In silico Investigation of Oxypeucedanin Hydrate from Traditional Chinese Medicine as a Potential PSAT1 Inhibitor in Early-Stage Ovarian Cancer

Abstract

Ovarian cancer is a major contributor to disease burden and fatalities globally, with earlystage cases posing substantial treatment hurdles due to restricted therapeutic modalities. Traditional Chinese Medicine (TCM) has gained attention for its diverse repertoire of natural compounds exhibiting possible antitumor effects. This research explores the therapeutic potential of oxypeucedanin hydrate (docking score of -9.75 kcal mol^-1). Computational approaches were employed to assess the interaction strength between oxypeucedanin hydrate and the critical protein phosphoserine aminotransferase 1 (PSAT1) involved in ovarian cancer progression. Molecular dynamics (MD) simulations revealed the stability and conformational behavior of the compoundtarget complexes, mainly focusing on Trp107. Pharmacokinetic evaluation showed high intestinal permeability (QPPCaco = 468.389) and good oral absorption (83.94%). WaterMap analysis further supported its binding stability with a binding free energy (dG) of -15.41 kcal mol^-1 and a favorable entropy contribution (-30.08 kcal mol^-1). A 100 ns MD simulation confirmed the conformational stability of the OPH-PSAT1 complex, with a mean protein backbone root mean square deviation (RMSD) of 1.70 Å and ligand RMSD of 3.09 Å. MM/GBSA (Molecular Mechanics/Generalized Born Surface Area) analysis yielded an average binding free energy (dG_bind) of -63.88 kcal mol^-1, outperforming reference ligand pyridoxamine phosphate (PMP) (-48.08 kcal mol^-1). These findings position oxypeucedanin hydrate as a novel therapeutic candidate capable of suppressing tumor growth and triggering programmed cell death in ovarian malignancies.

Keywords:
traditional Chinese medicine; PSAT1; molecular docking; MD simulation; virtual screening

Introduction

Ovarian cancer remains a pressing global health issue, primarily due to high mortality rates associated with latestage diagnoses and limited effective treatment options. The GLOBOCAN 2022 report estimated that 324,603 women were diagnosed with ovarian cancer in 2022, leading to 206,956 deaths. Ovarian cancer ranks as the 8^th most common cancer among women in terms of both incidence and mortality.¹ Notably, the prevalence is higher in industrialized regions of Europe and North America, while African countries generally report lower rates. In India, it is a leading cause of cancer-related deaths due to challenges in early detection. Early-stage ovarian cancer diagnosed at Stage I has a five-year survival rate of 94%, though only about 15% of cases are detected at this stage. Meanwhile, approximately 62% of cases are diagnosed at Stages III and IV, where the five-year survival rate drops to just 28%.²,³ Despite advances, the limited effectiveness of current treatments and the emergence of resistance underscore the urgent need for new therapeutic approaches.

Traditional Chinese Medicine (TCM) presents significant potential for identifying natural inhibitors of phosphoserine aminotransferase 1 (PSAT1) in earlystage ovarian cancer,⁴ leveraging its holistic principles and diverse natural compounds.⁵ PSAT1, recognized as a therapeutic biomarker in early ovarian malignancies.⁶,⁷ This enzyme is critical in the serine biosynthesis pathway,⁶ which supports cancer cell growth and survival, particularly in early stages. It facilitates the conversion of 3-phosphohydroxypyruvate (3-PHP) to L-phosphoserine within the serine-glycine biosynthesis pathway, utilizing glutamate as an amino donor. This pathway is essential for cancer cell proliferation and survival during early tumorigenesis. The Trp107 residue in PSAT1’s active site is crucial in positioning the cofactor and enabling the enzymatic reaction with 3-PHP.⁸ Inhibition of PSAT1 disrupts serine biosynthesis, leading to deoxyribonucleic acid (DNA) damage and triggering cell death through apoptosis, which underscores its therapeutic relevance. PSAT1 also contributes to treatment resistance in cancers such as melanoma, pancreatic cancer, and non-small cell lung carcinoma.⁹ Elevated expression of PSAT1 is associated with increased metastatic potential and poor prognosis in triple-negative breast cancer,¹⁰ further validating its role as a multifaceted oncogenic target. These findings highlight the synergy between TCM-derived compounds and modern biomarker-driven therapies in addressing PSAT1-driven cancer pathways.

This research investigates a new strategy for treating early-stage ovarian cancer by using computational techniques to identify potential TCM compounds that could inhibit the PSAT1 enzyme.⁸ Through computational methods such as virtual screening, molecular docking, and molecular dynamics (MD) simulations, a collection of TCM compounds was analyzed for their ability to bind to the active site of PSAT1. Integrating computational tools with experimental validation, the study aims to accelerate the drug discovery process. The findings highlight promising TCM-derived inhibitors of PSAT1, laying the groundwork for future preclinical and clinical research aimed at enhancing treatment options for early-stage ovarian cancer. A total of 381 TCM compounds were screened, and their docking results were compared to those of the native ligand to evaluate their binding strengths. Additionally, interaction fingerprints were analyzed, and pharmacokinetics were evaluated using the QikProp tool.¹¹ A 100 ns MD simulation was conducted to examine deviation fluctuations and intermolecular interactions, followed by MM/GBSA (Molecular Mechanics/Generalized Born Surface Area) analysis to validate the results.

