Structure-based computational screening of cyclotides as putative CYP19A1-binding scaffolds for modulation of estrogen biosynthesis in adenomyosis

Other OA: gold public-domain-us
⚙ AI-generated summary by qwen3.7-flash, 2026-11-20 ⓘ

Computational screening identified cycloviolacin O4 as a stable CYP19A1-binding cyclotide scaffold for modulating estrogen biosynthesis in adenomyosis, though biochemical and in vivo validation is required.

One-sentence paraphrase of the abstract; not a substitute for reading it. No clinical advice. How this works

⚙ AI-generated deep summary by qwen3.7-flash, 2026-11-20 · read from full text ⓘ

This study employed structure-based computational screening to evaluate plant-derived cyclotides, specifically cycloviolacin O4, as potential inhibitors of aromatase (CYP19A1), the enzyme responsible for local estrogen biosynthesis. Using molecular docking and dynamics simulations, the researchers assessed the binding stability and structural compatibility of these cyclic peptides with the aromatase active site, aiming to identify novel scaffolds that could modulate estrogen production without the side effects of traditional small-molecule inhibitors. The paper explicitly states that this in silico work is intended to prioritize candidates for future experimental investigation rather than to establish biological efficacy or therapeutic activity. This paper is centrally about adenomyosis — it investigates a mechanism-based drug discovery approach targeting aromatase expression, which is a key driver of the disease's estrogen-dependent pathology.

Read from the paper's body, not the abstract. Not a substitute for reading the paper. No clinical advice. How this works

Abstract

Adenomyosis is a chronic estrogen-dependent gynecological disorder associated with dysmenorrhea, abnormal uterine bleeding, chronic pelvic pain, infertility, and recurrent implantation failure. Because human placental aromatase cytochrome P450 (CYP19A1) plays a central role in local estrogen biosynthesis, this study employed an integrated in silico workflow to identify cyclotide scaffolds capable of interacting with CYP19A1. The receptor structure was prepared and validated, while protein-protein interaction analysis confirmed its involvement in steroidogenesis and estrogen biosynthesis. Twenty cyclotides were modeled and screened against the predicted active site of CYP19A1. Based on an integrated evaluation of HADDOCK and ClusPro docking, binding orientation, and interaction characteristics, cycloviolacin O4 was prioritized as the lead candidate. The predicted CYP19A1-cycloviolacin O4 complex exhibited a HADDOCK score of -82.8 ± 9.8, a cluster RMSD of 0.2 ± 0.1 Å, and a ClusPro energy score of -872.1. Docking against the reference inhibitor exemestane and the reference peptide SNA-60-367-10 provided additional computational benchmarking. Subsequent 500 ns molecular dynamics simulations, normal mode analysis, and MM-GBSA/MM-PBSA calculations supported the predicted structural stability and dynamic compatibility of the complex, yielding binding free energies of -66.58 ± 3.46 kcal/mol (MM-GBSA) and -4.56 ± 4.29 kcal/mol (MM-PBSA). Computational safety assessment predicted cycloviolacin O4 to be non-antigenic, non-allergenic, and non-toxic, while immune simulation suggested a balanced immune response. Collectively, these findings identify cycloviolacin O4 as a computationally predicted CYP19A1-binding cyclotide scaffold. However, these observations are derived exclusively from computational analyses and require biochemical, cellular, and in vivo validation before any conclusions regarding biological activity or therapeutic potential can be drawn.
Full text 83,757 characters · extracted from pmc-nxml · 5 sections · click to expand

Intro

Adenomyosis is an increasingly recognized estrogen-dependent uterine disorder characterized by the infiltration of endometrial glands and stroma into the myometrium. Despite increasing clinical recognition, the disease remains underdiagnosed and therapeutic options remain limited, particularly for women seeking fertility preservation. Its epidemiology is difficult to quantify precisely because historical estimates were based on hysterectomy specimens, whereas contemporary prevalence depends heavily on imaging criteria, population selection, and whether coexisting fibroids or endometriosis are counted [ 1 ]. Even so, modern reviews indicate that adenomyosis is not a rare condition: reported prevalence in reproductive-age women can range from approximately 12% to 58%, hysterectomy-based series commonly report substantial disease burden, and a 2023 meta-analysis estimated a pooled prevalence of isolated adenomyosis of about 10% among women with subfertility [ 2 ]. Clinically, the disorder is strongly associated with heavy menstrual bleeding, dysmenorrhea, chronic pelvic pain, dyspareunia, infertility, miscarriage, and adverse obstetric outcomes, thereby producing a considerable burden on quality of life, reproductive planning, and health-care utilization. The condition has also been described recently as an underacknowledged cause of abnormal uterine bleeding and pelvic pain, underscoring the need for disease-focused therapeutic innovation rather than continued symptom-only management [ 3 ]. Current therapeutic strategies for adenomyosis provide symptomatic benefit but remain limited, particularly for women seeking fertility preservation. Medical management is centered on symptom suppression, using nonsteroidal anti-inflammatory drugs for pain, levonorgestrel-releasing intrauterine systems and other progestin-based strategies for bleeding and dysmenorrhea, and gonadotropin-releasing hormone (GnRH) agonists or antagonists for more refractory disease [ 4 ]. However, evidence remains heterogeneous, lesion regression is inconsistent across therapies, and long-term hormonal suppression may be constrained by tolerability, hypoestrogenic symptoms, and concern for bone effects, particularly with prolonged GnRH-based approaches [ 5 ]. Surgically, hysterectomy remains the definitive treatment for severe symptomatic adenomyosis, but it permanently abolishes fertility and is therefore unacceptable for many younger patients. Uterus-sparing surgery can improve pain and bleeding in selected cases, yet it is technically demanding, carries meaningful recurrence risk, and in reproductive settings may be followed by uterine rupture or abnormal placentation in subsequent pregnancy [ 6 ]. Importantly, a recent EBioMedicine editorial emphasized that, despite the magnitude of the clinical burden, no approved adenomyosis-specific therapies currently exist, which sharpens the rationale for mechanism-based drug discovery in this disease. These therapeutic limitations make local estrogen biosynthesis an especially compelling mechanistic target. CYP19A1, commonly known as aromatase, is the cytochrome P450 enzyme responsible for the terminal and indispensable step in estrogen biosynthesis, catalyzing the aromatization of C19 androgens into C18 estrogens [ 7 ]. Structural biology has made this target unusually attractive for rational inhibitor design: the crystal structure of human placental aromatase in complex with androstenedione resolved the architecture of an androgen-specific active site and provided a direct template for structure-guided ligand discovery [ 8 ]. This structural insight is especially relevant to adenomyosis because aberrant aromatase expression has long been demonstrated in adenomyotic tissue and in eutopic endometrium from affected women, whereas normal endometrium lacks comparable expression, supporting the concept of autonomous local estrogen production within disease tissue [ 9 ]. Contemporary pathophysiologic reviews further emphasize that adenomyosis is characterized by a local hyperestrogenic milieu driven by increased estrogen synthesis, including aromatase-dependent mechanisms, together with disrupted estrogen-progesterone signaling; this endocrine microenvironment is thought to sustain lesion survival, inflammation, fibrosis, and impaired endometrial receptivity [ 10 ]. Thus, CYP19A1 is not merely a general steroidogenic enzyme but a biologically relevant nodal driver of adenomyosis-associated estrogen excess [ 11 ]. Clinical observations also lend credibility to the idea that aromatase inhibition could be therapeutically meaningful in adenomyosis. In a randomized controlled trial, letrozole was reported to be as effective as the GnRH agonist goserelin in reducing adenomyoma volume and improving symptoms in premenopausal women with uterine adenomyosis [ 12 ]. More recently, a prospective pilot study found that low-dose letrozole significantly reduced dysmenorrhea, menorrhagia, and sonographic disease severity in symptomatic women with adenomyosis awaiting in vitro fertilization, suggesting that aromatase suppression may have both symptom-modifying and reproductive-context relevance [ 13 ]. Reviews of current adenomyosis management now routinely cite aromatase inhibitors as promising mechanistically grounded options, while simultaneously acknowledging that their role is still limited by the relative scarcity of larger long-term trials and definitive fertility-outcome data. This combination of strong biologic rationale, validated target structure, and encouraging early clinical evidence makes human placental aromatase a credible receptor for exploration in silico, particularly when the aim is to identify alternative inhibitory chemotypes beyond classical small-molecule aromatase inhibitors. Within this context, medicinal plants and especially plant-derived cyclotides offer a compelling reservoir of bioactive scaffolds. Cyclotides are small macrocyclic peptides, typically 28–37 amino acids in length, defined by a head-to-tail cyclic backbone and a cystine-knot arrangement of three conserved disulfide bonds; this cyclic cystine knot confers extraordinary resistance to thermal, chemical, and proteolytic degradation, a highly desirable property for peptide drug discovery [ 14 ]. Cyclotides have been discovered across several plant families, most notably Violaceae and Rubiaceae, but also Cucurbitaceae, Fabaceae, and Solanaceae, often with dozens of related peptides in a single species. Their ethnopharmacologic roots are equally important: the discovery of the field traces to Oldenlandia affinis (Rubiaceae), a plant used in African traditional medicine as a uterotonic tea to facilitate childbirth, from which the prototypic cyclotide kalata B1 was isolated [ 15 ]. Among medicinal violets, Viola odorata has a long history of use in Persian, Greco-Arab, Ayurvedic, and Unani traditions for cough, sore throat, bronchitis, fever, headache, and inflammatory conditions, placing cyclotide-rich violets at an interesting intersection between ethnomedicine and modern peptide pharmacology [ 16 ]. Cycloviolacin O4 is one of the bracelet cyclotides catalogued from V. odorata , and it has also been identified from Pombalia calceolaria , where published experimental data showed cytotoxic activity against the triple-negative breast cancer cell line MDA-MB-231, with an IC₅₀ of 9.8 μM [ 17 ]. Although cycloviolacin O4 remains less extensively investigated than the prototype cycloviolacin O2, the broader cycloviolacin family is well known for membrane-disruptive, bactericidal, cytotoxic, and chemosensitizing behavior, suggesting that these peptides can engage biologically meaningful targets and survive demanding physicochemical environments. Collectively, these features make cycloviolacin O4 a rational plant-derived peptide scaffold to evaluate against CYP19A1 [ 18 ]. Because peptide–protein recognition is highly sensitive to conformation and dynamics, structure-based computer-aided drug design offers a particularly useful strategy for early-stage lead selection in this setting. Contemporary structure-based drug discovery frameworks use docking to propose binding poses and prioritize plausible binders, while molecular dynamics simulations provide atomistic information on binding stability, flexible site adaptation, and interaction persistence in explicit solvent [ 19 ]. For peptide–protein systems, this is especially important because peptides can mediate a large fraction of protein interactions but remain harder to model than small molecules owing to their flexibility and the complexity of receptor adaptation [ 20 ]. Normal mode analysis further complements docking and molecular dynamics by revealing low-frequency collective motions and flexible states accessible around an equilibrium conformation, thereby supporting interpretation of conformational adaptability at relatively low computational cost [ 21 ]. For peptide leads, in silico immunogenicity assessment is also a critical first-pass filter because T-cell epitope prediction provides an early estimate of HLA presentation and immune risk, while modern toxicity-prediction tools increasingly integrate sequence and structural information to triage unsafe candidates prior to experimental testing [ 22 ]. Accordingly, the aim of the present study was to computationally identify and characterize a cyclotide as a potential CYP19A1-binding scaffold using an integrated in silico workflow. The study was intended to prioritize candidates for future experimental investigation rather than to establish biological efficacy or therapeutic activity [ 23 ].

