binding and catalytic activity further supports these targets’ engagement in dynamic biochemical interactions.
48 Within the cellular component category, genes are localized in nuclear structures including nucleoplasm and transcription regulator complexes. These locations align with known activity of many hormone-related oncogenic proteins. The gene count and significance level indicate that multiple targets share similar functional contexts, reinforcing their selection. Enrichment of chromatin-associated complexes implies potential regulation at the epigenetic level. This functional mapping offers justification for focusing molecular docking and simulation efforts on these interconnected pathways.
Additional insights are derived from KEGG pathway enrichment, where targets are clustered in critical cancer-related signaling routes. Several proteins are involved in pathways such as endocrine resistance, prostate cancer, and transcriptional dysregulation, extending the relevance of these compounds beyond breast cancer alone. Other enriched pathways include microRNAs in cancer and proteoglycans in cancer, which often act in cell–cell communication and tumor microenvironment modulation. The prominence of cellular senescence, melanoma, and chronic myeloid leukemia pathways suggests broad therapeutic applicability.
49 Signaling through thyroid hormone and transcription factor-related networks indicates that these targets span diverse cellular control points. The convergence of multiple cancer types among the enriched pathways reveals shared vulnerability points in oncogenesis. Involvement in protein interaction and regulation complexes supports the selection of compounds that bind intracellular targets rather than surface receptors. Targets localized to the nucleus and regulatory chromatin sites are consistent with nuclear modulation mechanisms. These findings align with the polypharmacological nature of natural metabolites, especially those from
Citrus aurantium. The combination of functional and pathway-based insights reinforces the relevance of selected targets for structure-based drug discovery.
3.7 Molecular docking of Citrus aurantium metabolites with ESR1
From a total of 111 metabolites of
Citrus aurantium obtained through the KNAPSACK database, only 24 compounds were identified to have predictive interactions with the estrogen receptor alpha (ESR1). This filtering process was based on target prediction results that emphasize proteins linked to breast cancer progression. As shown in Table 3, ESR1 appeared most frequently among the predicted gene targets, making it a key molecular focus. The frequent involvement of ESR1 highlights its central regulatory role in estrogen-mediated signaling in luminal-type breast cancer.
50 Some metabolites also showed interactions with other cancer-related targets such as AKT1 and PIK3CA, suggesting their broader relevance. However, ESR1 was prioritized due to its strong disease linkage and tractable structure for ligand binding. The 24 ESR1-associated metabolites consist mainly of terpenoids and polyphenolic compounds, particularly flavonoid derivatives. These molecules present structural features favorable for interaction with nuclear hormone receptors. The emphasis on ESR1 selection is consistent with the aim of identifying hormone-related therapeutic modulators. Therefore, the screening outcome provides a rational starting point for molecular docking evaluation.
Binding affinity evaluation was conducted through molecular docking simulations using AutoDock Tools 1.5.7. The docking results enabled ranking of the 24 ESR1-binding metabolites based on their binding energy and inhibition constant. As presented in Table 4, seven compounds demonstrated the most promising docking scores and nanomolar to low micromolar inhibitory constants. These top-ranked ligands included 5,7,3′,4′-tetrahydroxyflavone (−8.26 kcal/mol), hesperetin (−8.24 kcal/mol), 4′,5,7,8-tetramethoxyflavone (−7.98 kcal/mol), alpha-copaene (−7.44 kcal/mol), apigenin (−7.37 kcal/mol),5,7,4′-trimethoxyflavone (−7.40 kcal/mol), and scutellarein 5,6,7,4′-tetramethyl ether (−7.52 kcal/mol). These values indicate favorable binding interactions with ESR1, especially considering the generally accepted threshold of−6.0 kcal/mol for ligand candidacy. Among them,5,7,3′,4′-tetrahydroxyflavone showed the lowest binding energy and sub-micromolar affinity, supporting its role as a lead compound. Hesperetin and apigenin also stood out with their balance of energy efficiency and predicted bioavailability. The remaining top ligands exhibited consistent scores within a narrow range, underscoring structural similarity and binding potential. As a whole, the docking data indicate strong ligand–receptor interactions for the seven selected candidates. These results serve as the foundation for further interaction and dynamics assessments.
