Molecular docking, DFT calculation and molecular dynamics simulation of Triazine-based covalent organic framework as a nanocarrier for delivering anti-cancer drug

Abstract

Covalent organic frameworks (COFs) have attracted attention as novel nanocarriers for anticancer drug delivery due to their properties such as large surface area, tunable porosity, and customizable surface capabilities. In this study, the absorption of imatinib (IMA) as an anticancer drug on a triazine-based COF was investigated through computational methods. Docking simulation was employed to identify potential binding sites and estimate the binding energy of imatinib on COF. The mean binding energy was calculated to be -6.43 kcal/mole, which is predominantly attributed to the π–π stacking interaction. Then, the most favorable complexes with good binding energies were selected for density functional theory (DFT) calculations. The calculations were performed using the ONIOM method in the presence of molecular water effects using the PCM model. The absorption energies ranged from 21.4 to 27.60 kcal/mol for the studied complex. For a deeper analysis of the interactions, natural bond orbital (NBO) analysis and quantum theory of atoms in molecules (QTAIM) were used. In addition, molecular dynamics (MD) simulations were performed to evaluate the IMA -COF complexes in aqueous environments. Based on the results, vdW interactions played an important role in the stability of the drug molecule in both inside and outside of the COF cavity. Also, the diffusion of IMA within the COF was investigated by mean square displacement (MSD) analysis which showed the restricted mobility of IMA within the COF compared to the aqueous media. The results of this study confirm the ability of COFs as drug delivery systems and suggest the targeted design of these structures according to the drug structure.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-26133-7.

Introduction

Cancer, as one of the major medical challenges of our time, requires new therapeutic approaches to overcome the side effects associated with traditional treatment methods^1,2 Nanocarriers, a diverse family of nanoscale materials, are increasingly used to deliver therapeutic agents with improved precision and safety. Biocompatible carriers load drugs and shield them from chemical degradation and enzymatic inactivation, enabling controlled, localized release at the tumor site to sustain therapeutic levels while reducing systemic toxicity. Surface engineering (e.g., ligand conjugation, stealth coatings) promotes passive accumulation via the enhanced permeability and retention effect or active targeting, enhancing uptake by cancer cells. They can also enable theragnostic and stimuli-responsive release in response to tumor microenvironment cues, supporting combination therapies. The vast diversity of nanocarrier architectures, ranging from solid nanoparticles and liposomes to polymeric nanogels, dendrimers, and nanoscale membranes, reflects their adaptable design strategies to optimize loading capacity, release kinetics, and targeting capabilities^3–7. Recently, covalent organic frameworks (COFs), a new class of synthesized polymers, were introduced as suitable candidates for carrying drug molecules due to their structural flexibility, high stability, large contact surface area, and potential for functionalization⁸. These porous and crystalline materials were constructed from light elements (e.g., carbon, hydrogen, oxygen, and nitrogen) through the reaction of organic linkers and building blocks, forming porous networks with various topologies such as Square Lattice, Hexagonal, Honeycomb, and kagome. Furthermore, the precise control over the pore size and shape, as well as modifiable chemical nanostructures, makes them promising materials for applications in fields of catalysis, gas storage, and gas separation, and drug delivery^9,10.

Previous studies have shown that COFs have been successful in encapsulating drugs such as doxorubicin and paclitaxel and protecting them from premature degradation^11,12. Furthermore, COF-drug complexes have shown enhanced cytotoxicity against cancer cells with minimal effects on healthy tissues, indicating their potential to overcome multidrug resistance¹³. Given the growing number and diverse applications of COFs, using computational methods offers a powerful way to not only predict their drug encapsulation capability, but also to visually analyze the various kinds of interactions that form between COFs and drug molecules¹⁴.

Theoretical simulations play a significant role in modeling and understanding biological and engineering systems. These research methods provide valuable information for discovering various physical and chemical properties at both the molecular and atomic levels¹⁵. Computational methods can be classified into two categories: quantum mechanics (QM) and molecular mechanics (MM). QM methods consider the interaction of electrons within the atoms, providing more accurate results at the atomic level. However, it is usually not possible to consider entire large biological systems due to the computational burden¹⁶. In contrast to QM methods, MM methods enable the analysis of large and complex biological behaviors at the molecular level. Molecular dynamics (MD) simulation is one of the computational approaches based on the principle of MM. In this method, the trajectory files (MD outputs) are generated by solving Newton’s equation of motion for each particle. The macroscopic properties of a system can be studied by applying statistical mechanics to the system’s trajectory¹⁷.

Imatinib is a clinically established tyrosine kinase inhibitor widely used in the treatment of chronic myeloid leukemia (CML) and gastrointestinal stromal tumors (GIST). Despite its therapeutic success, imatinib therapy faces several limitations including substantial interpatient variability in oral absorption, extensive hepatic metabolism (primarily via CYP enzymes), systemic toxicity at therapeutic doses, and challenges in achieving sustained, tumor-targeted drug concentrations. These pharmacokinetic and delivery-related issues can compromise treatment efficacy and increase adverse effects. Therefore, investigating new carriers for imatinib delivery is justified as a strategy to enhance drug stability, provide controlled and localized release, and potentially reduce systemic toxicity while improving therapeutic outcomes^18–20. In our previous study²¹, a biocompatible COF based on triazine and vanillin was synthesized and investigated as a drug carrier for the IMA. Experimental results showed the excellent performance of this nanocarrier for loading IMA with an exceptional encapsulation efficiency of 96%. Understanding the nature of interactions between the drug molecules and the nanocarrier can provide valuable insights into the mechanism of drug adsorption. In this study, the interaction of the IMA with the synthesized COF was investigated from a molecular perspective. For this purpose, various computational methods at different levels were used. Firstly, the molecular docking study was performed to obtain the best configurations of the IMA on the COF surface. Following that, to conduct more in-depth interaction analysis, density functional theory (DFT) calculations, natural bonding orbital (NBO) calculations, and quantum theory of atoms in molecules (QTAIM) were performed. Furthermore, Molecular dynamics (MD) simulations were used to study the stability of the drug-COF complex using a ten-layer COF. Finally, the drug diffusion process was investigated through mean square displacement (MSD) analysis.