Results

The three-dimensional structure of human placental aromatase cytochrome P450 (CYP19A1) was successfully retrieved from the Protein Data Bank under PDB ID 3S79 ( Fig 1A ). The receptor structure was subsequently prepared and purified for downstream computational analyses through removal of unnecessary structural components and optimization of the protein conformation for receptor-ligand interaction studies. Structural validation analysis demonstrated that the prepared CYP19A1 receptor possessed acceptable stereochemical quality and structural reliability for computational investigations. The ERRAT quality assessment generated an overall quality factor of 97.2851, indicating a highly reliable protein structure with favorable non-bonded atomic interactions and minimal structural inconsistencies ( Fig 1B ). Furthermore, Ramachandran plot analysis demonstrated that 86.6% of amino acid residues were positioned within the most favored regions, while 12.1% of residues were located in additionally allowed regions. Only 1.0% of residues were observed in generously allowed regions and 0.2% in disallowed regions, suggesting overall acceptable backbone geometry and conformational stability of the receptor structure. Collectively, these validation findings indicated that the CYP19A1 receptor structure was structurally suitable for subsequent active site prediction, molecular docking, interaction profiling, and molecular dynamics simulation analyses ( Fig 1C ). Protein-protein interaction network analysis demonstrated that CYP19A1 exhibited strong functional associations with multiple proteins involved in steroid hormone biosynthesis, androgen metabolism, estrogen regulation, and endocrine signaling pathways. The interaction network revealed a highly interconnected molecular profile surrounding CYP19A1, supporting its central biological role in estrogen biosynthesis and hormone-associated metabolic regulation relevant to adenomyosis pathophysiology. The generated interaction network identified several closely associated proteins, including CYP17A1, HSD17B1, HSD17B2, HSD17B3, HSD17B7, HSD3B1, HSD3B2, SRD5A1, SRD5A2, and SULT1E1. These proteins are functionally associated with steroidogenesis, androgen conversion, estrogen metabolism, and hormonal homeostasis. The observed interactions suggested that CYP19A1 operates within a coordinated steroid metabolic network involving multiple hydroxysteroid dehydrogenases and steroid-converting enzymes ( Fig 2 ). Among the identified interacting partners, CYP17A1 demonstrated a particularly important association due to its role in androgen precursor synthesis, which directly contributes to substrate availability for aromatase-mediated estrogen formation. Similarly, interactions with HSD17B family proteins indicated functional involvement in estradiol and estrone interconversion pathways, whereas associations with SRD5A1 and SRD5A2 highlighted potential linkage with androgen metabolic regulation. Interaction with SULT1E1 further suggested involvement in estrogen sulfation and hormone inactivation mechanisms. Overall, the PPI network analysis supported the biological relevance of CYP19A1 as a therapeutically important target associated with estrogen biosynthesis and endocrine pathway regulation. The highly connected interaction profile observed within the steroidogenic network further strengthened the rationale for investigating cyclotide-based inhibitors targeting CYP19A1 in the context of adenomyosis-associated estrogen modulation. A total of twenty cyclotides were successfully retrieved and structurally modeled for downstream computational analysis against human placental aromatase cytochrome P450 (CYP19A1). The selected cyclotides included cycloviolacin O1, cycloviolacin O2, cycloviolacin O3, cycloviolacin O4, cycloviolacin O5, cycloviolacin O6, cycloviolacin O7, cycloviolacin O11, cycloviolacin O12, cycloviolacin O13, cycloviolacin O14, cycloviolacin O15, cycloviolacin O16, cycloviolacin O17, cycloviolacin O18, cycloviolacin O19, cycloviolacin O20, cycloviolacin O21, cycloviolacin O22, and cycloviolacin O23. These cyclotides represented structurally diverse cyclic peptides possessing characteristic cysteine-rich frameworks and compact peptide conformations suitable for receptor interaction analysis. Three-dimensional structural modeling generated stable peptide conformations for all selected cyclotides with acceptable predicted confidence metrics. The obtained predicted template modeling scores indicated reliable structural folding behavior for the majority of the peptide models. The predicted ipTM values ranged from 0.54 to 0.63, suggesting overall acceptable structural confidence and conformational consistency among the modeled cyclotides. Several cyclotides, including cycloviolacin O6, cycloviolacin O11, and cycloviolacin O19, demonstrated comparatively higher predicted confidence values of 0.63, whereas cycloviolacin O12 and cycloviolacin O14 exhibited relatively lower but still acceptable confidence values of 0.54 ( S1 Table ). The predicted structures exhibited acceptable confidence (ipTM 0.54–0.63), supporting their suitability for subsequent docking analyses ( S1 Table ). The compact folding architecture and predicted conformational stability of these peptides suggested their suitability for subsequent protein-peptide docking and molecular dynamics investigations against CYP19A1. The finalized three-dimensional peptide models were therefore utilized for active site interaction analysis, docking evaluation, and computational stability assessment in the present study. Active site prediction analysis identified multiple potential binding pockets within the structure of human placental aromatase cytochrome P450 (CYP19A1). A total of two pockets were recognized during the prediction, as shown in Fig 3 . Among the predicted cavities, Pocket 1 was identified as the most probable and biologically relevant binding region for downstream receptor-peptide interaction analysis. This pocket demonstrated the highest prediction score of 48.59 along with a probability value of 0.974, indicating a highly reliable and structurally favorable binding cavity within the receptor ( Table 1 ). Pocket 1 (red) represents the highest-ranked predicted binding cavity and includes the experimentally characterized catalytic and substrate-recognition residues Asp309, Thr310, and Arg435 of the CYP19A1 active site, whereas Pocket 2 (yellow) showed substantially lower prediction confidence. Pocket 1 comprised residues A107, A115, A132–A134, A145, A148, A152, A199, A203, A221, A224, A302, A303, A306, A307, A309–A311, A314, A364, A369, A370, A372–A375, A429–A431, A435–A439, A442, A443, A446, A447, A477, and A478. Comparison with the experimentally characterized androstenedione-binding region of the 3S79 crystal structure showed that Pocket 1 contained Asp309, Thr310, and Arg435. Thus, all three catalytic and substrate-recognition residues highlighted for CYP19A1 were included within the highest-ranked predicted cavity. Asp309 and Thr310 are located within the conserved I-helix region associated with substrate positioning and oxygen-mediated catalysis, whereas Arg435 contributes to recognition and stabilization of the steroid 3-keto group. The inclusion of these residues indicates that Pocket 1 substantially overlaps the native androstenedione-binding region rather than representing an unrelated surface cavity. This correspondence, together with the high PrankWeb probability of 0.974, supported the selection of Pocket 1 for the subsequent docking analysis. Furthermore, comparison with the co-crystallized androstenedione present in the 3S79 crystal structure confirmed that the highest-ranked predicted pocket spatially corresponds to the experimentally established substrate-binding region, thereby supporting the biological relevance of the selected docking site. In comparison, Pocket 2 exhibited a substantially lower prediction score of 4.63 with a probability value of 0.213, indicating comparatively reduced binding confidence and structural suitability. This secondary pocket involved residues A354, A357, A358, A361, A418, A420–A422, A424, A427–A430, A440, A441, and A444. Based on the comparatively higher confidence score, residue distribution, and predicted binding reliability, Pocket 1 was selected as the primary docking region for subsequent protein-peptide interaction analyses. Molecular docking analysis demonstrated variable binding behavior among the screened cyclotides against the active region of human placental aromatase cytochrome P450 (CYP19A1). Comparative evaluation of docking scores indicated that several cyclotides exhibited favorable interaction energies and structurally stable docking conformations within the predicted binding cavity of the receptor. Among all screened peptides, cycloviolacin O4 demonstrated one of the most favorable docking profiles, with a HADDOCK score of −82.8 ± 9.8 and a low RMSD value of 0.2 ± 0.1 Å, suggesting stable clustering behavior and reliable docking orientation. Cross-validation analysis further supported the interaction potential of cycloviolacin O4 with a ClusPro energy score of −872.1. Similarly, cycloviolacin O6 and cycloviolacin O11 also exhibited favorable docking profiles with HADDOCK scores of −84.9 ± 12.7 and −84.7 ± 3.9, respectively, accompanied by low RMSD values of 0.2 ± 0.1 Å. Notably, cycloviolacin O11 demonstrated the most favorable ClusPro energy score of −897.8 among all analyzed cyclotides, indicating strong predicted interaction stability with CYP19A1 ( Table 2 ). Several additional cyclotides, including cycloviolacin O5, cycloviolacin O7, cycloviolacin O14, cycloviolacin O17, cycloviolacin O18, and cycloviolacin O23, also demonstrated comparatively favorable docking energies, suggesting potential interaction compatibility with the receptor binding region. In contrast, cycloviolacin O1, cycloviolacin O21, and cycloviolacin O22 exhibited comparatively less favorable docking energies, indicating relatively reduced predicted binding stability within the CYP19A1 binding cavity. RMSD-based cluster analysis further suggested that the majority of top-performing cyclotides maintained stable conformational clustering during docking, particularly cycloviolacin O4, O6, O11, O16, O17, and O22, all of which demonstrated low RMSD deviations close to 0.2 Å. To further assess the robustness of the docking protocol, the selected lead cyclotide was compared with two reference ligands, including the clinically approved aromatase inhibitor exemestane and the reported