To further validate these docking results, the interaction profiles between the top seven compounds and ESR1 residues were investigated. The interaction maps are detailed in Table 5, showing hydrogen bonds, hydrophobic interactions, and electrostatic forces. 5,7,3′,4′-Tetrahydroxyflavone exhibited four hydrogen bonds with HIS524, GLU353, and MET421, as well as multiple Pi-alkyl and Pi–Pi T-shaped interactions with PHE404 and LEU391. Hesperetin formed hydrogen bonds with ASP351, VAL534, TYR526, and VAL533, complemented by hydrophobic interactions with TRP383, LEU354, and ALA350. Alpha-copaene, although devoid of hydrogen bonds, was stabilized through Pi-alkyl contacts with PRO324, MET537, and TRP360. Apigenin displayed strong binding through hydrogen bonding at ARG394 and GLU353, as well as π-interactions and van der Waals forces. Scutellarein 5,6,7,4′-tetramethyl ether formed mixed electrostatic and hydrophobic contacts with TRP393, GLU353, and ARG394, enhancing its binding profile.4′,5,7,8-Tetramethoxyflavone shared key interactions with TRP383 and LEU354, showing favorable alignment with conserved ESR1 residues.5,7,4′-Trimethoxyflavone exhibited similar interaction types, confirming its compatibility with the ESR1 pocket. These interaction characteristics affirm the structural suitability of these compounds as ESR1 ligands. Altogether, the presence of stable contacts with catalytically relevant residues supports further exploration.
The spatial orientation and molecular surface engagement of the best two compounds, 5,7,3′,4′-tetrahydroxyflavone and hesperetin, were visualized through docking pose analysis. As shown in Fig. 5, the top panel illustrates the binding of 5,7,3′,4′-tetrahydroxyflavone to ESR1's ligand-binding domain. The planar flavone scaffold allowed this molecule to form extended π-conjugation with aromatic residues such as PHE404 and LEU387. A network of four hydrogen bonds stabilized its orientation within the hydrophobic pocket, including contributions from HIS524 and GLU353. Hydrophobic interactions further supported docking conformation, with residues like LEU391 and ALA350 reinforcing the binding. The ligand's position was surrounded by favorable electrostatic regions, suggesting enhanced complementarity. Hesperetin adopted a bent conformation, accommodating the cavity shape of ESR1. This compound engaged TYR526, ASP351, and VAL533 in multiple hydrogen bond formations. Its phenolic groups contributed to both polar and non-polar interactions across the pocket surface. The consistent interaction geometry confirms the docking reliability for both ligands.
While 5,7,3′,4′-tetrahydroxyflavone showed the highest docking score, hesperetin demonstrated an equally favorable interaction pattern with essential ESR1 residues. The comparative analysis reveals that hesperetin's flexibility enabled it to fit tightly within the receptor's binding groove. Although hesperetin formed fewer hydrogen bonds than 5,7,3′,4′-tetrahydroxyflavone, it compensated with a broader spectrum of hydrophobic contacts. Residues such as TRP383, LEU536, and LEU354 anchored the ligand via van der Waals forces, enhancing its spatial occupancy. The electron-rich regions on hesperetin's B-ring allowed interaction with polar residues through π-stacking. This binding landscape supports hesperetin's potential as an ESR1 modulator, especially for hormone-responsive breast cancers. The visualization reinforces the notion that structural orientation and interaction distribution are critical for ligand efficacy. Furthermore, the conserved residue interactions across both ligands indicate key anchoring points on ESR1. This dual analysis of docking score and interaction quality helps refine candidate prioritization. These observations align with the hydrogen bonding and Pi-system contributions identified in the interaction profile.