Methodology

A detailed representation of the methodological workflow employed in this study is shown in Figure 1. It encapsulates the entire process, starting from data acquisition and extending through various analytical stages such as protein and ligand preparation, molecular docking, pharmacokinetic assessment, molecular interaction fingerprinting, MD simulations, and the MM/GBSA analysis.¹²,¹³ Each of these stages is elaborated in the following sections to ensure a clear understanding of the techniques and rationale guiding our research methodology.

Figure 1
Methodological pipeline adopted in the current study.

Data collections and protein preparations

The three-dimensional structure of human PSAT1 in complex with pyridoxamine phosphate (PMP) was sourced from the RCSB Protein Data Bank (accession number 8A5V). This crystallographic model consists of eight individual monomers, which are paired to form four dimers within the asymmetric unit. Each monomer exhibits a classic α/β structural motif, typical of foldtype I PLP-dependent enzymes, and is organized into two distinct domains. The major domain, encompassing amino acid residues 19 to 265, consists of seven β-strands surrounded by six α-helices, and is primarily responsible for dimerization. The smaller domain includes residues 266 to 307 and the N-terminal residues 1-18. This region features three antiparallel β-strands and three α-helices. Together, individual monomers associate to form four stable, S-shaped dimers.⁸

In order to perform computational studies, the crystal structure was loaded into the Maestro interface of the Schrödinger Suite (version 14.0.136).¹⁴ Only chains A and B were selected for further analysis, and all heteroatoms were excluded. The Protein Preparation Wizard in Schrödinger Suite 2024-2 was employed to process the structure. This preparation involved reconstructing any absent side chains, assigning proper bond orders with the Chemical Component Database (CCD), and adding hydrogen atoms. Disulfide bridges were defined, and missing loops were modeled using Prime.¹⁵ The protonation states of the protein were adjusted to reflect a physiological pH of 7.4 ± 2.0, and hydrogen bonding patterns were optimized using PROPKA.¹⁶,¹⁷ To refine the structure, energy minimization was performed using the OPLS4 force field.¹⁸ Additionally, water molecules situated more than 5 Å away from any heteroatoms were eliminated to complete the model preparation.

Natural compound library collection and preparations

A review of the scientific literature was done to find possible PSAT1 inhibitors from TCM. A variety of TCMs have been explored for their potential in cancer therapy. These include herbal agents such as Allium sativum,¹⁹ Astragalus membranaceous (Huang Qi),²⁰ Panax ginseng (Ren Shen),²¹ Curcuma longa,²² Aloe vera, Angelica dahurica, Betula dahurica, Chelidonium majus, Coriandrum sativum, Ephedra sinica, Magnolia denudate, Magnolia fargesii, Magnolia kobus, Prunus dulcis, as well as animal-derived substances like toad venom (Chan Su) and Mylabris phalerata (Ban mao),²³ were considered. Molecular data for the plant-derived compounds were obtained from Dr. Duke’s Phytochemical and Ethnobotanical Databases,²⁴ yielding a total of 381 molecules. These compounds were downloaded in 2D SDF format from PubChem. For virtual screening, the molecules were processed using the LigPrep²⁵ module (version 2024-2) within the Schrödinger Suite (Maestro 14.0.136). The OPLS4 force field was applied during preparation, and potential ionization states at physiological pH (7.4 ± 2.0) were predicted using Epik.¹⁶ Additional steps included desalting, tautomer generation, and maintaining original stereochemistry while generating up to 32 stereoisomers per compound. This preparation process yielded a total of 767 ligand structures in SDF format.

Molecular Docking

A receptor grid was established for molecular docking using the Receptor Grid Generation wizard, centering the grid at coordinates (X: -3.49, Y: -56.17, Z: 4.74) with dimensions of 10 × 10 × 10 Å. These parameters were chosen based on the position of the co-crystallized ligand within the protein structure.

For the docking process, the Virtual Screening Workflow (VSW) was employed. Before docking, compounds were filtered for drug-likeness using QikProp¹¹ and Lipinski’s Rule, ensuring only suitable candidates were considered. The VSW’s extra precision (XP) docking protocol was used, generating up to four binding poses for each ligand. All docked compounds were retained for further analysis and subjected to MM/GBSA postprocessing to refine binding affinity predictions. The top 3 results were utilized for induced-fit docking.

Induced-fit docking

Traditional docking approaches typically assume the protein receptor remains rigid, only allowing flexibility in the ligand. While this method is computationally efficient, it may overlook important conformational changes in the protein that occur upon ligand binding. Consequently, receptor-binding pockets are now widely acknowledged as flexible structures capable of adopting multiple conformations in response to different ligand architectures. Recognizing that protein binding sites are dynamic and can adapt to different ligands, Induced Fit Docking (IFD)²⁶ was performed for the top three docking hits. The IFD protocol in Maestro (version 14.0.136) was utilized with extended sampling, allowing up to 80 potential binding poses per ligand using automated settings. The docking site was defined by the location of the co-crystallized ligand, and no additional constraints were applied. Ligand conformational sampling was set with a 2.5 kcal mol^-1 energy window to explore ring flexibility. Additionally, residues within 5 Å of the ligand were refined using Prime to better accommodate ligand-induced structural changes.

