Molecular dynamics and free energy analysis of curcumin binding near the DNA-recognition interface of NF-κB Authors: Vandana Kumari, Susruta Samanta Published: 2026 ABSTRACT AbstractCurcumin, a natural polyphenolic compound derived from Curcuma longa, is widely reported to modulate NF-κB signaling, primarily through indirect effects on upstream regulatory pathways. However, residue-level information on possible curcumin association near structurally relevant regions of NF-κB remains limited. Here, we investigated the predicted binding behavior of curcumin near the DNA-recognition region of the NF-κB p50 homodimer using molecular docking, 100 ns all-atom molecular dynamics simulation, and MM-PBSA free-energy analysis. Docking with the p50 homodimer model derived from PDB ID 1SVC identified a preferred pose near the Rel Homology Domain with an AutoDock Vina score of − 6.0 kcal/mol. Molecular dynamics simulations showed that the curcumin-bound model maintained persistent local contacts and exhibited reduced residue-level fluctuations near the predicted binding region. The complex showed a compact structural profile, with Radius of gyration (Rg) values in the range of 2.50–2.60 nm, and lower solvent-accessible surface area relative to the apo model. MM-PBSA analysis indicated a favorable relative binding free energy dominated by van der Waals and electrostatic contributions. These findings provide a computational residue-level characterization of curcumin association near the NF-κB p50 DNA-recognition region.Keywords: NF-κB signaling, Molecular dynamics simulation, MM-PBSA binding free energy, Anti-inflammatory therapeutics, Natural bioactive compoundsSubject terms: Biochemistry, Biophysics, Cancer, Chemistry, Computational biology and bioinformatics, Drug discovery, Structural biology Introduction Inflammation is a coordinated protective response to infection, tissue injury, and cellular stress. Numerous inflammatory mediators are synthesized and secreted by different cells in this intricate process, which results in a variety of cellular effects1. Cytokines like interleukins (ILs), tumor necrosis factor (TNF), and chemokines, are examples of these inflammatory mediators2,3. Because they attract immune cells to the location of inflammation, facilitate cellular communication, and control the level of inflammation, cytokines are essential for coordinating the immune response. Among the transcriptional regulators involved in this response, nuclear factor kappa B (NF-κB) plays a central role in controlling genes associated with inflammation, immune regulation, apoptosis, and cell survival4–7. NF-κB can be activated by cytokines, pathogens, oxidative stress, and other cellular stimuli, after which it translocates to the nucleus and binds specific κB DNA sequences to regulate target-gene transcription4. Numerous diseases, including cancer, autoimmune diseases, chronic inflammatory disorders, type 2 diabetes mellitus, rheumatoid arthritis, asthma, cardiovascular disease, and neurodegenerative conditions have been linked to the progression of NF-κB signaling dysregulation and uncontrolled inflammation5. NF-κB remains an important target for mechanistic and therapeutic investigation8–10. Several pharmacological strategies have been explored to regulate NF-κB signaling, including inhibition of IκB kinase activity, prevention of IκBα degradation, proteasome inhibition, and modulation of nuclear translocation1–3. These approaches show that NF-κB signaling can be chemically regulated at multiple levels, although compounds acting on this pathway differ substantially in mechanism, selectivity, toxicity, and clinical relevance11. This has also encouraged interest in natural bioactive compounds that influence inflammation-associated signaling pathways, including curcumin. Curcumin has attracted attention because of its reported ability to regulate NF-κB signaling through multiple upstream and downstream mechanisms4,5,12. Curcumin, a polyphenolic compound derived from Curcuma longa, has been widely studied for anti-inflammatory, antioxidant, and anticancer activities13,14. Several experimental studies report that curcumin suppresses NF-κB signaling primarily through indirect mechanisms, including inhibition of IκB kinase activity, reduced phosphorylation and degradation of IκBα, modulation of PI3K/Akt and MAPK pathways, and reduced nuclear translocation of NF-κB6–8. Curcumin also affects downstream NF-κB-regulated processes, including the expression of pro-inflammatory cytokines such as TNF-α and IL-6. Despite these reports, residue-level information on possible curcumin association near structurally relevant regions of NF-κB remains limited. Curcumin and its analogues have also been examined in several cellular and disease models, including inflammatory and cancer-related systems, but their molecular effects are often pleiotropic and context-dependent9. Therefore, computational studies can be useful for generating residue-level hypotheses regarding possible curcumin interactions with selected protein targets. In the present work, we focus specifically on the NF-κB p50 homodimer model and evaluate curcumin binding near the DNA-recognition region using docking, molecular dynamics simulations, and end-point free-energy analysis. A structural overview of curcumin and NF-κB p50 homodimer with the predicted curcumin-binding region used for subsequent analysis is shown in Fig. 1. Structural overview of curcumin and the NF-κB p50 homodimer model. The chemical and optimized structures of curcumin are shown on the left. The NF-κB p50 homodimer structure derived from PDB ID 1SVC is shown on the right, with representative residues near the predicted curcumin-binding region highlighted. NF-κB comprises a family of related subunits that can form multiple homo- or heterodimers. A heterodimer of two subunits, p50 (encoded by the NFKB1 gene) and p65 (encoded by the RELA gene), is the most common form of NF-κB. Additionally, subunits such as p52, RelB, and c-Rel can