control peptide SNA-60-367-10. As summarized in Table 2 , exemestane produced a HADDOCK score of −22.9 ± 2.2 with an interface RMSD of 0.1 ± 0.0 Å, whereas SNA-60-367-10 yielded a more favorable HADDOCK score of −59.4 ± 3.2 with an interface RMSD of 1.0 ± 0.6 Å. Independent validation using ClusPro generated docking energies of −678.1 and −720.8, respectively. These reference docking analyses provide additional computational benchmarks for interpreting the predicted interaction profile of cycloviolacin O4 within the CYP19A1 binding pocket. Although the ligands differ substantially in molecular size and chemical class, inclusion of both a clinically established inhibitor and a reported peptide control enhances the overall reliability of the docking workflow and facilitates broader assessment of the predicted receptor–ligand interactions. Overall comparative analysis of docking energies, RMSD values, and cross-validation scores suggested that cycloviolacin O4 exhibited a balanced and computationally favorable interaction profile against CYP19A1. Although cycloviolacin O6 and cycloviolacin O11 demonstrated slightly stronger HADDOCK and ClusPro energies, cycloviolacin O4 displayed highly stable docking behavior with consistent interaction characteristics, leading to its prioritization as the lead cyclotide candidate for downstream molecular dynamics simulations and advanced computational analyses. Cycloviolacin O4 adopted a well-defined binding orientation within the active cavity of CYP19A1, demonstrating favorable spatial accommodation and structural complementarity with the receptor binding environment ( Fig 4A ). Detailed interaction analysis revealed multiple intermolecular contacts between cycloviolacin O4 and key amino acid residues lining the predicted binding pocket, supporting stable receptor–peptide recognition ( Fig 4B ). The interaction map demonstrated multiple hydrogen bonding interactions and hydrophobic contacts contributing to stabilization of the docked complex. Several active site residues surrounding the binding cavity were observed to participate in intermolecular interactions with the lead cyclotide, supporting the predicted receptor-peptide binding stability obtained from docking analysis. The combined three-dimensional ( Fig 4C ) and two-dimensional interaction analyses collectively suggested that cycloviolacin O4 may establish stable computational interactions within the CYP19A1 active region relevant to estrogen biosynthesis modulation. To further investigate the dynamic stability and conformational behavior of the prioritized CYP19A1–cycloviolacin O4 complex under simulated physiological conditions, a 500 ns molecular dynamics simulation analysis was performed. Although several cyclotides, including cycloviolacin O6 and cycloviolacin O11, produced competitive docking scores, candidate prioritization was not based exclusively on docking energy. The final selection considered the overall quality of the docked complexes, including binding orientation within the predicted active site, interaction consistency with catalytically relevant residues, agreement between HADDOCK and ClusPro docking analyses, and the suitability of the complex for subsequent molecular dynamics investigation. Based on this integrated evaluation, cycloviolacin O4 demonstrated the most favorable overall receptor–ligand interaction profile and was therefore selected as the lead candidate for detailed molecular dynamics simulations and downstream computational analyses. The majority of amino acid residues demonstrated relatively low fluctuation behavior, indicating overall structural stability of the receptor throughout the simulation trajectory ( Fig 5A ). Most residues remained below approximately 1.5 Å fluctuation, suggesting limited local flexibility across the protein backbone. However, a few terminal and loop-associated regions exhibited comparatively elevated fluctuations, particularly near the N-terminal region and several distal loop segments, where fluctuations approached approximately 3.0–4.0 Å. These localized flexible regions may reflect naturally mobile surface-exposed segments rather than global instability of the complex. Importantly, the active site-associated regions demonstrated comparatively controlled fluctuations, suggesting that the binding of cycloviolacin O4 maintained stable local receptor conformational behavior during the simulation. The principal component analysis (PCA) trajectory of the simulated complex was projected along the first three principal components. The PCA distribution demonstrated gradual conformational transitions across the simulation trajectory, reflecting the intrinsic dynamic motion of the receptor-peptide complex. The clustering pattern observed during later simulation stages suggested attainment of relatively stable conformational states following the initial conformational adaptation period. The gradual shift in conformational sampling indicated that the complex underwent controlled structural rearrangements rather than abrupt destabilizing motions during simulation ( Fig 6A ). The protein backbone RMSD initially increased during the equilibration phase and subsequently stabilized with relatively consistent fluctuations throughout the remaining simulation period, indicating maintenance of overall structural integrity of the receptor complex ( Fig 5B ). The ligand RMSD exhibited moderate conformational adjustment during the early stages of simulation followed by stable trajectory behavior, suggesting sustained accommodation of cycloviolacin O4 within the receptor binding environment. Similarly, active site residue RMSD values remained comparatively controlled during the simulation, indicating preservation of the receptor binding cavity architecture during peptide interaction. Collectively, the RMSD analysis suggested that the CYP19A1–cycloviolacin O4 complex achieved a relatively stable dynamic state following initial equilibration. The Rg values remained within a relatively narrow fluctuation range throughout the trajectory, indicating maintenance of overall structural compactness of the complex ( Fig 5C ). Although moderate fluctuations were observed during intermediate simulation intervals, no major structural expansion or collapse behavior was detected. A slight increase in compactness variation was observed around the mid-simulation region; however, the system subsequently regained relatively stable conformational behavior. These observations suggested that the receptor-peptide complex preserved an overall compact structural arrangement during long-timescale simulation. Persistent hydrogen-bond interactions were observed during the simulation, indicating continuous receptor-peptide contact across the trajectory. The number of intermolecular hydrogen bonds fluctuated during the simulation, with the rolling mean generally remaining within approximately 8–15 hydrogen bonds. The persistence of these interactions suggested maintenance of stable intermolecular binding throughout the simulation despite normal dynamic fluctuations, supporting continuous association between cycloviolacin O4 and the CYP19A1 binding site ( Fig 6C ). These fluctuations are expected during dynamic simulations and reflect continuous formation and disruption of hydrogen-bond interactions rather than loss of binding. The persistent intermolecular hydrogen-bonding interactions support the stable association of cycloviolacin O4 with the CYP19A1 binding region throughout the simulation period. Only hydrogen bonds formed directly between CYP19A1 and cycloviolacin O4 were considered in this interpretation, whereas intramolecular hydrogen bonds within the receptor or peptide were not included. Apo protein (receptor). The dynamic cross-correlation matrix (DCCM) analysis of residue motion correlations within the CYP19A1–cycloviolacin O4 complex demonstrated the presence of both positively and negatively correlated residue motions distributed across the receptor structure. Several regions exhibited coordinated motion behavior, indicating dynamic communication among structurally associated residues during simulation. The predominance of moderate positive correlation patterns suggested relatively synchronized residue movements contributing to maintenance of structural stability within the receptor-peptide complex. In contrast, limited negatively correlated motions were observed in certain distal regions, potentially reflecting localized compensatory structural adjustments during the dynamic simulation process. Overall, the DCCM analysis suggested coordinated internal residue dynamics within the CYP19A1–cycloviolacin O4 complex during the 500 ns simulation trajectory ( Fig 6B ). To evaluate the intrinsic conformational stability and dynamic behavior of the unbound receptor, a separate 500 ns molecular dynamics simulation of apo CYP19A1 was performed. Fig 7A illustrates the root mean square deviation (RMSD) profile of the apo CYP19A1 receptor during the 500 ns simulation trajectory. During the initial simulation phase, the receptor maintained relatively stable RMSD values with moderate fluctuations, indicating an equilibrated structural state. However, a noticeable increase in RMSD was observed during the later stages of simulation, particularly after approximately 300 ns, where the receptor adopted a comparatively elevated conformational state with persistent fluctuations. Despite this structural transition, the receptor maintained an overall stable trajectory without evidence of abrupt destabilization or excessive structural deviation. The observed conformational shift may reflect intrinsic dynamic flexibility of the unbound CYP19A1 receptor in the absence of stabilizing ligand interactions. Residue-level flexibility analysis demonstrated limited fluctuations throughout most of the receptor, while the slightly higher mobility observed in terminal and surface-exposed loop regions reflected intrinsic protein flexibility rather than instability of the binding interface ( Fig 7B ). However, certain regions exhibited comparatively elevated fluctuations, particularly around the mid-residue region where fluctuations approached approximately 3.5 Å. These localized fluctuations may correspond to flexible loop or surface-exposed regions contributing to the natural conformational adaptability of the receptor. Overall, the RMSF profile suggested that the apo receptor retained acceptable structural integrity despite the presence of several dynamically flexible regions. The radius of gyration of the apo receptor remained