WaterMap analysis

WaterMap²⁷ is an advanced computational technique used in drug discovery to analyze the behavior and thermodynamics of water molecules within protein binding pockets. Since water plays a critical role in modulating how drugs interact with their targets, understanding these interactions can greatly enhance the design and optimization of therapeutics. In our investigation, we leveraged the WaterMap Perform Calculation tool, which utilizes Desmond²⁸ as its simulation engine, to study the hydration landscape surrounding the ligand within the active site of the protein. For this analysis, we selected the ligand from its docked configuration and ensured it remained fixed during the calculations to maintain accuracy. The study focused on a 10 Å region around the binding site, capturing the most relevant water-ligand interactions. A 5-nanosecond MD simulation was performed using the OPLS4 force field, with water modeled as the explicit solvent. The simulation was run without saving trajectory files to streamline the workflow and focus on hydration analysis.

Post-simulation, the WaterMap tool was used to interpret the hydration data, employing a combined scoring system that integrates MMGBSA and WaterMap results to identify the most favorable ligand binding poses. The Variable Solvent Generalized Born (VSGB) solvation model was applied for energy calculations, with nearby protein residues considered rigid to simplify the analysis. Throughout the process, key structural elementsincluding the receptor, ligand, hydrogen bonds, and water markerswere retained for visualization and assessment of hydration energetics.

This comprehensive approach provided a detailed understanding of how water molecules within the binding site affect ligand affinity and stability. Such information is invaluable for guiding the rational development of new drug candidates or refining existing molecules to improve their efficacy.

Molecular Dynamics (MD) simulation

MD simulation is a robust computational method employed to observe the time-dependent behavior and interactions within biomolecular systems. This approach is invaluable for probing the structural dynamics, stability, and interaction patterns of biomolecules, which can inform drug design, validate molecular hypotheses, and deepen our understanding of disease mechanisms. For this research, we employed the academic version of Desmond,²⁸,²⁹ a simulation software developed by D. E. Shaw Research group. The MD simulation workflow was classified into two main phases: system setup and production run. Using Maestro’s System Builder, we constructed the simulation environment by placing the molecular complex in an orthorhombic simulation box filled with Simple Point Charge (SPC) water, ensuring a 10 Å buffer on all sides. The system was neutralized by adding one sodium ion for the PMP system and one chloride ion for the oxypeucedanin hydrate (OPH) system. The gamma-glutamyl cysteine (GGC) system was already neutral and required no further adjustment. The final atom counts for the systems were 74,941 for PMP, 74,970 for OPH, and 74,978 for GGC. The production phase of the simulation was run for 100 ns, with snapshots recorded every 100 ps, resulting in 1,000 frames for each system. Simulations were performed under constant number of particles, pressure, and temperature (NPT) conditions, maintaining a constant temperature of 300 K and a pressure of 1.01325 bar³⁰ with an energy recording interval set at 1.2 ps. Upon completion, the resulting trajectories were analyzed using the Simulation Interaction Diagram tool, which enabled a detailed assessment of structural deviations, flexibility, and the nature of molecular interactions throughout the simulation.

Additionally, MM/GBSA analysis was carried out on MD simulation trajectories to estimate the binding free energy and total energy of the complex. The analysis was executed using the following commands on Ubuntu 22: export SCHRODINGER=/opt/Schrodinger2024-2 $SCHRODINGER/run thermal_mmgbsa.py desmond_md_ job_FILE-NAME-out.cms

Results and Discussion

Induced Fit Docking

Induced Fit Docking was performed on the top 3 results from docking studies, the results of which are mentioned in Table 1 (among the top five poses for each ligand from the induced-fit docking (IFD) analysis, the poses with the highest docking scores were identified). Also, the interactions of IFD are represented in Figure 2.

Table 1
Representing Induced Fit Docking (IFD) and docking scores of the top 3 ligands after performing Induced-Fit Docking

Ligand	Docking score / (kcal mol-1)	IFD score / (kcal mol-1)	Bonds (single-letter codes for amino acid residues)
PMP	-7.322	-30890.58	H-bond (N241, T242, C80, G79, Q199, K200, D176), Pi-cation (W107), Pi-Pi stacking(W107), Salt Bridge (K200, D176)
GGC (Isomer 1) (PubChem ID: 123938)	-8.266	-30863.77	H-bond (Q199, G79, C80, K200, N241, R45, HIE44) Pi-Pi cation (W107), Salt Bridge (R45, R336)
GGC (Isomer 2) (PubChem ID: 123938)	-7.401	-30877.20	H-bond (K200, Q199, G79, T242, R45), Pi-cation (W107), Salt Bridge (R45, R336, R342)
OPH (PubChem ID: 17536)	-9.750	-30804.32	H-bond (D176, N154, C80, N241, T242), Pi-cation (K200)

Figure 2
2-D interaction diagram of Induced-Fit docked ligands namely (a) PMP, (b) GGC (Isomer 1), and (c) OPH.