be combined to create distinct dimeric forms of NF-κB. The conserved domains seen in each of these subunits are essential for NF-κB’s structural stability and functional capacities10,11,15. All NF-κB subunits contain an N-terminal Rel Homology Domain (RHD), which mediates dimerization, DNA recognition, nuclear localization, and interaction with IκB protein. Structural studies of NF-κB–DNA complexes have shown that residues within the RHD contribute to both base-specific recognition and phosphate-backbone contacts with κB DNA sequences9,15. These structures provide a useful framework for identifying residues located near DNA-recognition regions and for examining whether small molecules can associate with nearby protein-surface sites. The crystal structure of the p50 homodimer bound to DNA has been solved at a resolution of 2.6 Å (PDB ID: 1SVC)13. In the present work, the p50 homodimer structure was used as a protein model to examine predicted curcumin association near this region. Because DNA was removed before docking and simulation, the model should be interpreted as a protein-only system and not as a direct simulation of the DNA-bound transcriptional complex. The crystal structures of NF-κB subunits and their complexes with DNA and transcriptional coactivators have provided significant insights into the molecular mechanisms behind NF-κB-mediated gene expression. These findings have facilitated the development of innovative therapy strategies that target NF-κB signaling in a variety of diseases, including inflammatory and cancerous disorders. Here, we investigate the predicted binding behavior of curcumin near the DNA-recognition region of the NF-κB p50 homodimer using molecular docking, 100 ns molecular dynamics simulations, and end-point binding free-energy analysis. The objective is to characterize residue-level contacts, local conformational dynamics, and relative energetic contributions associated with the curcumin-bound model. Methods Molecular docking All docking and MD analyses in this study were performed using the NF-κB p50 homodimer structure derived from PDB ID 1SVC as the receptor model. Molecular docking was performed using AutoDock Vina program because of its sampling efficiency and suitability for flexible ligand docking14. Standard software such as VMD, PyMOL, and Discovery Studio were used for visualization, analysis, and interpretation16. For the preparation of target proteins, specific PDB structures were used as references. The PDB structure used for NF-κB is 1SVC and it was obtained from standard repository (http://www.rcsb.org)13. This structure was selected because it provides a high-resolution crystal structure of the NF-κB p50 homodimer bound to κB DNA, enabling identification of residues located near the DNA-recognition region. Since the objective of the present work was to examine the predicted association of curcumin with the protein surface near this region, the p50 homodimer was used as the receptor model after removing DNA and crystallographic solvent molecules. Removal of DNA may alter local electrostatics and conformational preferences; therefore, the results are interpreted as a protein-only computational model rather than a direct simulation of the DNA-bound transcriptional complex. The protein structure was set up by removing all nonreceptor atoms including water molecules or ligands, followed by addition of hydrogen atoms using AutoDock Tools and VMD. Curcumin was prepared in its enol form, and the ligand geometry was energy-minimized before docking. Gasteiger charges were assigned, and both receptor and ligand structures were converted to PDBQT format. Because PDB ID 1SVC does not contain a co-crystallized small-molecule ligand, conventional ligand redocking validation was not applicable. Docking was therefore performed using a grid covering the protein surface near the DNA-recognition region identified from the crystal structure. Docked poses were ranked using the AutoDock Vina scoring function and further inspected for residue contacts, hydrogen-bonding interactions, and spatial proximity to residues associated with DNA recognition. MD simulation Following molecular docking to determine binding affinity and to identify interaction sites of curcumin with NF-κB, MD simulations of the NF-κB p50 homodimer and p50 homodimer–curcumin complex were performed for 100 ns to investigate the stability of binding and influence on protein dynamics. The NF-κB p50 homodimer structure (PDB ID: 1SVC) was retrieved from the Protein Data Bank. Missing atoms or residues were added, hydrogen atoms were positioned, and protonation states were adjusted based on the physiological pH. The protein topology was generated using the GROMACS software package and the GROMOS96 force field, whereas the ligand topology was generated using the SwissParam server (https://www.swissparam.ch/)17,18. Protonation states corresponding to pH 7.4 were assigned using pdb2gmx and visually inspected to confirm the binding site. The system was neutralized, solvated with SPC water, and adjusted to 0.15 M NaCl using ion parameters compatible with the SPC model19. It was centered in a cubic box with a minimum solute–edge distance of 1.2 nm. Periodic boundary conditions were applied in all three dimensions. The steepest descent method was used to minimize energy until the maximum force dropped below 1000 kJ·mol⁻¹·nm⁻¹. Positional restraints on solute heavy atoms (k = 1000 kJ·mol⁻¹·nm⁻²) were used during equilibration in two steps: (i) 500 ps of NVT at 300 K using the V-rescale thermostat (τT = 0.5 ps) and (ii) 2 ns at 300 K and 1 bar using the Parrinello–Rahman barostat (τP = 2.0 ps, compressibility 4.5 × 10⁻⁵ bar⁻¹)20. Equilibration was assessed from the stability of temperature, pressure, density (~ 1000 kg m⁻³), and potential energy before initiating production simulations. For nonbonded interactions, the Verlet neighbor scheme was used with a 1.0 nm real-space cutoff for both the Coulomb and Lennard-Jones terms. ong-range electrostatics