The integration of gene-target prediction, docking scores, molecular interaction data, and structural modeling demonstrates the potential of Citrus aurantium metabolites as ESR1-targeted ligands. The information consistently points to a subset of polyphenolic compounds with robust pharmacological promise. Notably, 5,7,3′,4′-tetrahydroxyflavone and hesperetin show strong binding, stable interaction networks, and spatial compatibility with ESR1's binding site. These characteristics suggest their functional capacity as selective estrogen receptor modulators. Additional compounds such as apigenin, alpha-copaene, and scutellarein also merit further investigation due to their favorable docking metrics and interaction profiles. This multi-layered computational approach provides a rigorous filter for experimental validation. By narrowing candidates through network pharmacology and docking simulations, the study supports natural-product-based drug discovery against breast cancer. Continued research will involve dynamic simulation to assess conformational stability. These findings underscore the therapeutic potential of Citrus aurantium and support its development in hormone-targeted oncology.
3.8 Molecular dynamics analysis of ESR1 complexes
The RMSD results illustrated in Fig. 6 provide insight into the stability of ESR1 when complexed with different ligands across 300 ns. The ESR1–hesperetin system displayed the highest deviations, fluctuating near 0.40–0.45 nm, suggesting greater conformational adjustments. In contrast, the complex with the native ligand maintained the lowest RMSD values, stabilizing close to 0.30 nm throughout the simulation. The trajectory of 5,7,3′,4′-tetrahydroxyflavone revealed intermediate stability, averaging between 0.35 and 0.38 nm. Importantly, all systems equilibrated after ~100 ns, indicating that ESR1 maintained its overall fold despite ligand binding. The persistent low RMSD of the native ligand reflects its evolutionary fit with the receptor's binding cavity. The 5,7,3′,4′-tetrahydroxyflavone showed slightly higher variability, particularly beyond 200 ns, reflecting potential conformational rearrangements. Hesperetin, although producing higher displacements, still reached a plateau and remained dynamically stable until 300 ns. These differences emphasize that stability arises not only from minimized deviations but also from the ability to equilibrate. Furthermore, the RMSD outcomes suggest hesperetin maintains a flexible yet acceptable mode of interaction with ESR1.
As shown in the RMSF profile (Fig. 7), residue-level dynamics of ESR1 varied slightly depending on the bound ligand. Overall fluctuations for all systems remained below 0.3 nm, indicating structural conservation. Notable peaks were detected at the C-terminal region, particularly residues 525–550, which are naturally flexible. Hesperetin induced slightly higher fluctuations at residues around PHE404 and LEU525, reflecting localized mobility. By contrast, the native ligand and 5,7,3′,4′-tetrahydroxyflavone maintained more restrained residue movements in these segments. Binding pocket residues such as GLU353 and HIS524 were consistently stable, confirming their central role in ligand anchoring. The presence of 5,7,3′,4′-tetrahydroxyflavone introduced more pronounced deviations at residues LEU346, THR347, LEU387, and HIS524. Meanwhile, hesperetin caused only small shifts at key residues, which may favor its pharmacophoric fit. The similar RMSF patterns across all systems demonstrate that ESR1 retains its secondary structure despite ligand differences. Thus, RMSF data support that hesperetin binding minimizes conformational perturbations while sustaining overall stability.
Analysis of protein compactness through radius of gyration plots revealed that the ESR1–native ligand complex was the most compact, averaging 1.84–1.86 nm. The 5,7,3′,4′-tetrahydroxyflavone system showed slightly higher values around 1.85–1.87 nm, while hesperetin remained the least compact at ~1.88–1.90 nm. These findings indicate differential effects of ligands on ESR1 compactness. The SASA profiles(Fig. 8) showed 5,7,3′,4′-tetrahydroxyflavone consistently exposed more surface area (> 135 nm
2), suggesting greater solvent accessibility. In contrast, the native ligand reduced SASA values, reflecting deep insertion into the binding pocket. Hesperetin maintained intermediate solvent exposure, consistent with a more flexible but persistent binding interaction. Compactness alone did not correlate with higher stability, since 5,7,3′,4′-tetrahydroxyflavone displayed RMSD fluctuations despite maintaining low Rg. Conversely, hesperetin's less compact but equilibrated structure suggests adaptability within the binding site. The combined interpretation of Rg and SASA demonstrates that each ligand stabilizes ESR1 differently. These results indicate hesperetin achieves stable binding by balancing compactness with solvent exposure.