During IFD, no major deviation in protein backbone from the reference structure was observed except in strand 3 chain B of OPH-bound PSAT1, followed by minor deviations in strand 6 chain B, helix 3 chain B, and loop 14 chain B, as shown in Figure 3. However, deviations in the binding site side chains were observed in residues Cys80, Asn241, Arg45, Arg336, Hie335, Val157, Thr156, Trp107, and Lys200, as shown in Figure 3.

Figure 3
(a) Protein backbone changes due to induced-fit docking by superimposing the protein structures. (b) Protein sidechain flexibility due to inducedfit docking by superimposing the protein structures.

Pharmacokinetics and interaction fingerprints studies

The QikProp descriptors and their standard values were computed for the GGC (Isomer 1), PMP, and OPH mentioned in Table S1 (Supplementary Information (SI) section). The analysis suggests that this compound generally aligns with the standard values for various descriptors. For example, with respect to chemical functionalities such as the number of acidic (#acid), amide (#amide), amidine (#amidine), and amine (#amine) groups, all selected molecules fall within the accepted ranges. Likewise, binary descriptors like #in34 and #in56, along with numeric descriptors including #noncon, #nonHatm, #rin-gatoms, #rotor, #rtvFG, and #stars, conform to the indicated standards. In the case of OPH, it shows high permeability (QPPcaco = 1369.050) and low metabolic liability (#metab = 4), suggesting it will be less prone to drug-drug interactions. Octanol/water partition coefficient (QPlogPo/w) of 1.571 is well-balanced, aiding in good absorption without excessive hydrophobicity. QPlogKhsa of -0.31 indicates moderate human serum albumin binding. Blood-brain barrier permeability is low, QPlogBB (-0.977) suggests minimal central nervous system (CNS) penetration, ideal for a non-CNS drug. Overall, OPH achieves high permeability and moderate protein binding, with good metabolic stability, making it the top candidate. Furthermore, the generated molecular interaction fingerprints (see Figure 4) highlight a variety of amino acid residues involved and their frequencies, offering valuable information about the possible interaction patterns between the studied compounds and the protein target. In molecular interaction fingerprints, OPH shows the highest number of ligand interactions, and the interactions are similar to PMP except in the case of interaction with PHE83, where no interaction was observed between GCC (Isomer 1) and OPH, but can be observed in the case of PMP (Figure 4). However, PHE83 is not involved in any activity or ligand positioning in PSAT1.

Figure 4
Molecular interaction fingerprints of the docked poses are presented, indicating the specific interacting residues and their respective frequencies. To enhance clarity in visualizing the interaction patterns, the protein’s N- to C-terminus is color-coded from blue to red (PHE38 is the only residue that is associated with only PMP and not with other inhibitors).

WaterMap analysis

OPH has the most favorable binding free energy, indicating the strongest binding affinity among the three molecules. GGC (Isomer 1) has a moderate binding affinity, while PMP has the weakest binding affinity. WaterMap enthalpy measures the enthalpic contribution of water displacement upon binding. Lower positive values (like in GGC (Isomer 1)) suggest that fewer water molecules with strong enthalpic contributions need to be displaced, which can be energetically favorable. OPH has a relatively high enthalpic value, meaning it may need to displace more tightly bound water molecules, but this is compensated by its binding energy. Entropy here indicates the change in water entropy upon ligand binding. More negative entropy values (like in OPH) can mean a significant ordering of water molecules in the binding site or restricted freedom of water molecules, which contributes to an overall more stable binding configuration. OPH has the most negative entropy, suggesting a favorable entropy contribution to binding stability. WaterMap dG represents the combined Gibbs free energy contribution of the solvation process. The more negative values for GGC (Isomer 1) and OPH indicate that these molecules favorably interact with water molecules upon binding, with OPH having the most favorable waterrelated free energy change, followed closely by GGC (Isomer 1). PMP’s WaterMap dG is less negative, indicating weaker water interactions (Table 2) (Figure 5).

Table 2
WaterMap analysis results and scores of PMP, OPH, and GGC (Isomer 1)

Descriptor	PMP	GGC (Isomer 1)	OPH
WM/MM2.1 dG bind / (kcal mol-1)	-10.833	-14.207	-29.424
WaterMap enthalpy / (kcal mol-1)	17.999	6.704	15.066
WaterMap entropy / (kcal mol-1)	-22.440	-19.729	-30.076
WaterMap dG / (kcal mol-1)	-4.441	-13.025	-15.410

Figure 5
Hydration sites around the binding pocket of ligand (a) PMP, (b) GGC (Isomer 1), (c) OPH (red means favorable hydration sites and green refers to unfavorable hydration sites).

Molecular dynamics simulation

Molecular dynamics simulations are a cornerstone in modern drug discovery, enabling an in-depth exploration of the dynamic properties and interaction mechanisms between drugs and their biological targets. In this study, we carried out a 100-nanosecond simulation to monitor and analyze the structural variations, conformational flexibility, and internal interaction networks of the drugtarget complexes over the course of the simulation. The following sections present a detailed summary of our observations and findings.