were treated using the particle-mesh Ewald (PME) method with fourth-order interpolation and a Fourier grid spacing of 0.12 nm. Long-range dispersion corrections were applied to energy and pressure21. Although the GROMOS96 force field was originally developed with reaction-field electrostatics, PME was used here because it is widely applied in contemporary biomolecular simulations. The stability of density, pressure, and energy profiles was monitored during equilibration and production runs. During equilibration, positional restraints were applied to the heavy atoms of the protein and ligand to allow solvent relaxation. Production MD simulations were performed for 100 ns using a 2 fs integration timestep, with LINCS constraints applied to all covalent bonds22. Trajectory coordinates were saved at 10 ps intervals for subsequent analysis. Analyzing the resulting trajectory data allowed for the calculation of all reported parameters, including root mean square deviation (RMSD), root mean square fluctuation (RMSF), radius of gyration (Rg), solvent-accessible surface area (SASA), hydrogen-bond interactions, and contact profiles. The stability and convergence of the trajectories were assessed using RMSD, RMSF, Rg, SASA, hydrogen-bond persistence, contact analysis, block-averaged RMSD values, and running averages of the binding free energy calculated over the analyzed simulation window. Free energy calculation The calculation of binding free energies between proteins and ligands is useful for evaluating the relative strength and energetic favorability of the protein–ligand interactions. In this study, the MM-PBSA approach was employed to estimate the binding free energy of the NF-κB p50 homodimer–curcumin complex. The calculations were performed using the gmx_MMPBSA tool integrated with GROMACS23,24. The binding free energy was calculated from the difference between the free energy of the protein–ligand complex and the sum of the free energies of the isolated protein and ligand, as given below: The binding free energy was further decomposed into molecular mechanics energy, solvation free energy, and entropic contribution: Here, and represent the van der Waals and electrostatic contributions, respectively, while and represent the polar and nonpolar solvation energy terms. The polar solvation energy was computed using the Poisson–Boltzmann equation, while the nonpolar contribution was estimated from the solvent-accessible surface area (SASA) using γ = 0.0072 kcal mol− 1 Å−2 and β = 0.0 kcal mol-1. The dielectric constants were set to εin = 2 and εout = 80. Binding free energy calculations were performed using snapshots extracted from the equilibrated portion of the MD trajectory. The remaining parameters were kept as implemented in the gmx_MMPBSA workflow unless otherwise specified. In the present calculation, the entropic contribution () was not included; therefore, the reported values should be interpreted as relative MM-PBSA binding free energy estimates rather than absolute experimental binding free energies. Principal component analysis Principal component analysis (PCA) was used to examine dominant collective motions sampled by the apo NF-κB p50 homodimer and the curcumin-bound complex during MD simulation. In the context of MD simulations, PCA reveals coordinated atomic motions and major conformational changes. The method involves diagonalizing the covariance matrix to extract the principal components of the simulation trajectory. PCA was performed using the GROMACS tools gmx_covar and gmx_anaeig25. To compute free energy landscapes (FEL), the GROMACS tool gmx_sham was utilized. FEL was calculated using the equation: Here, PC1 and PC2 denote the reaction coordinates, kB is the Boltzmann constant, T is the simulation temperature, and P(PC1,PC2) represents the probability distribution. PCA was carried out on the backbone atoms of the protein after least-squares fitting of the trajectory to remove overall translational and rotational motion. The covariance matrix of positional fluctuations was diagonalized to obtain eigenvectors and eigenvalues representing the dominant collective motions sampled during the simulation. The trajectories of the apo protein and curcumin-bound complex were projected onto the first two principal components, PC1 and PC2, which were used as reaction coordinates for FEL construction. The percentage variance captured by PC1 and PC2 was calculated from the corresponding eigenvalues to assess the contribution of these modes towards total conformational fluctuation. Results Docking analysis of curcumin with the NF-κB p50 homodimer Molecular docking was performed to examine the predicted association of curcumin with the NF-κB p50 homodimer structure used in this study. The docking analysis generated multiple curcumin poses with AutoDock Vina scores ranging from − 5.3 to − 6.0 kcal/mol. The most favorable pose showed a docking score of − 6.0 kcal/mol and involved hydrogen-bonding interactions with Arg157 and Lys149. Other representative poses showed hydrogen-bonding contacts with Lys52, Arg59, Lys79, His67, Arg164, His118, and Glu120, as summarized in Table 1. These interactions suggest that curcumin can form polar contacts with residues located near the predicted binding region of the p50 homodimer. The docking results were used as the starting point for subsequent molecular dynamics simulations to evaluate the persistence of the predicted protein–ligand contacts under dynamic conditions. The interaction models with most favorable docking scores and binding residues are shown in Fig. 2 and the details are provided in Table 1. Table 1AutoDock Vina docking scores and predicted hydrogen-bond interactions between curcumin and residues of the NF-κB p50 homodimer.Binding affinity* (kcal/mol)H-Bond interactionNo. of H-BondsDistance between the donor and acceptor (Å)− 5.6LYS5241.92ARG591.96LYS792.24HIS672.60− 5.3ARG16432.34HIS1182.89GLU1202.20− 6.0ARG15722.17LYS1492.45*Binding affinity values denote docking scores from