within a relatively narrow range during the initial part of the trajectory, indicating maintenance of overall structural compactness ( Fig 7C ). A noticeable increase in Rg values was observed after approximately 300 ns, indicating a transient reduction in structural compactness associated with conformational rearrangement of the receptor structure. Although the receptor later demonstrated partial stabilization, the apo protein continued to exhibit comparatively higher compactness fluctuations than typically observed in ligand-stabilized systems. These observations suggested that the absence of ligand interaction may permit greater structural flexibility within the CYP19A1 receptor. Principal component analysis demonstrated that the apo receptor sampled a comparatively broad conformational space during the trajectory ( Fig 7D ). Distinct cluster separation observed within the trajectory suggested that the apo receptor underwent noticeable conformational transitions during simulation. The wider conformational spread further supported the relatively dynamic nature of the unbound receptor structure. Dynamic cross-correlation analysis revealed coordinated residue motions within the receptor, suggesting preservation of the native dynamic communication network within the unbound receptor ( Fig 7E ). The correlation heatmap demonstrated the presence of both positively and negatively correlated residue movements distributed throughout the receptor structure. Several regions exhibited coordinated residue motion patterns, whereas additional areas showed anti-correlated dynamic behavior, suggesting structural communication between distant regions of the receptor. The comparatively diverse correlation pattern observed in the apo receptor may reflect increased intrinsic flexibility and dynamic adaptability in the absence of ligand-mediated stabilization. Overall, the DCCM analysis suggested complex internal motion behavior within the unbound CYP19A1 receptor during long-timescale simulation. To further evaluate the independent conformational stability and intrinsic dynamic behavior of the lead cyclotide, a separate 500 ns molecular dynamics simulation of cycloviolacin O4 was performed. Fig 8A presents the RMSD profile of cycloviolacin O4 throughout the 500 ns simulation period. During the initial simulation phase, the ligand maintained relatively stable RMSD values with moderate fluctuations, indicating preservation of the peptide conformation under solvated conditions. A noticeable conformational transition was observed after approximately 300 ns, where the ligand adopted a comparatively elevated RMSD state followed by sustained fluctuations around the newly stabilized conformational region. Despite this structural transition, the peptide maintained an overall equilibrated trajectory without evidence of major destabilization or structural collapse. The observed behavior suggested that cycloviolacin O4 possesses intrinsic conformational flexibility while retaining overall structural stability during long-timescale simulation. Cycloviolacin O4 maintained an overall stable conformational framework throughout the 500 ns simulation while retaining localized flexibility characteristic of cyclic peptides. Most residues exhibited limited fluctuations, indicating preservation of the compact cystine-knot architecture, whereas slightly increased mobility observed within several central peptide residues likely reflects intrinsic loop flexibility and solvent exposure rather than structural instability ( Fig 8B ). The peptide also preserved its overall structural compactness during the simulation despite moderate conformational adjustments. Although a transient increase in the radius of gyration was observed after approximately 300 ns, the peptide rapidly reached a new equilibrium without evidence of significant unfolding or loss of structural integrity, suggesting controlled conformational adaptation rather than destabilization ( Fig 8C ). Principal component analysis demonstrated that cycloviolacin O4 sampled several closely related conformational states during the simulation, indicating dynamic flexibility while remaining confined within a limited conformational space. This behavior suggests that the peptide underwent gradual structural adaptation without large-scale conformational transitions ( Fig 8D ). The internal dynamic behavior of cycloviolacin O4 was further supported by dynamic cross-correlation analysis, which revealed coordinated residue motions together with localized anti-correlated movements. These complementary motion patterns are consistent with the intrinsic flexibility of cyclic peptides and indicate that local structural adjustments occurred while preserving the overall conformational integrity of the peptide throughout the simulation ( Fig 8E ). Overall comparative analysis of the 500 ns molecular dynamics simulations suggested that the CYP19A1–cycloviolacin O4 complex maintained comparatively favorable structural stability and interaction persistence relative to the apo receptor and isolated ligand systems. The complex simulation demonstrated controlled RMSD fluctuations, sustained intermolecular hydrogen bonding interactions, stable active site behavior, and relatively preserved structural compactness throughout the simulation trajectory. In contrast, the apo CYP19A1 receptor exhibited comparatively greater conformational transitions and increased compactness fluctuations during later simulation stages, indicating higher intrinsic flexibility in the absence of ligand-mediated stabilization. Similarly, the isolated cycloviolacin O4 ligand demonstrated moderate conformational adaptability and dynamic structural rearrangements during the simulation period, reflecting the natural flexibility of the peptide framework under solvated conditions. Comparative RMSF analysis further indicated that the receptor-ligand complex maintained relatively stabilized residue fluctuations within important structural regions compared with the apo receptor system. The radius of gyration profiles additionally suggested that ligand binding contributed to preservation of overall receptor compactness during long-timescale simulation. Principal component and DCCM analyses collectively demonstrated that the complex adopted coordinated conformational motions with comparatively controlled dynamic transitions, whereas the apo receptor and isolated ligand displayed broader conformational sampling behavior. Furthermore, the persistence of intermolecular hydrogen bonding interactions throughout the simulation supported stable computational association between cycloviolacin O4 and CYP19A1 during the entire 500 ns trajectory. Collectively, these findings suggested that cycloviolacin O4 may contribute to stabilization of the CYP19A1 receptor environment following binding interaction. The favorable dynamic behavior, sustained intermolecular contacts, and comparatively stable conformational profiles observed during simulation further supported the potential computational compatibility of cycloviolacin O4 as a cyclotide-based CYP19A1 inhibitory candidate for modulation of estrogen biosynthesis associated with adenomyosis. To provide a quantitative computational assessment of the predicted CYP19A1–cycloviolacin O4 interaction beyond the docking scores, endpoint binding free-energy calculations were performed using both the MM-GBSA and MM-PBSA approaches. The individual energy components and total binding free energies obtained from both methods are summarized in Table 3 , whereas the corresponding energy decomposition profiles are presented in Fig 9 . Negative values indicate computationally favorable association. The reported values are method-dependent endpoint free-energy estimates and should not be interpreted as experimentally measured binding affinities. Negative energy components contribute favorably to complex formation, whereas positive values represent unfavorable solvation penalties. The MM-GBSA calculations predicted a favorable mean binding free energy (ΔG bind ) of −66.58 ± 3.46 kcal/mol, indicating energetically favorable receptor–peptide association. The interaction was predominantly driven by favorable gas-phase interactions (ΔG gas  = −128.57 ± 4.80 kcal/mol), resulting from substantial van der Waals (−62.50 ± 3.09 kcal/mol) and electrostatic (−66.06 ± 5.07 kcal/mol) contributions. These attractive interactions were partially opposed by the implicit solvation penalty (ΔG solv  = +61.99 ± 3.59 kcal/mol), primarily arising from the unfavorable polar solvation component (+70.32 ± 3.58 kcal/mol). However, the favorable nonpolar solvation contribution (−8.33 ± 0.18 kcal/mol) partially compensated for this penalty, resulting in a strongly negative overall free energy that supports favorable computational binding of cycloviolacin O4 to CYP19A1. The MM-PBSA analysis demonstrated a similar energetic trend for the gas-phase interactions, with identical van der Waals, electrostatic, and ΔG gas values to those obtained using MM-GBSA. However, the calculated solvation energy was considerably larger (ΔG solv  = +124.01 ± 4.48 kcal/mol) owing to the greater polar solvation penalty (EPB = +93.41 ± 4.21 kcal/mol) together with the cavity (−43.64 ± 0.43 kcal/mol) and dispersion (+74.23 ± 1.39 kcal/mol) contributions incorporated within the Poisson–Boltzmann model. Consequently, the predicted total binding free energy was substantially reduced (−4.56 ± 4.29 kcal/mol) compared with the MM-GBSA estimate. Comparison of the two endpoint free-energy approaches demonstrated that both methods consistently predicted energetically favorable association of cycloviolacin O4 with CYP19A1, as both yielded negative mean binding free energies. However, the magnitude of the predicted binding differed substantially between the two methods. As summarized in Fig 10 , the MM-GBSA approach predicted a markedly more favorable binding free energy (−66.58 ± 3.46 kcal/mol) than MM-PBSA (−4.56 ± 4.29 kcal/mol). The identical gas-phase interaction energies obtained by both methods indicate that receptor–peptide association is primarily driven by van der Waals and electrostatic interactions. The observed difference in the final binding free energy originates from the distinct implicit-solvation models employed by the Generalized Born and Poisson–Boltzmann approaches. Specifically, the Poisson–Boltzmann model estimated a substantially greater polar desolvation penalty, resulting in a considerably less favorable net binding free energy than that predicted by MM-GBSA. Both approaches predicted