Root-mean-square deviation

Root mean square deviation (RMSD) analysis was employed to monitor structural fluctuations in both ligands and proteins throughout the nanosecond timescale of the simulation. This method offers detailed insight into molecular stability and structural dynamics during the simulation process. In the case of the PSAT1-PMP complex, the initial deviation observed at 0.10 ns was 1.17 Å for protein backbone and 0.67 Å for PMP. At 100 ns, only minor deviations from initial values were observed, with protein backbone deviating to 1.55 Å and PMP deviating to 0.97 Å (Figure 6a). Furthermore, for PSAT1-GGC (Isomer 1) complex, initial deviation at 0.10 ns for protein backbone was observed at 1.34 Å and for GGC (Isomer 1) at 1.43 Å, upon further stabilization, no major fluctuation was observed till 100 ns and RMSD values observed for protein backbone and GGC (Isomer 1) was found to be 1.76 and 1.79 Å, respectively (Figure 6b).

Figure 6
Root mean square deviation (RMSD) of the PSAT1 protein in complex with (a) PMP, (b) GGC (Isomer 1), and (c) OPH. In these plots, red represents ligand deviations, blue corresponds to Cα atoms, and light green indicates the protein backbone. The root mean square fluctuation (RMSF) analysis is also shown, where the dark green lines illustrate interaction patterns of (d) PMP, (e) GGC (Isomer 1), and (f) OPH with PSAT1.

However, in the PSAT1-OPH complex, initial deviation at 0.10 ns was found to be 1.10 Å for the protein backbone and 1.65 Å for OPH. As the system stabilized, no major fluctuation was observed until 92.80 ns, where the protein backbone remained stable, but the OPH molecule showed an RMSD value of 3.44 Å. Upon trajectory analysis of frames, it was observed that at frame 928 or 92.80 ns, a fluctuation in Trp107 led to breakage in the bonds with OPH and allowed that site to be occupied by a water molecule, preventing further consistent bond formation with Trp107. However, at the end of the simulation, at around 100 ns, the bonding with Trp107 was observed again, leading to an RMSD value for protein backbone to be 1.70 Å and for OPH to be 3.09 Å (Figure 6c).

Root-mean-square fluctuation

Root mean square fluctuation (RMSF) measures the average positional variations of atoms within a molecule over the course of a simulation, reflecting the flexibility of specific regions. This analysis is essential for understanding protein dynamics and identifying highly mobile segments.

In the PSAT1-PMP complex formation, the residues involved in the structural dynamics of the complex (showing deviations greater than 2 Å) include Ala310, Lys311, Gly312, Asp313, Gln369 and Leu370 from chain A, and Gln6, Leu370 from chain B, whereas the residues that interacted with PMP and plays role in stability, functionality of the complex includes His44, Arg45, Tyr240, Asn241, Thr242 from chain A and Pro12, Gly13, Gly79, Cys80, Phe83, Trp107, Lys110, Glu155, Thr156, Val157, Asp176, Ser178, Gln199, Lys200, Thr208 and Tyr346 from chain B of PSAT1 (Figure 6d).

On the other hand, PSAT1 in complex with GGC (Isomer 1) the residues that deviated for more than 2 Å include Gly170, Asn309, Ala310, Leu370 from chain A and Gly34, Asn309, Ala310, Leu370 from chain B, whereas the residues that interact with GGC (Isomer 1) include Ser43, His44, Arg45, Tyr240, Asn241, Thr242 from chain A and Pro12, Gly13, Cys80, Trp107, Thr156, Gly198, Gln199, Lys200, Val207, His335, Arg342 from chain B (Figure 6e).

In the PSAT1-PMP interaction, the residues exhibiting deviations greater than 2 Å include Asn30, Ala310, Lys311, Leu370 from chain A and Gln6, Asn309, Ala310, Gln369, Leu370 from chain B. The residues that interacted with OPH include Ser43, His44, Arg45, Thr242 from chain A and Pro12, Gly13, Gly77, Gly79, Cys80, Trp107, Asn154, Glu155, Thr156, Gly206, Val207, Arg342 from chain B of PSAT1 (Figure 6f).

Simulation interaction analysis

The simulation interaction diagram provides a visual representation of the dynamic interactions between the protein and ligand throughout the simulation. It captures various orientations, conformations, and optimal binding poses occurring during their interaction.

Several H-bonds are formed between various residues of PSAT1 with PMP namely Asp179, Lys200, Thr156, Asn241, Gly79, Gln199, Thr242, Cys80, Thr208 and 4 water molecules. Also, Pi-Pi stacking and Pi-cation bonding can be seen between Trp107 and PMP (Figure 7a).

Figure 7
Simulation Interaction Diagrams of PSAT1 in complex with (a) PMP, (b) GGC (Isomer 1), and (c) OPH. The corresponding interaction counts are presented for (d) PMP, (e) GGC (Isomer 1), and (f) OPH. In these diagrams, red represents ionic interactions, blue denotes water bridges, grey indicates hydrophobic contacts, and light green highlights hydrogen bonds.