AutoDock Vina binding affinity estimates for the corresponding docked poses. Hydrogen-bond interactions and donor–acceptor distances were identified from the docked pose using Discovery Studio and are not independently calculated affinity values. AutoDock Vina docking scores and predicted hydrogen-bond interactions between curcumin and residues of the NF-κB p50 homodimer. *Binding affinity values denote docking scores from AutoDock Vina binding affinity estimates for the corresponding docked poses. Hydrogen-bond interactions and donor–acceptor distances were identified from the docked pose using Discovery Studio and are not independently calculated affinity values. Representative docked poses and interaction profiles of curcumin with the NF-κB p50 homodimer. Curcumin is shown in the predicted binding region of the p50 homodimer, and selected hydrogen-bonding and non-covalent contacts are shown for representative docking poses. Hydrophobic interactions are shown in pink-violet; electron donor-donor interactions are shown in red; π-sulfur interactions are shown in orange; hydrogen bonds are shown in light green; van der Waals interactions are shown in brown. The docking scores indicate moderate predicted affinity, which is this is expected behavior of a flexible natural polyphenol interacting with a protein-surface region and this is consistent with the earlier reports26. The docking results were used to identify representative binding poses for subsequent MD-based stability analysis. The surface depiction surrounding the ligand shows regions involved in the predicted protein–ligand interaction (details are shown in Fig. 3). The dark green sections on the surface represent sites that can accept hydrogen bonds, indicating their involvement in forming hydrogen bonds with the ligand. These interactions may contribute to the local stability of the complex between the ligand and protein, although docking alone cannot quantify their energetic contribution. The existence of hydrogen bond acceptors on the protein’s surface corresponds with the polar functional groups found in curcumin, such as hydroxyl and methoxy groups, promoting specific interactions. These observations highlight the significance of hydrogen bonding in molecular recognition and support the selection of this docking pose for subsequent dynamic stability analysis. Surface representation of the predicted curcumin-binding region. Hydrogen-bond donor and acceptor regions are shown using distinct surface colors, and the interacting residues are labelled for clarity. MD-based stability and flexibility analysis The trajectories of 100 ns all-atom MD simulations were analyzed using GROMACS utilities. Root Mean Square Deviations (RMSD) and Root Mean Square Fluctuations (RMSF) were investigated to evaluate protein stability and flexibility. The Radius of Gyration (Rg) and Solvent Accessible Surface Area (SASA) profiles were analyzed to assess the compactness and flexibility of the proteins. We also analyzed the Hydrogen Bond (H-bond) formations, number of contacts, and their minimum distances to identify interactions between curcumin with NF-κB p50 homodimer. RMSD values were calculated using gmx_rms after least-squares fitting to the protein backbone. Unless otherwise stated, RMSD was calculated for protein backbone atoms. Hydrogen atoms were excluded from the RMSD calculation. The RMSD of simulated NF-κB p50 homodimer remained relatively stable after the initial equilibration period, indicating that the protein model retained its overall fold during the simulation which is in agreement with the experimentally resolved crystal structure13. The RMSD plot of apo protein showed initial fluctuations during the equilibration phase, stabilizing after approximately 20 ns, indicating that the protein reached a stable conformation within the simulation timescale. The relatively stable RMSD profile suggests that the apo p50 homodimer did not undergo large-scale structural deviation under the applied simulation conditions indicating that the force field and simulation parameters were well-chosen for this system. In contrast, the RMSD analysis for the protein-ligand complex exhibited an increased RMSD compared to the unbound protein as shown in Fig. 4a. These higher values and dynamic behavior suggest that the binding of curcumin induces conformational adjustments in the protein. RMSD analysis reported values of approximately 0.2–0.4 nm during the early part of the simulation, followed by higher values of approximately 0.6–0.8 nm during the subsequent trajectory for the complex. The sharp increase in the RMSD values of the backbone of this complex was observed between 95 ns and 100 ns of simulation time, indicating late-stage conformational fluctuation within the simulated timescale. These observations are consistent with previous studies, which report that RMSD increased for protein-ligand systems due to induced fit and dynamic flexibility upon ligand binding27,28. To obtain a quantitative assessment of stability, block averaging was used to further examine the NF-κB-curcumin complex’s RMSD profile. After splitting the trajectory into five equal blocks, the average RMSD values were 1.29, 1.90, 2.03, 2.33, and 2.00 nm, respectively. The system achieved a relatively stable conformation during the simulation, despite moderate fluctuations, as indicated by the overall RMSD of 1.92 ± 0.34 nm. The dynamic nature of the RMSD highlights that curcumin binding is associated with local conformational adjustment of the p50 homodimer. (a) The evolution of RMSD throughout the course of the simulation. (b) The evolution of RMSF throughout the course of the simulation. (c) The RMSF of the residues of the binding domain. (d) The evolution of the radius of gyration throughout the course of the simulation. (e) The evolution of solvent accessible surface area (SASA) and the number of H-Bonds throughout the course of the simulation. (f) The SASA of NF-κB is shown in black and that of NF-κB-curcumin complex is shown