favorable receptor–peptide association, although the estimated binding strength differed owing to differences in the implicit solvation models. Collectively, the MM-GBSA and MM-PBSA analyses provide complementary quantitative evidence supporting the predicted interaction between cycloviolacin O4 and CYP19A1 beyond the docking scores alone. Although the absolute binding energies differed because of the distinct solvation models employed, both approaches consistently identified favorable receptor–peptide association and further supported the structural stability observed throughout the molecular dynamics simulations. As with all endpoint free-energy methods, these values represent computational estimates of binding energetics and should therefore be interpreted in conjunction with the molecular docking and molecular dynamics analyses rather than as experimentally determined binding affinities. The lead cyclotide, cycloviolacin O4, was further evaluated for its predicted antigenicity, allergenicity, and toxicity profiles to assess its potential biophysical compatibility and computational safety characteristics. The obtained results suggested an overall favorable biosafety profile of the selected peptide candidate for downstream therapeutic consideration. Antigenicity prediction analysis generated an overall score of 0.4384, classifying cycloviolacin O4 as a probable non-antigen. This finding suggested that the peptide may possess relatively low predicted immunogenic potential under the evaluated computational conditions. Reduced antigenic behavior is considered favorable for peptide-based therapeutic candidates because excessive immune recognition may contribute to undesirable immunological responses following administration. Allergenicity assessment further classified cycloviolacin O4 as a non-allergenic peptide candidate. The absence of predicted allergenic behavior suggested comparatively lower potential for hypersensitivity-associated immune responses, thereby supporting the computational biocompatibility of the lead cyclotide. In addition, toxicity profiling identified cycloviolacin O4 as a non-toxic peptide according to the applied predictive models, indicating the absence of major predicted toxic characteristics within the evaluated sequence framework. Collectively, the combined biophysical evaluation analyses suggested that cycloviolacin O4 possesses comparatively favorable predicted safety characteristics, including low antigenic potential, absence of predicted allergenicity, and non-toxic behavior. These findings further supported the suitability of the lead cyclotide for subsequent immune simulation and advanced computational evaluation in the present in silico investigation. The immune simulation analysis of cycloviolacin O4 was performed to evaluate its predicted immunological behavior and potential host immune response under simulated physiological conditions. Overall, the obtained immune simulation results suggested the absence of excessive or abnormal immune activation against the lead cyclotide. The observed immunological patterns indicated comparatively controlled immune system interaction behavior, supporting the predicted biocompatibility and computational safety profile of cycloviolacin O4 as a potential therapeutic peptide candidate. The immune simulation predicted a controlled immunological response following exposure to cycloviolacin O4, characterized by progressive antigen clearance without evidence of prolonged antigen persistence or excessive immune activation. The gradual reduction in antigen levels together with a balanced immunoglobulin response suggests that the lead cyclotide is unlikely to trigger exaggerated humoral immune stimulation under the simulated conditions, supporting its computationally predicted low antigenic potential ( Fig 11A ). Similarly, the simulated cytokine profile demonstrated a regulated immune response characterized by moderate transient cytokine and interleukin production followed by stabilization during the later stages of the simulation. The absence of sustained pro-inflammatory signaling or cytokine storm-like behavior indicates a balanced pattern of immune communication rather than excessive inflammatory activation. Collectively, these findings suggest that cycloviolacin O4 exhibits favorable computationally predicted immunological compatibility under the simulated conditions; however, these observations should be interpreted as in silico predictions that require experimental validation ( Fig 11B ). The adaptive immune simulation predicted a balanced B-cell response following exposure to cycloviolacin O4. Total B-cell populations remained relatively stable throughout the simulation, with only moderate expansion during antigen exposure and controlled development of active and memory B-cell populations. The absence of excessive B-cell proliferation or abnormal humoral activation further supports the computationally predicted low immunogenic potential of the lead cyclotide ( Fig 11C ). A similarly regulated pattern was observed for helper T-cell populations, where both helper and memory T cells gradually increased without evidence of abrupt immune expansion or excessive activation. This controlled cellular response suggests that cycloviolacin O4 is unlikely to provoke exaggerated T-cell-mediated immune stimulation under the simulated conditions, indicating balanced adaptive immune recognition ( Fig 11D ). Cytotoxic T-cell dynamics further supported this observation by demonstrating relatively stable cell populations with only moderate fluctuations throughout the simulation. The absence of sustained cytotoxic expansion or excessive immune activation suggests limited cell-mediated immune reactivity toward cycloviolacin O4, consistent with its computationally predicted favorable immunological safety profile ( Fig 11E ). The dendritic cell response was likewise characterized by predominance of resting cells, whereas activated dendritic cell populations remained comparatively limited during the simulation. Because dendritic cells play a central role in antigen presentation and initiation of adaptive immune responses, this balanced activation pattern suggests a relatively low computationally predicted immune stimulatory potential for cycloviolacin O4 under the simulated conditions. Nevertheless, these immune simulation results represent in silico predictions and should be interpreted cautiously until validated experimentally ( Fig 11F ). Normal mode analysis of the CYP19A1–cycloviolacin O4 docked complex was performed to further investigate the intrinsic flexibility, collective molecular motions, structural deformability, and dynamic stability of the receptor-peptide complex. The obtained NMA results suggested that the lead docked complex maintained comparatively stable dynamic behavior with coordinated residue motions and controlled structural flexibility under simulated conditions. Normal mode analysis indicated that the CYP19A1–cycloviolacin O4 complex maintained an overall rigid structural framework while preserving localized flexibility required for biologically relevant conformational adaptation. Most residues exhibited low to moderate deformability, whereas only a few surface-exposed and loop-associated regions displayed comparatively higher mobility, suggesting that structural flexibility was confined to naturally dynamic regions without compromising the overall integrity of the complex ( Fig 12A ). The predicted residue fluctuation profile was further supported by the calculated B-factor distribution, which showed good agreement with the intrinsic structural flexibility expected for the receptor. Although several regions exhibited moderate fluctuations, the majority of residues maintained controlled motion amplitudes, indicating that cycloviolacin O4 binding did not induce excessive structural instability ( Fig 12B ). The calculated eigenvalue (3.607006 × 10 −4 ) suggested that the receptor–peptide complex can undergo collective conformational transitions with relatively low energetic cost while preserving overall structural stability. This balance between conformational adaptability and structural rigidity is consistent with stable receptor–ligand interactions and supports the dynamic compatibility of cycloviolacin O4 with the CYP19A1 binding site ( Fig 12C ). The cumulative variance profile further demonstrated that the collective motions of the complex were distributed smoothly across multiple low-frequency normal modes, indicating coordinated internal dynamics rather than abrupt conformational rearrangements. Together, the normal mode analyses support the computationally predicted structural stability and conformational adaptability of the CYP19A1–cycloviolacin O4 complex under the simulated conditions ( Fig 12D ). The covariance analysis revealed coordinated residue motions throughout the CYP19A1–cycloviolacin O4 complex, indicating efficient dynamic communication between different structural regions of the receptor–peptide system. The predominance of positively correlated motions suggests synchronized conformational behavior that may contribute to maintaining the integrity of the binding interface and supporting persistent receptor–peptide interactions during the simulated dynamics ( Fig 12E ). The elastic network model further demonstrated extensive mechanical connectivity across the receptor–peptide complex, reflecting strong residue coupling and an organized internal communication network. Such dense elastic interactions are consistent with preservation of the overall structural architecture while permitting coordinated conformational adjustments required for stable molecular recognition. Collectively, these findings support the computationally predicted structural robustness and dynamic stability of the CYP19A1–cycloviolacin O4 complex under the simulated conditions ( Fig 12F ). Collectively, the normal mode analysis results suggested that the CYP19A1–cycloviolacin O4 complex maintained favorable dynamic stability, coordinated residue motion behavior, and acceptable conformational flexibility. The controlled deformability, low-energy collective motion profile, coordinated covariance behavior, and extensive elastic network connectivity collectively supported the predicted computational stability of the lead docked complex. These findings further strengthened the potential suitability of cycloviolacin O4 as a computationally capable cyclotide-based inhibitor targeting CYP19A1 for modulation of estrogen biosynthesis associated with adenomyosis.