In case of GGC (Isomer 1), H-bonds can be observed in Arg45, Gly13, Cys80, Gly79, Asn241, Gln199, Lys200, Pro12 and 8 water molecules. Also, Pi-cation bond is formed between Trp107 and GGC (Isomer 1) (Figure 8). PSAT1 residues also form H-bonds with OPH, namely, Gln199, Gly79, Thr156, Asn154, Asp176, and 6 water molecules. Also, 2 Pi-Pi stacking bonds are observed between Trp107 and OPH (Figure 7c).

Figure 8
Binding free energy and total complex energy profiles of the PSAT1 complex with PMP, GGC (Isomer 1), and OPH at a 100-frame interval

Molecular mechanics and MM/GBSA studies

MM/GBSA analysis of MD simulation trajectories offers valuable insights into the binding free energies and stability of ligand-receptor complexes. By evaluating individual frames from the simulation, this method quantifies binding affinities and energetics at an atomic level, making it a critical tool in drug discovery, lead optimization, and the exploration of molecular mechanisms.

Table S2 (SI section) summarizes the comparative MM/GBSA results for GGC (Isomer 1), OPH, and PMP, each analyzed over 1001 structural snapshots. PMP demonstrated an average binding free energy (dG) of -48.079 kcal mol^-1, with a standard deviation (SD) of 3.282, ranging from -56.547 to -37.377 kcal mol^-1. Its average non-solvated binding energy (dG(NS)) was -51.189 kcal mol^-1 (SD: 3.094), with a range of -59.882 to -40.193 kcal mol^-1. The average total complex energy (dG complex) was calculated as -21297.068 kcal mol^-1, with a standard deviation of 88.471 and values spanning from -21538.852 to -20955.712 kcal mol^-1.

GGC (Isomer 1) showed an average dG of -40.732 kcal mol^-1 (SD: 3.428), with values between -51.206 and -28.331 kcal mol^-1. The corresponding dG(NS) averaged -44.493 kcal mol^-1, with a standard deviation of 3.706, ranging from -56.054 to -29.499 kcal mol^-1. The complex energy averaged -21235.733 kcal mol^-1 with a standard deviation of 81.193 and values ranging from -21475.287 to -20889.007 kcal mol^-1.

OPH exhibited the strongest binding, with an average dG of -63.877 kcal mol^-1 (SD: 5.198), ranging from -81.246 to -41.247 kcal mol^-1. Its average dG(NS) was -67.724 kcal mol^-1 with a standard deviation of 5.429, spanning from -84.904 to -43.113 kcal mol^-1. The complex energy for OPH averaged -21248.135 kcal mol^-1 (SD: 94.629), ranging between -21534.511 and -20978.727 kcal mol^-1.

The analysis of the PMP, GGC (Isomer 1), and OPH complexes, based on their MM/GBSA binding free energy and total complex energy across various frames (Figure 8), offers important insights into their stability. In the case of OPH, it shows a significantly better dG bind throughout from frame 0 to frame 1000 with minor fluctuations around frame 300 with dG bind of -56.7 kcal mol^-1, but overall, relatively stable binding is observed. The complex energy of PSAT1 and OPH is also comparable, with a minor difference between PMP and GGC (Isomer 1), indicating very similar complex stability among the 3 ligand complexes. In contrast, the dG bind of PMP turns out to be more stable compared to the other ligands, with minor fluctuations at frame 400 with dG bind of -42.2 kcal mol^-1, but relatively stable throughout the frames. Also, the overall complex energies of PSAT1-PMP are slightly better than other complexes (Figure 9b). GGC (Isomer 1) shows the lowest dG bind of all the ligands, lagging slightly behind PMP with minor fluctuations at frame 300 and frame 900, but overall stable binding.

Figure 9
(a) Binding free energy across all simulation frames and (b) total complex energy of the PSAT1 complex with PMP, GGC (Isomer 1), and OPH over the course of the simulation.

OPH shows significantly stronger binding affinity with an average dG bind of -63.877 kcal mol^-1 when compared to PMP (-48.079 kcal mol^-1) and GGC (Isomer 1) (-40.732 kcal mol^-1). Also, the dG bind (NS) of OPH (-67.724 kcal mol^-1) is significantly better when compared to GGC (Isomer 1) (-44.493 kcal mol^-1) and PMP (-51.189 kcal mol^-1), providing additional evidence for the enhanced binding affinity of OPH. MM/GBSA results suggest that OPH forms a more stable complex, exhibiting stronger and uniform binding affinity compared to GGC (Isomer 1) (Figure 9). The fluctuations observed in total complex energy and binding free energy across simulation frames underscore the dynamic nature of molecular interactions and the range of conformations adopted throughout MD simulation. Despite these variations, OPH steadily shows a more robust and stable binding profile than both GGC (Isomer 1) and PMP-an important factor in drug development and lead optimization.