in red. The Root Mean Square Fluctuation (RMSF) analysis highlighted reduced flexibility of amino acid residues upon curcumin binding compared to the protein alone, and it is shown in Fig. 4b and c. RMSF values were calculated for Cα atoms of Chain A of the NF-κB p50 homodimer. The reduction in RMSF, particularly in regions near the ligand-binding site, suggests that curcumin binding restricts local motions, potentially enhancing the protein’s stability. Reduced RMSF values have also been reported in similar protein-ligand studies, where ligand binding restricts the mobility of key residues27. The RMSF profile for residues 50–300 showed that several residues located near the predicted curcumin-binding region displayed altered fluctuation patterns in the curcumin-bound model. Fluctuations observed at residues 52, 59, 79, 118, 120, 149–167, 249, and 285 were reduced in the curcumin-bound trajectory relative to the apo model, suggesting localized stabilization of residues associated with the predicted binding region. Residues showing higher RMSF values correspond mainly to flexible loops or terminal regions. These results suggest that curcumin is associated with reduced local flexibility near the predicted binding region under the modeled conditions. Hydrogen-bond and contact persistence during MD simulation Hydrogen bonds (H-bonds) play a critical role in stabilizing the protein-ligand interactions contributing to the observed reduction in flexibility. Throughout the whole simulation process, multiple hydrogen bonds were observed forming and breaking between the protein and ligand, reflecting the dynamic nature of their interaction. Hydrogen bonds were analyzed using gmx_hbond with a donor–acceptor distance cutoff of 0.35 nm and a hydrogen–donor–acceptor angle cutoff of 30°29. The simulation revealed three stable H-bonds between curcumin and NF-κB p50 homodimer throughout the trajectory, indicating recurrent interactions. Representative structures of these interactions are shown in Fig. 5. The most persistent interactions were found with Gly116 and Ser66. With donor–acceptor lengths of 2.27 Å and 2.09 Å, two hydrogen bonds were continuously maintained between curcumin and Ser66, suggesting stable interactions. To further stabilize the ligand inside the binding site, a third H-bond was also established with Gly116, with an average distance of 2.51 Å. Representative structures showing the hydrogen bonds formed and consistent hydrophobic interactions in curcumin-NF-κB. H-bonds are shown in dark blue and hydrophobic interaction is shown by dotted lines. The frequent involvement of Ser66 in hydrogen bonding is of particular interest, because this residue is located near the DNA-recognition region of the p50 homodimer. The positioning of Ser66 in a flexible loop region may contribute to curcumin’s capacity to modulate conformational dynamics of the NF-κB dimer, which is critical for DNA binding and transcriptional regulation. Apart from hydrogen bonding, curcumin was stabilized within the NF-κB binding pocket by hydrophobic interactions. During the 100 ns simulation, two important residues, Val115 and Lys117, were consistently involved in non-polar interactions with the ligand. Val115 and curcumin’s aromatic ring system remained close, with an average hydrophobic contact distance of 3.13 Å. It is well known that valine’s branched alkyl side chain participates in alkyl–π and van der Waals interactions, which support ligand stability, especially in hydrophobic cores of transcription factor binding areas. Interestingly, Lys117, typically associated with polar or electrostatic interactions, also exhibited a consistent hydrophobic contact with curcumin at an average distance of 3.49 Å. While lysine side chains are charged, the aliphatic portion of the chain (CH₂ groups before the terminal amine) can participate in hydrophobic interactions when the terminal NH₃⁺ group is not directly involved in ionic bonding. This suggests that curcumin may be interacting with the lysine side chain in a way that exploits both hydrophobic and potential electrostatic complementarity. These H-bonds seem essential for maintaining the binding affinity and stability of the complex. The presence of stable H-bonds supports the hypothesis that curcumin can effectively bind and potentially modulate NF-κB activity. Previous studies have emphasized the importance of stable H-bonds in enhancing the inhibitory potential of molecules against NF-κB30. These outcomes are consistent with reported research showing curcumin blocks NF-κB’s nuclear translocation and transcriptional activity by directly interacting with its DNA-binding or transactivation domain31,32. Radius of gyration and solvent-accessible surface area The radius of gyration (Rg) measures the overall compactness and stability of the system. The Rg values for the protein-ligand complex were lower and more stable compared to the protein alone, suggesting that curcumin binding induces a more compact and stable structural arrangement of NF-κB (Fig. 4d). The Rg values of the curcumin-bound complex remained within approximately 2.50–2.60 nm during the 100 ns simulation. The lower Rg value of the curcumin-bound complex around ~ 18 ns reflects a transient compact conformation relative to the apo structure. This may arise from local rearrangement around the predicted ligand-binding region and partial reduction in solvent-exposed conformations. Since Rg is a global compactness descriptor, this observation is interpreted as a structural stability trend rather than direct evidence of functional inhibition. These findings align with earlier reports, where ligand binding resulted in reduced Rg, signifying enhanced structural integrity27. The reduced and stable Rg values indicate that curcumin contributes to the stabilization of NF-κB. During the molecular dynamics simulation, the solvent accessible surface area (SASA) of NF-κB in the unbound state was observed to be higher, indicating greater surface exposure to the solvent