Conclusions

This study employed an integrated structure-based computational workflow to identify cyclotides with the potential to interact with human placental aromatase cytochrome P450 (CYP19A1), a biologically relevant target associated with estrogen-dependent adenomyosis. Through molecular docking, molecular dynamics simulations, normal mode analysis, endpoint MM-GBSA/MM-PBSA free-energy calculations, and in silico safety assessment, cycloviolacin O4 emerged as the most favorable lead candidate based on its overall predicted interaction profile, structural stability, and dynamic compatibility with the CYP19A1 binding site. The incorporation of pharmacological and peptide-based reference controls further strengthened the computational benchmarking of the docking workflow and provided additional context for interpreting the predicted receptor–peptide interactions. Collectively, the computational findings suggest that cycloviolacin O4 represents a structurally promising cyclotide scaffold for future CYP19A1-targeted inhibitor development. However, these findings should be interpreted as computational predictions rather than evidence of biological efficacy or therapeutic activity. Experimental validation, including biochemical aromatase inhibition assays, binding studies, cell-based functional investigations, and appropriate in vivo models, will be essential to determine whether the predicted molecular interactions translate into measurable biological activity and therapeutic potential. Therefore, the present study should be regarded as a hypothesis-generating computational investigation that provides a rational foundation for future experimental development of cycloviolacin O4 as a putative cyclotide-based aromatase inhibitor for adenomyosis.