Ovarian cancer remains a formidable gynecological malignancy, with high mortality largely attributed to latestage detection, making early diagnosis and treatment an urgent global health priority. This study explores an innovative computational strategy aimed at early-stage intervention by identifying potential TCM compounds that target PSAT1-a critical enzyme involved in the serine biosynthesis pathway, known to support cancer cell proliferation and survival.

The approach integrates traditional medicinal knowledge with modern computational tools to address limitations of existing therapies. The workflow began with the retrieval of PSAT1’s structure from the RCSB Protein Data Bank, followed by the screening of 381 TCM-derived compounds. Molecular docking, carried out using Glide, predicted ligand binding to PSAT1, leading to the selection of OPH and gamma-glutamyl cysteine (GGC, isomer 1) as promising candidates. Their pharmacokinetic properties were further assessed via the QikProp module in Maestro. To evaluate the dynamic stability and interaction profiles of the selected complexes, 100 ns MD simulations were conducted. Results from docking revealed strong binding affinities for both OPH and GGC (Isomer 1), with key interactions involving residues such as Trp107, essential for PSAT1’s activity. Notably, interactions including p-p stacking and p-cation contacts were observed, reinforcing their potential as inhibitors. QikProp analysis confirmed that both compounds generally conform to drug-likeness criteria, although slight variations in descriptors pointed to differences in pharmacokinetic behavior. Molecular interaction fingerprints further revealed a wide range of interacting amino acid residues, deepening the understanding of ligand-protein interactions. The MD simulations confirmed complex stability over time, with initial fluctuations (as shown by RMSD and RMSF analyses) stabilizing in later stages. Interaction diagrams supported the docking outcomes, with overlapping interaction residues during both docking and simulation phases. MM/GBSA energy calculations provided a robust validation step, showing OPH to possess the most favorable binding energy profile among the tested ligands. Furthermore, the combined results from WaterMap and MM/GBSA analyses highlighted OPH’s superior thermodynamic stability, with a dG bind of -63.88 kcal mol^-1 and favorable hydration free energy (-15.41 kcal mol^-1), indicating a strong and energetically favorable interaction within the PSAT1 active site. The stability observed across 100 ns MD simulations, supported by minimal backbone RMSD deviation (ca. 1.7 Å), underscores the structural resilience of the OPH-PSAT1 complex. These findings, together with its high predicted oral absorption (83.94%) and permeability indices, suggest that OPH not only binds effectively but also exhibits properties consistent with drug-like behavior. Collectively, this integrative in silico investigation provides mechanistic insights into PSAT1 inhibition and establishes oxypeucedanin hydrate as a promising lead molecule for further preclinical exploration in early-stage ovarian cancer.

Conclusions

The computational findings support the hypothesis that OPH could act as a potent inhibitor of PSAT1, especially during the early stages of ovarian cancer. As a naturally derived compound, OPH exhibits low toxicity, making it a promising lead for further drug development. Through a comprehensive computational pipeline, including molecular docking, induced-fit docking, WaterMap analysis, 100 ns MD simulations, and MM/GBSA energy calculations, OPH demonstrated superior binding affinity (dG bind = -63.88 kcal mol^-1) and conformational stability compared with both PMP and GGC (Isomer 1). Its favorable pharmacokinetic profile, low predicted toxicity, and strong interaction with catalytically significant residues (Trp107, Lys200, Asn241, Thr242) collectively underscore its potential as a natural inhibitor. These computational findings establish a solid foundation for further in vitro and in vivo investigations to confirm the inhibitory mechanism of OPH and therapeutic viability in ovarian cancer management.

Supplementary Information

Supplementary information is available free of charge at http://jbcs.sbq.org.br as file.

Acknowledgments

Research on Fundamental Technical Quality Indicator System for Inclusion of State-approved Miao Ethnomedicine “Fenghe Chubi Tincture” in Chinese Pharmacopoeia, Qiankehe Support (2022) General 289; Technological Innovation Team for Material Basis, Pharmacodynamic Evaluation and Mechanism of Action of Characteristic Miao Ethnomedicine, Qianjiaoji (2023) No. 099.