and it is shown in Fig. 4e and f. This suggests a more dynamic and flexible structure in the absence of the ligand. In contrast, the NF-κB-curcumin complex exhibited consistently lower SASA, reflecting reduced solvent accessibility. This decrease can be attributed to curcumin binding, which is associated with partial burial of surface residues near the predicted binding region. The reduced SASA in the complex highlights the role of curcumin in promoting reduced solvent exposure which may eventually interfere with its ability to interact with DNA and thus contributing towards stability. The curcumin-bound complex shows predominantly stable behavior throughout the duration of simulation. The average solvent-accessible surface areas (SASA) of NF-κB-curcumin complex was determined to be 190 nm2. Contact analysis To track the proximity and possible binding stability, the minimum distance between the chosen groups was calculated along the simulation trajectory. The number of contacts that were created within a 0.6 nm cutoff distance was measured. The gmx mindist tool from the GROMACS package was used to study the minimum distance and number of contacts to assess the intermolecular interactions between selected residue groups in the p50 homodimer–curcumin system. A constant high number of contacts and the maintenance of a minimum distance below 0.6 nm throughout the simulation indicated strong and stable interactions between the binding partners. The results provide an additional measure of the persistence of contacts in the important binding regions during the trajectory. A contact was defined when any atom of the ASP-GLU group was found within 0.6 nm of any atom from the LYS-ARG group. MM-PBSA binding free-energy analysis The binding free energy of the p50 homodimer–curcumin complex was estimated using the MM-PBSA implementation in gmx_MMPBSA33–39. This analysis was used to estimate the relative energetic favorability of the predicted protein–ligand complex and they are shown in Table 2. Because entropic contributions were not included, the calculated values are interpreted as relative end-point binding free-energy estimates rather than absolute experimental binding free energies. In the single-trajectory approach, bonded terms cancel during complex, receptor, and ligand energy subtraction and are therefore not discussed as contributors to binding. Table 2MM-GBSA/MM-PBSA binding free-energy components for the NF-κB–curcumin complex. Values are reported as mean ± SD with SEM calculated over the analyzed trajectory frames. Entropic contributions were not included. All values are in kcal mol–1.Energy componentAverageSDSEMΔVDWAALS− 24.942.650.26ΔEEL− 20.816.020.60ΔEPB33.785.270.53ΔESURF− 3.770.240.02ΔGGAS− 45.755.600.56ΔGSOLV30.005.190.52ΔTOTAL− 15.752.180.22 MM-GBSA/MM-PBSA binding free-energy components for the NF-κB–curcumin complex. Values are reported as mean ± SD with SEM calculated over the analyzed trajectory frames. Entropic contributions were not included. All values are in kcal mol–1. Energy components were visualized using gmx_MMPBSA_ana. Figure 6 shows the variation in energy values for the combined components (∆GGAS, ∆GSOLV, and ∆GTOTAL). A fluctuation in binding free energy and the corresponding running average fluctuation is seen over the final 100 ns of the MD simulation trajectory. The p50 homodimer–curcumin complex showed an average ΔTOTAL of − 15.75 kcal mol–1, indicating favorable relative binding energetics. The favorable gas-phase contribution (ΔGGAS = − 45.75 kcal mol–1), mainly arising from van der Waals and electrostatic terms, was partially offset by the positive solvation contribution (ΔGSOLV = 30.00 kcal mol–1). This energetic pattern suggests that curcumin association is driven primarily by nonbonded interactions, while polar solvation opposes binding. Variation and running average of MM-PBSA binding free-energy components for the p50 homodimer–curcumin complex. Discussion NF-κB signaling is closely associated with inflammatory and cancer-related processes, and several natural compounds have been reported to modulate this pathway through diverse mechanisms40,41,57. Curcumin is among the most widely studied natural products in this context, but its effects on NF-κB signaling are generally understood to occur mainly through indirect regulation of upstream signaling components, including IKK activity, IκBα degradation, and nuclear translocation of NF-κB42–50. The present study evaluates whether curcumin can form a computationally predicted association near the DNA-recognition region of the NF-κB p50 homodimer and how such association affects local structural dynamics within the selected protein-only model. The combined use of docking, MD simulation, PCA/FEL analysis, and MM-PBSA binding free-energy estimation provides complementary information on the predicted p50 homodimer–curcumin complex. Docking was used to generate an initial binding hypothesis and identify plausible residue contacts. MD simulation was then used to evaluate the persistence of these contacts under explicit-solvent dynamic conditions, while MM-PBSA analysis provided a relative estimate of the energetic contributions to the predicted complex. PCA and FEL analysis further helped describe conformational sampling within the simulated timescale. Each of these methods has limitations: docking scores are approximate, MD sampling depends on trajectory length and replication, and endpoint free-energy estimates are sensitive to force-field parameters and omission of entropy. The results are therefore interpreted as a computational model of local curcumin association with the p50 homodimer rather than as direct experimental evidence of biological inhibition. Molecular Docking identified a moderate-affinity curcumin pose near the DNA-recognition region of the p50 homodimer, with the most favorable AutoDock Vina score of − 6.0 kcal/mol supporting a plausible association. The predicted pose involved hydrogen-bonding and non-covalent contacts with residues