Materials|Methods

The three-dimensional structure of human placental aromatase cytochrome P450 (CYP19A1) was retrieved from the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB) database ( https://www.rcsb.org ). The experimentally resolved structure was selected based on its crystallographic quality and suitability for downstream protein-peptide interaction analyses [ 24 ]. Prior to computational investigations, the retrieved CYP19A1 structure (PDB ID: 3S79) was subjected to protein preparation and purification using BIOVIA Discovery Studio Visualizer. The receptor structure was carefully inspected, and all non-essential heteroatoms, including crystallographic water molecules, co-crystallized ligands, solvent molecules, buffer ions, and other non-protein residues, were removed to eliminate potential interference during active-site prediction, molecular docking, and molecular dynamics simulations. Structural editing and visualization were performed to generate a clean receptor model suitable for computational analysis. Following initial purification, structural refinement was performed using PDBFixer. The prepared protein structure was examined for missing heavy atoms, incomplete amino acid residues, absent side chains, and other structural inconsistencies commonly encountered in experimentally determined crystal structures. Missing atoms and residues were repaired where required, and hydrogen atoms were added to generate a chemically consistent protein model suitable for downstream simulations. The refined structure was subsequently saved in PDB format for further analyses [ 25 ]. To evaluate the stereochemical quality and structural reliability of the prepared receptor, validation analyses were performed using the Structural Analysis and Verification Server (SAVES v6.0) ( https://saves.mbi.ucla.edu ). Ramachandran plot analysis generated through PROCHECK was used to assess backbone dihedral angle distributions and overall residue conformational quality. In addition, the ERRAT algorithm was employed to evaluate the quality of non-bonded atomic interactions and identify potential structural irregularities within the receptor structure. The validated CYP19A1 model was subsequently utilized for active-site prediction, molecular docking, interaction analysis, and molecular dynamics simulations [ 26 ]. Protein-protein interaction (PPI) network analysis of human placental aromatase cytochrome P450 (CYP19A1) was performed using the STRING database version 12.0 ( https://string-db.org ). PPI network analysis is important for identifying the functional interaction landscape of a target protein and understanding its involvement in biological pathways, molecular regulation, and disease-associated cellular processes. In the context of adenomyosis and estrogen biosynthesis, investigation of CYP19A1-associated interacting proteins may provide insight into the molecular environment associated with estrogen regulation and endocrine signaling pathways [ 27 ]. The CYP19A1 protein was queried in the STRING database using Homo sapiens as the selected organism to ensure species-specific interaction profiling. The interaction network was generated based on experimentally validated evidence, curated database information, co-expression patterns, neighborhood associations, gene fusion events, and co-occurrence data integrated within the STRING platform. Default confidence parameters provided by the server were applied to obtain biologically relevant interaction partners associated with CYP19A1. The interaction map obtained from STRING was subsequently utilized to support the biological relevance of CYP19A1 as a therapeutic target in the present computational investigation [ 28 ]. The cyclotide sequences utilized in this study were retrieved from the CyBase database ( https://www.cybase.org.au ), a specialized repository containing experimentally reported cyclotide sequences and structural information. The selection of cyclotides was performed to explore naturally occurring cyclic peptides as potential inhibitors of CYP19A1 due to their structural stability, compact disulfide-rich framework, and previously reported biological activities. Retrieval of accurate peptide sequences is an important step in peptide-based computational investigations because the quality of downstream structural modeling and docking predictions depends on the correctness of the primary amino acid sequence [ 29 ]. The FASTA sequences of the selected cyclotides were obtained directly from the CyBase repository and subsequently used for three-dimensional structure prediction. Since experimentally resolved structures were not available for all selected cyclotides, computational structure modeling was performed using AlphaFold for reliable peptide conformation prediction. Structural modeling of peptides is essential in molecular docking studies because the spatial arrangement, folding pattern, and conformational stability of ligands significantly influence receptor-binding interactions and predicted binding affinity. The three-dimensional structures of all cyclotides were generated using the AlphaFold protein structure prediction framework. The obtained peptide structures were carefully examined for structural integrity, cyclic peptide folding consistency, and disulfide bond organization prior to downstream analysis. Predicted confidence metrics generated by the modeling framework were also considered to assess the reliability of the generated peptide conformations. The finalized three-dimensional peptide models were subsequently utilized for active site docking, interaction profiling, and molecular dynamics simulations against the CYP19A1 receptor [ 29 ]. The potential ligand-binding pockets of human placental aromatase cytochrome P450 (CYP19A1) were predicted using the PrankWeb server ( https://prankweb.cz ). Active site prediction is an important step in structure-based computational screening because it helps identify probable functional cavities and binding regions that may participate in receptor-ligand recognition. This information supports more focused docking analysis and improves the biological relevance of predicted binding orientations [ 30 ]. The prepared and validated CYP19A1 structure was submitted to the PrankWeb server in PDB format. The server predicted possible binding pockets using a machine learning-based approach that evaluates the local structural environment of the protein surface. The predicted pockets were ranked by the server according to their probability scores and residue composition. The most relevant binding pocket was selected for downstream docking analysis based on its predicted confidence, pocket ranking, and spatial suitability for peptide interaction with CYP19A1. The amino acid residues forming the selected pocket were recorded and used to guide the interpretation of molecular docking and interaction analysis [ 31 ]. To assess the biological relevance of the computationally predicted pocket, the residues assigned to the highest-ranked PrankWeb cavity were compared with the experimentally characterized androstenedione-binding region of the CYP19A1 crystal structure, PDB ID 3S79. Particular attention was given to Asp309, Thr310, and Arg435, which participate in substrate recognition and catalytic organization within the aromatase active site. Residue numbering and amino-acid identities were verified against chain A of the 3S79 structure. The presence of these experimentally established active-site residues within the predicted pocket was used as structural evidence that the selected cavity overlapped the native substrate-binding region. The validated three-dimensional structure of human placental aromatase cytochrome P450 (CYP19A1) and the AlphaFold-predicted cyclotide structures were prepared for protein–peptide docking using the High Ambiguity Driven Biomolecular Docking (HADDOCK 2.4) server ( https://wenmr.science.uu.nl/haddock2.4 ). Docking was performed in protein–peptide interaction mode to predict the binding orientation and intermolecular interactions of cyclotides within the experimentally supported active-site region of CYP19A1. Prior to docking, all cyclotide structures were generated as complete three-dimensional PDB models using AlphaFold and were submitted directly to HADDOCK without further structural modification [ 32 ]. Because cyclotides possess a highly constrained cyclic cystine-knot framework, the submitted AlphaFold models inherently preserved the native head-to-tail cyclized backbone and disulfide-bond connectivity. Consequently, no additional user-defined covalent, distance, or flexibility restraints were applied during docking. HADDOCK performed rigid-body docking followed by its standard semi-flexible refinement and explicit solvent refinement protocol to generate the final receptor–peptide complexes. The resulting docking solutions were ranked according to the HADDOCK scoring function, interface quality, and structural clustering characteristics [ 33 ]. To improve the robustness of the docking workflow, all predicted complexes were independently validated using the ClusPro protein–protein docking server under identical receptor–ligand conditions. Candidate prioritization was based on an integrated assessment rather than docking score alone, considering HADDOCK and ClusPro performance, binding orientation within the predicted catalytic pocket, interactions with catalytically relevant residues, interface quality, and overall structural plausibility. Based on these complementary criteria, cycloviolacin O4 was selected as the lead candidate for subsequent molecular dynamics simulations and detailed computational analyses [ 34 ]. To further benchmark the docking protocol, exemestane (PubChem CID: 60198), a clinically approved CYP19A1 inhibitor, and the reported reference peptide SNA-60-367-10 were additionally docked using the same computational workflow [ 35 ]. The structural information for SNA-60-367-10 was retrieved from the Natural Products Magnetic Resonance Database (NP-MRD) and is based on the experimentally characterized peptide originally reported by Esumi et al. ( https://np-mrd.org/ ). Their docking results were compared with those of the prioritized cyclotide to provide pharmacological and peptide-based reference controls for interpreting the predicted interaction profile of cycloviolacin O4. These reference ligands were included for comparative docking analysis and were not part of the initial cyclotide screening library. To evaluate the structural stability and dynamic behavior of the CYP19A1–cycloviolacin O4 complex under physiological conditions, molecular dynamics (MD) simulations were performed using the OpenMM simulation engine. Three independent simulation systems were constructed using an identical protocol, including the docked CYP19A1–cycloviolacin O4 complex, the apo CYP19A1 receptor, and the isolated cycloviolacin O4 peptide. Each system was subjected to a 500 ns production simulation to comparatively evaluate receptor stability, peptide conformational behavior, and receptor–peptide interaction persistence [ 36 ]. The lead cyclotide selected for molecular dynamics simulations was not determined solely on the basis of docking score. Candidate prioritization was performed using a multi-parameter evaluation strategy that considered HADDOCK docking score, ClusPro validation, binding orientation within the predicted active site, interactions with catalytically relevant residues, interface quality, and overall structural plausibility of the docked complex. The AMBER ff14SB force field was employed for protein and peptide parameterization. All systems were solvated in an explicit TIP3P water box with a minimum 10 Å buffer surrounding the solute. Counter ions (Na ⁺ /Cl⁻) were added to neutralize the systems, and the ionic strength was adjusted to 0.15 M NaCl to mimic physiological conditions. Long-range electrostatic interactions were treated using the Particle Mesh Ewald (PME) method with a 10 Å non-bonded cutoff. Following energy minimization, each system was equilibrated sequentially under NVT and NPT ensembles at 300 K and 1 atm using a Langevin thermostat and a Monte Carlo barostat, respectively. Covalent bonds involving hydrogen atoms were constrained using the SHAKE algorithm, and a 2 fs integration time step was employed throughout the production simulations. Post-simulation trajectory analyses were performed to characterize the structural and dynamic properties of the simulated systems. Root mean square deviation (RMSD), root mean square fluctuation (RMSF), radius of gyration (Rg), solvent-accessible surface area (SASA), intermolecular hydrogen bond analysis, free-energy landscape (FEL), principal component analysis (PCA), and dynamic cross-correlation matrix (DCCM) analyses were performed to evaluate structural stability, residue flexibility, molecular compactness, conformational dynamics, and interaction persistence throughout the 500 ns simulation. Only intermolecular hydrogen bonds between CYP19A1 and cycloviolacin O4 were quantified, using a donor–acceptor distance cutoff of 3.5 Å and a donor–hydrogen–acceptor angle cutoff of 135°, with representative interactions independently verified using LigPlot+ and PyMOL [ 28 ]. Endpoint binding free-energy calculations were conducted for the CYP19A1–cycloviolacin O4 complex using the MM-GBSA and MM-PBSA approaches. Representative conformations were extracted from the equilibrated molecular dynamics trajectory, and explicit water molecules and counterions were removed before calculation. Binding free energy was estimated according to: The molecular-mechanics contribution comprised van der Waals and electrostatic interaction energies. Polar solvation was estimated using the Generalized Born model for MM-GBSA and the Poisson–Boltzmann model for MM-PBSA. Nonpolar solvation was calculated from solvent-accessible surface-related terms in MM-GBSA and from the corresponding cavity and dispersion contributions in MM-PBSA. Negative mean values of ΔG_bind were interpreted as computationally favorable association. Because endpoint free-energy estimates depend on the selected implicit-solvent model and trajectory-sampling procedure, the calculated values were interpreted as comparative computational indicators rather than experimentally measured binding affinities [ 37 ]. The selected lead cyclotide was further subjected to biophysical and immunological safety evaluation to assess its predicted antigenicity, allergenicity, and toxicity profiles using computational prediction tools. Biophysical evaluation is an important step in peptide-based therapeutic development because biologically unfavorable immune or toxic responses may limit the translational applicability of peptide candidates despite favorable receptor-binding characteristics. Early-stage computational screening of these properties assists in identifying peptide candidates with potentially improved biosafety and therapeutic suitability. The antigenicity profile of the lead cyclotide was predicted using the VaxiJen v2.0 server ( http://www.ddg-pharmfac.net/vaxijen/VaxiJen/VaxiJen.html ). VaxiJen is an alignment-independent prediction platform that evaluates the antigenic potential of peptides based on physicochemical property analysis. Assessment of antigenicity is important because highly antigenic peptide candidates may trigger undesirable immune recognition and immunogenic responses following therapeutic administration [ 38 ]. The allergenicity potential of the lead cyclotide was evaluated using the AllerTOP v2.0 server ( https://www.ddg-pharmfac.net/AllerTOP ). Prediction of allergenic behavior is essential in peptide therapeutic studies because allergenic compounds may induce hypersensitivity reactions and immune-associated complications. AllerTOP performs allergenicity prediction based on auto-cross covariance transformation of amino acid properties and machine learning classification approaches [ 39 ]. In addition, toxicity profiling of the lead cyclotide was carried out using the ToxinPred server ( https://webs.iiitd.edu.in/raghava/toxinpred2 ). Toxicity prediction is important to identify peptide candidates that may exhibit harmful biological effects or cellular toxicity under physiological conditions. The ToxinPred platform evaluates peptide toxicity using sequence-based computational models and physicochemical descriptors associated with toxic peptide behavior [ 40 ]. The immunological response profile of the selected lead cyclotide was evaluated through immune simulation analysis using the C-ImmSim immune simulation server ( https://kraken.iac.rm.cnr.it/C-IMMSIM ). Immune simulation studies are important in peptide-based therapeutic investigations because they provide computational insight into the potential interaction of peptide candidates with components of the host immune system. Such analyses may help estimate the immunological behavior, immune activation tendency, and possible adaptive immune responses associated with peptide administration under simulated biological conditions [ 41 ]. The amino acid sequence of the lead cyclotide was submitted to the C-ImmSim server for simulation of host immune responses under default physiological parameters. The simulation platform utilizes machine learning algorithms and position-specific scoring matrix approaches to mimic mammalian immune system behavior, including antigen processing, immune cell activation, cytokine release, and antibody production. The simulation was configured to evaluate the predicted immunological interactions of the lead cyclotide over sequential exposure intervals within a virtual immune environment. Various immune-associated parameters generated during the simulation process were monitored to assess the computational immune response behavior of the peptide candidate. These included predicted immunoglobulin responses, B-cell and T-cell population dynamics, cytokine-associated responses, and antigen clearance behavior throughout the simulation timeline. The immune simulation analysis was subsequently used to provide additional computational insight into the potential immunological compatibility and predicted host response characteristics of the lead cyclotide candidate [ 42 ]. Normal Mode Analysis (NMA) of the lead cyclotide-CYP19A1 docked complex was performed using the iMODS server ( https://imods.iqfr.csic.es ). NMA is an important computational approach used to evaluate the intrinsic flexibility, collective molecular motions, and structural deformability of biomolecular complexes. Unlike conventional molecular docking, which provides a static representation of receptor-ligand interactions, NMA enables investigation of the possible dynamic behavior and motion-associated stability of the docked complex under simplified physiological conditions [ 43 ]. The docked complex of the lead cyclotide with human placental aromatase cytochrome P450 (CYP19A1) was submitted to the iMODS server in PDB format for dynamic analysis. The server applies internal coordinate normal mode calculations to estimate the large-scale collective motions and conformational flexibility of protein complexes. Structural dynamics assessment through NMA is particularly useful for evaluating the potential stability and flexibility of protein-peptide interactions while requiring lower computational cost compared with long-timescale atomistic simulations. Several dynamic parameters generated by the iMODS platform were analyzed to investigate the structural behavior of the docked complex. These included deformability profiles, B-factor estimations, eigenvalue calculations, covariance matrices, elastic network models, and collective motion patterns associated with the receptor-ligand complex. The eigenvalue parameter was utilized to estimate the predicted motion stiffness and energy requirement for conformational transitions within the complex, whereas covariance and elastic network analyses were examined to evaluate correlated residue motions and structural connectivity. The obtained NMA results were subsequently used to support the dynamic stability assessment of the lead cyclotide interaction with CYP19A1 [ 31 ]. Language editing of the manuscript text was performed using the generative artificial intelligence tool ChatGPT, based on the GPT-5.6 model ( https://chatgpt.com ). Draft sections prepared by the authors were submitted to the tool with instructions limited to grammatical correction, sentence restructuring, and refinement of clarity and readability, and each returned passage was subsequently compared with the original wording and verified by the authors before incorporation into the manuscript. The tool was not applied to cyclotide sequence retrieval, three-dimensional structure modeling, active-site prediction, molecular docking, molecular dynamics simulation, binding free-energy calculation, or in silico safety assessment, nor to the preparation of figures and tables, each of which was carried out using the computational platforms described above.