The hydrogen bonding patterns observed across 300 ns highlight distinct stabilization mechanisms among the ligands (Fig. 9). The 5,7,3′,4′-tetrahydroxyflavone formed up to four hydrogen bonds, indicating frequent and diverse interactions. The native ligand consistently maintained one to three hydrogen bonds, displaying steady anchoring. In contrast, hesperetin generally exhibited fewer interactions, typically one or two bonds at intermittent intervals. Despite lower frequency, hesperetin's interactions showed persistence, supporting stable anchoring throughout the trajectory. These results confirm that hydrogen bond count is not directly proportional to binding stability. The 5,7,3′,4′-tetrahydroxyflavone, although forming multiple bonds, still demonstrated RMSD fluctuations, reflecting compensatory interactions. The native ligand's consistent hydrogen bonds align with its naturally optimized binding. Hesperetin's fewer but more persistent hydrogen bonds emphasize the importance of quality and orientation over quantity. Such interaction patterns reveal different strategies by which ligands stabilize ESR1. Altogether, hydrogen bond dynamics suggest hesperetin employs a selective yet reliable binding mode.
Residue-specific hydrogen bond contributions were quantified, as summarized in Table 6. The 5,7,3′,4′-tetrahydroxyflavone exhibited the highest total occupancy (228.93%), dominated by GLU353 and HIS524, reflecting strong dependence on these residues. The native ligand showed moderate occupancy (122.62%), with GLU353 and HIS524 again serving as principal contributors. Hesperetin, in contrast, displayed markedly lower occupancy (24.38%), engaging residues such as CYS417, GLU353, and GLY420. Despite its reduced occupancy, hesperetin maintained stable binding throughout the simulation, emphasizing the sufficiency of selective contacts. The 5,7,3′,4′-tetrahy-droxyflavone's over-reliance on GLU353 may explain its structural fluctuations in RMSD, as excessive bonding suggests compensatory rather than optimal interactions. The native ligand's balanced hydrogen bonding highlights its evolutionarily refined binding behavior. Hesperetin's selective engagement strategy reflects an efficient pharmacophoric fit within ESR1. These findings demonstrate that occupancy quantity does not equate to stability or binding efficiency. Taken together, hydrogen bond occupancy analysis reinforces hesperetin's ability to sustain ESR1 stability through strategic residue interactions.
3.9 Binding free energy analysis using MM-PBSA
The MM-PBSA analysis provided quantitative insight into the thermodynamic contributions underlying ESR1–ligand interactions(Table 7). The 5,7,3′,4′-tetrahydroxyflavone complex demonstrated the most favorable binding free energy with a ΔTOTAL of−79.695 kJ/mol, strongly supported by van der Waals forces (ΔEVDW =−103.460 kJ/mol) and electrostatic energy (ΔEEEL =−81.418 kJ/mol). These negative values indicate that both hydrophobic packing and polar interactions played a major role in stabilizing the 5,7,3′,4′-tetrahydroxyflavone complex. The native ligand followed with a ΔTOTAL of−72.521 kJ/mol, driven by a large van der Waals contribution(−110.764 kJ/mol) combined with moderate electrostatic stabilization(−25.544 kJ/mol). However, its polar solvation energy (ΔEGB =74.583 kJ/mol) acted as a destabilizing factor, partially offsetting the favorable interactions. Surface energies (ΔESURF) remained relatively small for both ligands, suggesting minimal additional stabilization from non-polar solvation. These findings suggest that the strong electrostatic and hydrophobic forces of 5,7,3′,4′-tetrahydroxyflavone contribute to its superior affinity. By comparison, the native ligand showed a more balanced but slightly weaker energetic profile, consistent with its role as an optimized natural binder.