located near the selected binding region, including Arg157 and Lys149. These interactions provided a structurally reasonable starting pose for subsequent MD simulation. Also, this analysis supports the hypothesis that curcumin may affect NF-κB’s DNA binding function, hence inhibiting its transcriptional activity as suggested by prior studies51. The RMSD profile of the curcumin-bound model showed larger fluctuations than the apo p50 homodimer, indicating that ligand association is accompanied by conformational adjustment within the simulated timescale. These changes suggest local structural rearrangements in the protein-only which may eventually result in impaired DNA recognition or altered interaction with cofactors. RMSF analysis indicated residue-level changes in flexibility upon curcumin binding. Higher fluctuations were observed mainly in loop or terminal regions, including residues 120–140 and 160–170, whereas several residues close to the predicted binding region showed reduced fluctuations in the curcumin-bound trajectory. This suggests localized stabilization near the predicted binding site. This reduced RMSF is interpreted as a change in local dynamics and indicates decrease in NF-κB activity. The Rg profile of the p50 homodimer–curcumin complex remained within a narrow range of approximately 2.50–2.60 nm throughout the trajectory, indicating that no major unfolding or global expansion occurred during the simulation. Together with the SASA profile, this supports a more compact and less solvent-exposed curcumin-bound model. These global descriptors are useful indicators of structural stability, and support the hypothesis that curcumin stabilizes the structure of NF-κB. Hydrogen-bond analysis showed recurrent interactions between curcumin and residues such as Ser66 and Gly116, along with hydrophobic contacts involving Val115 and Lys117. These contacts support the local stability of the predicted curcumin-bound pose during the 100 ns trajectory and the indicate inhibition of NF-κB function. Structural studies of NF-κB–DNA complexes have identified several residues involved in DNA recognition. In the p50 subunit, residues such as Arg54, Arg56, Tyr57, Glu60, and His64 contribute to base-specific recognition in the major groove, while residues including Cys59, Thr142, Lys144, Lys145, and Lys241 participate mainly in contacts with the DNA phosphate backbone13. These crystallographic observations provide a useful reference for interpreting the location of the predicted curcumin-binding region relative to known DNA-contacting residues. In our simulation study, curcumin showed persistent contacts with Ser66 and Lys117 as shown in Fig. 5. These residues are not classical DNA-contacting residues in the crystallographic NF-κB–DNA complex, but they are positioned near regions involved in DNA recognition or local structural stabilization13. Ser66 lies close to Glu60 and His64, which are associated with DNA recognition in the p50 subunit. Lys117 is located in the Rel Homology Domain and may contribute to the local electrostatic environment of the protein surface. Therefore, the observed contacts suggest that curcumin associates near a structurally relevant region of the p50 homodimer. However, because DNA was absent from the simulation, these observations cannot be interpreted as evidence that curcumin competes with DNA or disrupts DNA binding. MM-PBSA analysis provided a relative estimate of the energetic favorability of the p50 homodimer–curcumin complex. The average total binding free energy was − 15.75 kcal mol–1, with favorable van der Waals and electrostatic terms partially offset by the polar solvation contribution. This pattern indicates that nonbonded interactions, especially van der Waals contacts and electrostatic contributions, are the major stabilizing components in the predicted complex. Because entropic contributions were not included, the calculated value should be interpreted as a relative end-point free-energy estimate rather than an absolute binding affinity. The energetic pattern observed here is broadly consistent with previous computational studies in which hydrophobic, van der Waals, electrostatic, and hydrogen-bonding interactions contribute to small-molecule association with NF-κB-related targets or NF-κB pathway proteins52. However, the present calculation does not demonstrate that curcumin lowers NF-κB DNA-binding affinity. Instead, it identifies residues and energetic terms that may contribute to the stability of the predicted protein–ligand complex under the modeled conditions. PCA and FEL analyses were used to examine whether curcumin binding is associated with changes in dominant collective motions within the 100 ns trajectory. The curcumin-bound system sampled a comparatively restricted region of conformational space relative to the apo model, and the FEL showed more localized low-energy basins. These observations suggest stabilization of a subset of conformational states in the curcumin-bound protein-only model. Similar PCA- and conformational landscape-based analyses have been used to examine ligand- or modification-associated changes in biomolecular motions and conformational state distributions53,54. However, the FEL results should not be interpreted as evidence that curcumin locks NF-κB into an inactive state or directly reduces DNA-binding activity. The combined results from the docking, MD, MM-PBSA, and PCA/FEL results suggest that curcumin can maintain a predicted association near the p50 DNA-recognition region and is associated with localized changes in structural dynamics. These results support a computational hypothesis for local curcumin-associated stabilization near a structurally relevant region of the p50 homodimer. They do not establish that curcumin directly inhibits NF-κB, blocks DNA binding, or alters transcriptional activity. PCA was used to examine whether curcumin binding is associated with changes in the dominant collective motions of NF-κB within the simulated timescale. Projections