Supplementary Material

(DOCX)

Text is read by the "Ask this paper" AI Q&A widget below. Extraction quality varies by source — PMC NXML preserves structure cleanly, OA-HTML may include some navigation residue, and OA-PDF can have broken hyphenation. The publisher copy (via DOI) is the canonical version.

My notes (saved in your browser only)

⚙ Ask this paper AI returns verbatim quotes from the full text · source: pmc-nxml ⓘ

Answers must be backed by verbatim quotes from this paper's full text. Hallucinated quotes are dropped automatically; if no verbatim passage answers the question, we say so. How this works

MeSH descriptors

Adenomyosis Adenomyosis Adenomyosis Adenomyosis Adenomyosis Aromatase Aromatase Aromatase Aromatase Aromatase Aromatase Cyclotides Cyclotides Cyclotides Cyclotides Cyclotides Cyclotides Estrogens Estrogens Estrogens

Citation neighborhood (2-hop)

Papers in the corpus that this work cites (lower rings, blue) and that cite this one (upper rings, green). Dot size scales with the paper's in-corpus citation count — bigger dot = more influential within the endo/adeno field. Click a dot to open that paper. Outer rings show 2-hop neighbours — papers reached through the immediate citers/citees. [ collapse to 1-hop ]

References (53)

Source provenance

europepmc
last seen: 2026-11-21T06:17:04.340742+00:00
openalex
last seen: 2026-11-21T06:07:56.064915+00:00
pubmed
last seen: 2026-11-21T06:09:43.705417+00:00
License: public-domain-us · commercial use OK · attribution required
Courtesy of the U.S. National Library of Medicine