Hesperetin exhibited the weakest binding free energy among the three ligands, with a ΔTOTAL of−55.737 kJ/mol, reflecting reduced stabilizing contributions. Its van der Waals component (ΔEVDW =−90.389 kJ/mol) and electrostatic energy (ΔEEEL =−12.952 kJ/mol)were markedly smaller than those of 5,7,3′,4′-tetrahydroxyflavone and the native ligand. Despite this, hesperetin still produced a negative overall free energy, confirming that binding to ESR1 remains thermodynamically favorable. The solvation penalty (ΔEGB = 56.277 kJ/mol)was smaller compared to the other ligands, yet not sufficient to compensate for weaker attractive forces. These results suggest that hesperetin interacts with ESR1 through fewer stabilizing contacts but maintains binding stability, as supported by MD trajectories. The relatively small differences between the native ligand and 5,7,3′,4′-tetrahydroxyflavone imply that both form strong complexes, while hesperetin stabilizes ESR1 less efficiently. Importantly, the integration of MM-PBSA with dynamic results reinforces that hesperetin's stability derives more from selective residue interactions than from strong energetic contributions. Thus, MM-PBSA clarifies the energetic foundations behind the observed structural and dynamic stability of ESR1–ligand complexes.
4 Discussion
Estrogen receptor 1 (ESR1) is a vital molecular target in estrogen-dependent breast cancer due to its central role in regulating cell proliferation and survival. The selective binding of 24 metabolites from
Citrus aurantium to ESR1, as opposed to the 111 compounds initially screened, indicates strong structural compatibility. This selectivity is particularly evident among flavonoids and methoxylated compounds which often share core features with endogenous estrogens. These features include planar aromatic rings and hydroxyl substitutions that enable hydrogen bonding and pi–pi stacking interactions. Previous studies showed that 3,5,7,4′-tetrahydroxyflavone (Kaempferol), a structurally similar flavonoid, inhibited ESR1-positive breast cancer cell growth.
51–53 The presence of similar chemical scaffolds in
Citrus compounds suggests potential as selective estrogen receptor modulators. Such natural SERMs are of interest due to their dual agonist–antagonist activity with fewer adverse effects. The pharmacophore alignment with the ESR1 binding pocket explains their prioritization for docking and dynamic simulations. These findings demonstrate the relevance of
Citrus aurantium as a natural source of therapeutic agents for breast cancer. Thus, the early selection step serves both as a filter and as a validation of pharmacological relevance.
Molecular docking studies revealed favorable binding affinities for 5,7,3′,4′-tetrahydroxyflavone, hesperetin, and alpha-copaene, each demonstrating consistent interactions with critical ESR1 residues such as GLU353, HIS524, and ARG394. These amino acids are essential within the ligand-binding domain and play a central role in regulating receptor activation and coactivator recruitment. The interaction profiles identified in this study are consistent with previously reported flavonoid–ESR1 complexes, where compounds established stable hydrogen bonds and hydrophobic contacts at similar residue positions. Notably, prior analyses involving structurally related flavones also described π-stacking and hydrogen bonding interactions at GLU353 and HIS524, indicating a conserved binding mode.
54,55 In the present study, 5,7,3′,4′-tetrahydroxyflavone and hesperetin exhibited superior ligand efficiency compared to alpha-copaene, underscoring the importance of hydroxyl and methoxy groups in enhancing electrostatic compatibility. These functional groups contribute to more persistent receptor engagement and greater conformational stabilization. The similarities between the current results and earlier docking investigations affirm the reliability of the docking protocol and support the predicted biological relevance of the interactions. Variability in interaction types, including both polar and nonpolar contacts, also suggests that each compound may influence ESR1 through distinct modulation mechanisms.