of the trajectories along PC1 and PC2 showed that the curcumin-bound system sampled a comparatively restricted conformational space relative to the apo protein. The dominant motions were mainly associated with loop and interfacial regions close to the predicted curcumin-binding region, consistent with the residue-level fluctuations observed in the RMSF analysis. The FEL constructed from PC1 and PC2 further showed more localized low-energy basins for the curcumin-bound trajectory, suggesting stabilization of a subset of conformational states under the modeled conditions as shown in Fig. 7. Similar PCA- and conformational landscape-based analyses have been used to examine ligand- or modification-associated changes in biomolecular motions and conformational state distributions55,56. These results support the interpretation that curcumin binding is associated with altered local conformational sampling near the predicted binding region; however, they should not be interpreted as direct evidence of altered DNA-binding activity in the absence of explicit DNA-bound simulations. The NF-κB-curcumin complex’s FEL from 10 ns to 100 ns showing energy minima. The present work is based on a 100 ns MD trajectory for the apo protein and the curcumin-bound complex. Although longer simulations and independent replicas would provide more exhaustive conformational sampling, the objective of the present simulation was to examine the short-timescale stability of the predicted curcumin-bound state and the persistence of local protein–ligand interactions under the modeled conditions. The RMSD profile showed an initial adjustment followed by a comparatively stable conformational regime, while the radius of gyration remained within a narrow range of approximately 2.50–2.60 nm for the curcumin-bound complex. The SASA profile and hydrogen-bond/contact analyses also showed persistence of the ligand-associated interaction pattern during the trajectory. In addition, block averaging of the RMSD profile and the running average of the binding free energy indicated that the major structural and energetic trends were maintained over the analyzed timescale. Therefore, the 100 ns simulation is considered sufficient for assessing local binding stability and residue-level dynamical trends within this computational model. However, it does not provide exhaustive sampling of the full NF-κB conformational landscape, and the absence of independent replicate simulations remains a limitation. The simulations were performed using the NF-κB p50 homodimer model derived from PDB ID 1SVC after removal of DNA; therefore, the electrostatic and conformational effects of DNA binding were not explicitly represented. Also, curcumin binding near the DNA-recognition region is proposed from computational modelling and has not been experimentally validated in this work. Consequently, the results are interpreted as a computational hypothesis for curcumin-associated local conformational stabilization rather than definitive evidence of altered NF-κB DNA-binding activity. The present findings are consistent with the broader literature describing curcumin as a modulator of NF-κB-associated signaling. However, the current results are computational and are based on a protein-only p50 homodimer model. Further studies using explicit DNA-bound NF-κB models, longer and replicated simulations, entropy-inclusive free-energy calculations, and experimental binding or functional assays will be required before conclusions can be drawn regarding direct effects on NF-κB DNA recognition or transcriptional activity. Conclusion In this study, we investigated the interaction of curcumin with the NF-κB p50 homodimer using molecular docking, 100 ns all-atom molecular dynamics simulations, and MM-PBSA free-energy analysis. Molecular docking identified a preferred binding pose within the Rel Homology Domain, proximal to the DNA-recognition region. Subsequent molecular dynamics simulations showed that curcumin binding is associated with localized conformational stabilization, reflected by reduced residue-level flexibility near the predicted binding region, decreased solvent-accessible surface area, and maintenance of a compact structural profile throughout the trajectory. Energy decomposition analysis indicates that the interaction is primarily driven by favorable van der Waals and electrostatic contributions, partially offset by polar solvation effects, resulting in a moderately favorable binding free energy. The absence of significant structural destabilization during simulation suggests that curcumin forms a dynamically stable complex with NF-κB p50 homodimer under the modeled conditions. Collectively, these findings provide a residue-level and energetic characterization of predicted curcumin binding near the NF-κB DNA-recognition region. While the present work offers mechanistic insight into the molecular determinants of this interaction, further investigations incorporating explicit DNA-bound NF-κB models, entropy estimation, longer and replicated simulation timescales, and experimental validation will be necessary to establish quantitative binding affinities and functional consequences. Author contributions S.S. conceived and designed the study, planned the computational strategy, and supervised the research. V.K. set up the computational systems, performed the molecular docking and molecular dynamics simulations, and analyzed the simulation data. Both authors wrote, critically reviewed, edited, and finalized the manuscript. All authors read and approved the final manuscript. Funding Open access funding provided by Manipal University Jaipur. This research received no specific grant or funding from any funding agency in the public, commercial, or not-for-profit sectors. Data availability Data generated in this study are available from the corresponding author upon request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement Data generated in this study are available from the corresponding author upon request.