Data Availability Statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

References

• ¹ International Agency for Research on Cancer; Cancer Today [Link] accessed in December 2025
  » Link
• ² Berek, J. S.; Bast Jr., R. C. In Holland-Frei Cancer Medicine, 6^th ed.; Kufe, D. W.; Pollock, R. E.; Weichselbaum, R. R., eds.; BC Decker: Hamilton, Canada, 2003, ch. 118.
• ³ The ICON and AGO Collaborators; Lancet 2003, 361, 2099. [Crossref]
  » Crossref
• ⁴ Alam, S. S. M.; Samanta, A.; Uddin, F.; Ali, S.; Hoque, M.; Pharmacol. Rep. 2023, 75, 907. [Crossref]
  » Crossref
• ⁵ Zhang, R.; Gao, X.; Bai, H.; Ning, K.; Front. Pharmacol. 2020, 11, 538. [Crossref]
  » Crossref
• ⁶ Zhang, Y.; Li, J.; Dong, X.; Meng, D.; Zhi, X.; Yuan, L.; Yao, L.; OncoTargets Ther. 2020, 13, 5443. [Crossref]
  » Crossref
• ⁷ Zhang, Y.; Qazi, S.; Raza, K.; Saudi J. Biol. Sci. 2021, 28, 4069. [Crossref]
  » Crossref
• ⁸ Marchesani, F.; Zangelmi, E.; Murtas, G.; Costanzi, E.; Ullah, R.; Peracchi, A.; Bruno, S.; Pollegioni, L.; Mozzarelli, A.; Storici, P.; Campanini, B.; Protein Sci. 2023, 32, e4609. [Crossref]
  » Crossref
• ⁹ Ross, K. C.; Andrews, A. J.; Marion, C. D.; Yen, T. J.; Bhattacharjee, V.; Mol. Cancer Ther. 2017, 16, 1596. [Crossref]
  » Crossref
• ¹⁰ Metcalf, S.; Dougherty, S.; Kruer, T.; Hasan, N.; Biyik-Sit, R.; Reynolds, L.; Clem, B. F.; Clin. Exp. Metastasis 2020, 37, 187. [Crossref]
  » Crossref
• ¹¹ QikProp, Schrödinger Release 2024-2; Schrödinger, LLC, New York, USA, 2024.
• ¹² Pan, Y.; Gangwar, S.; Abbas, M.; Raza, K.; Sci. Rep. 2025, 15, 11708. [Crossref]
  » Crossref
• ¹³ Zhang, J.; Gangwar, S.; Bano, N.; Ahmad, S.; Alqahtani, M. S.; Raza, K.; PLoS One 2025, 20, e0313585. [Crossref]
  » Crossref
• ¹⁴ Maestro, Schrödinger Release 2024-2; Schrödinger, LLC, New York, USA, 2024.
• ¹⁵ Prime, Schrödinger Release 2024-2; Schrödinger, LLC, New York, USA, 2024.
• ¹⁶ Johnston, R. C.; Yao, K.; Kaplan, Z.; Chelliah, M.; Leswing, K.; Seekins, S.; Watts, S.; Calkins, D.; Chief Elk, J.; Jerome, S. V.; Repasky, M. P.; Shelley, J. C.; J. Chem. Theory Comput. 2023, 19, 2380. [Crossref]
  » Crossref
• ¹⁷ Søndergaard, C. R.; Olsson, M. H. M.; Rostkowski, M.; Jensen, J. H.; J. Chem. Theory Comput. 2011, 7, 2284. [Crossref]
  » Crossref
• ¹⁸ Lu, C.; Wu, C.; Ghoreishi, D.; Chen, W.; Wang, L.; Damm, W.; Ross, G. A.; Dahlgren, M. K.; Russell, E.; Von Bargen, C. D.; Abel, R.; Friesner, R. A.; Harder, E. D.; J. Chem. Theory Comput. 2021, 17, 4291. [Crossref]
  » Crossref
• ¹⁹ Shukla, Y.; Kalra, N.; Cancer Lett. 2007, 247, 167. [Crossref]
  » Crossref
• ²⁰ Astragalus membranaceus; [Link] accessed in January 2026
  » Link
• ²¹ Hofseth, L. J.; Wargovich, M. J.; J. Nutr. 2007, 137, 183S. [Crossref]
  » Crossref
• ²² Kuttan, R.; Bhanumathy, P.; Nirmala, K.; George, M. C.; Cancer Lett. 1985, 29, 197. [Crossref]
  » Crossref
• ²³ Ye, Q.; Zhou, X.; Han, F.; Zheng, C.; Front. Chem. 2023, 11, 1137547. [Crossref]
  » Crossref
• ²⁴ Dr. Duke’s Phytochemical and Ethnobotanical Databases. [Link] accessed in January 2026
• ²⁵ LigPrep, Schrödinger Release 2024-2; Schrödinger, LLC, New York, USA, 2024.
• ²⁶ Induced Fit Docking, Schrödinger Release 2024-2; Schrödinger, LLC, New York, USA, 2024.
• ²⁷ WaterMap, Schrödinger Release 2024-2; Schrödinger, LLC, New York, USA, 2024.
• ²⁸ Desmond Molecular Dynamics System, Schrödinger Release 2024-2; Schrödinger, LLC, New York, USA, 2024.
• ²⁹ Bowers, K. J.; Sacerdoti, F. D.; Salmon, J. K.; Shan, Y.; Shaw, D. E.; Chow, E.; Xu, H.; Dror, R. O.; Eastwood, M. P.; Gregersen, B. A.; Klepeis, J. L.; Kolossvary, I.; Moraes, M. A.; Sacerdoti, F. D.; Proceedings of the 2006 ACM/IEEE conference on Supercomputing; ACM Press: New York, USA, 2006. [Crossref]
  » Crossref
• ³⁰ McDonald, I. R.; Mol. Phys. 1972, 23, 41. [Crossref]
  » Crossref

Edited by

• Editor handled this article:
  Paulo Augusto Netz (Associate)

Publication Dates

• Publication in this collection
  09 Mar 2026
• Date of issue
  2026

History

• Received
  26 Sept 2025
• Accepted
  23 Jan 2026