56 Among the tested ligands, 5,7,3′,4′-tetrahydroxyflavone displayed the deepest pocket engagement and the most stable docking conformation. By contrast, hesperetin formed fewer key interactions and was positioned more superficially within the binding site. This distinction was further confirmed by MM/PBSA calculations, which showed that the most effective ligands derived their stability primarily from van der Waals and electrostatic forces.
Molecular dynamics simulations validated the structural stability and dynamic interactions of
Citrus aurantium metabolites with ESR1 over the 300-ns trajectory. The RMSD and RMSF analyses showed that the native ligand maintained the most consistent binding pose with minimal deviation, while 5,7,3′,4′-tetrahydroxyflavone exhibited intermediate fluctuations and hesperetin displayed higher displacements, though it equilibrated after 100 ns. Compactness and solvent accessibility analysis revealed that the native ligand complex was the most compact, tetrahydroxyflavone demonstrated greater solvent exposure, and hesperetin showed moderate structural adaptation. RDF confirmed dense atomic clustering for tetrahydroxyflavone around the receptor pocket, indicating persistent spatial occupancy. Hydrogen bond analysis indicated that tetrahydroxyflavone engaged ESR1 with the highest number and occupancy, particularly through GLU353 and HIS524, while hesperetin formed fewer but selective interactions with residues such as CYS417 and GLU353. These dynamic patterns are consistent with prior molecular dynamics reports, where stable flavonoid–receptor interactions were correlated with improved pharmacological efficacy.
57,58 This was further supported by MM-PBSA analysis, which showed 5,7,3′,4′-tetrahydroxyflavone achieving the most favorable binding free energy (−79.695 kJ/mol), followed by the native ligand(−72.521 kJ/mol), while hesperetin displayed the weakest affinity(−55.737 kJ/mol). The stronger van der Waals and electrostatic contributions in 3,5,7,4′-tetrahydroxyflavone highlight its enhanced thermodynamic stabilization, whereas hesperetin's lower values explain its reduced binding energy despite dynamic stability. Taken together, these findings suggest that 3,5,7,4′-tetrahydroxyflavone and hesperetin remain viable ESR1 binders, with the former showing superior thermodynamic favorability and the latter offering selective but stable pharmacophoric interactions.
5 Conclusions
In conclusion, this study highlighted Citrus aurantium-derived flavonoids as promising ESR1 modulators relevant to breast cancer therapy. Molecular dynamics analyses revealed that the native ligand maintained the most stable trajectory, 3,5,7,4′-tetrahydroxyflavone exhibited intermediate fluctuations with strong receptor engagement, and hesperetin showed higher deviations yet equilibrated after 100 ns. RMSF, Rg, and SASA analyses confirmed that tetrahydroxyflavone maintained persistent atomic clustering and solvent exposure, while hesperetin provided moderate adaptation supported by selective residue contacts. Hydrogen bond and occupancy studies showed that 3,5,7,4′-tetrahydroxyflavone engaged ESR1 through sustained interactions at GLU353 and HIS524, whereas hesperetin relied on fewer but targeted bonds at residues like CYS417 and GLU353. Importantly, MM-PBSA calculations indicated that 3,5,7,4′-tetrahydroxyflavone achieved the most favorable binding free energy (−79.695 kJ/mol), followed by the native ligand (−72.521 kJ/mol), while hesperetin displayed the weakest affinity (−55.737 kJ/mol). These results demonstrate that 3,5,7,4′-tetrahydroxyflavone benefits from strong van der Waals and electrostatic stabilization, while hesperetin achieves binding through efficient pharmacophoric fit despite weaker energetics. The hydroxyl-rich aromatic structures of both flavonoids contributed to selective ESR1 engagement, consistent with prior studies linking flavonoid–receptor stability to pharmacological potential. Together, these findings provide a solid foundation for experimental validation and future development of Citrus aurantium metabolites as potential therapeutic agents in estrogen receptor-positive breast cancer.
2026 Publishing services by Elsevier B.V. on behalf of Higher Education Press and KeAi Communications Co. Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).