Methods

In this study, to perform both molecular dynamic simulations and DFT calculations, the best configuration of the drug on the COF was obtained through the molecular docking study. To analyze the interaction between IMA and the COF surface, docking was performed for a single COF layer (IMA @COF). Furthermore, molecular dynamics simulations were performed to investigate the dynamic behavior of a complex of 10-layer COF with IMA (IMA @10COF). The COF model was created as a multilayered structure, resembling a realistic nanoscale carrier with a periodic, porous framework.

The COF structure was built using GaussView 5.0 software²², while the molecular structure of IMA was retrieved from the DrugBank database. To ensure accurate modeling, the drug and COF monolayer structures were geometrically optimized using Gaussian 09 software²³ at the B3LYP level of theory with the 6-31G(d, p) basis set. Docking studies were performed with AutoDock4 software²⁴ through 200 independent runs using the Lamarckian genetic algorithm. Each run included up to 2,500,000 energy evaluations over 27,000 generations, ensuring comprehensive sampling of the configuration space. IMA’s geometry was optimized using the MMFF94 force field, and partial charges were assigned via the Gasteiger method. The COF structure was treated as a rigid receptor, with its atomic charges derived from prior quantum mechanical calculations. The dimensions of the grid box were adjusted to encompass the entire COF structure, with a size of (6 × 6 × 3) nm for the IMA@COF system. For IMA@10COF, a single grid box was employed to encompass the entire cavity, measuring (6 × 6 × 6) nm. Additionally, to explore potential binding sites without considering any specific regions, another box with dimensions of (4 × 4 × 5) nm was used to cover the interior cavity of the ten-layer COF. The identified docking positions were used as starting points for subsequent DFT studies and MD simulations.

To evaluate the interaction energies between IMA and the COF surface, DFT calculations were employed, incorporating solvent effects through the polarizable continuum model (PCM) for water as the solvent. For this purpose, five complexes with the best binding energies of the highest population were selected from the IMA@COF docking study. The calculations were performed using ONIOM implemented in Gaussian 09, in which the system was investigated at two computational levels (high and low level)²⁵. In this study, the high layer consisted of IMA and the COF moieties that interacted directly with the drug, was optimized using the M06-2X method and the 6-311G(2d, p) basis set. M06-2X was chosen as a proposed method to capture the effects of electron correlation in the medium range for non-covalent interactions^26,27. 6-311G(2d, p) as triple-ζ split-valence basis was selected to achieve a balance between computational cost and reliable description of electronic structure for the drug and COF interface region. This basic set includes additional polarization functions (two sets of d functions on heavy atoms and p on hydrogens) that improve the description of anisotropic electron density and noncovalent interactions compared with common double-ζ sets (e.g., 6-31G(d)), while remaining far less expensive than large quadruple-ζ or explicitly correlated approaches. Since the QM layer models bond-breaking/forming, charge redistribution, and short-range noncovalent interactions essential to the mechanism; 6-311G(2d, p) provides a level of accuracy appropriate for these purposes²⁸.The MM layer supplies structural and steric context rather than highly accurate QM layer. UFF is computationally inexpensive and stable for large MM regions, allowing to include sufficient structural context around the QM region while keeping geometry optimizations tractable²⁹.

The interaction energy (E_int) between IMA and COF was determined using the following equation:

Here, E_IMA@COF represents the total energy of the IMA@COF complex, while E_IMA and E_COF denote the energies of isolated IMA and COF, respectively. The basis set superposition error (BSSE) was included to correct for any overestimation of interaction energies due to basis set limitations³⁰.

In order to provide detailed information on the orbital interactions and electronic charge transfer between IMA and the COF framework, NBO analysis was performed using NBO 6.0 software³¹. Also, QTAIM calculations were performed at the same level of theory using Multiwfn software³². The QTAIM method allowed for the identification of interaction pathways and critical bonding points (BCPs). Parameters such as electron density (ρ), Laplacian (∇²ρ), and total energy density (H(r)) were calculated to categorize bonds as either covalent or non-covalent, providing a comprehensive analysis of the binding characteristics in the IMA@COF system.

This study used molecular dynamics (MD) simulations to investigate the stability of the drug on the COF as well as the dynamic diffusion of the drug into the COF cavity. GROMACS 2022 software³³ using the CHARMM36 force field^34,35 was used for MD simulations. Topology and parameter files for both the COF and IMA molecules were prepared via the SwissParam server³⁶, which is compatible with the CHARMM all-atom force field. The systems were solvated in triclinic boxes filled with SPC-modeled water molecules³⁷. For the IMA@10COF systems, simulation boxes with dimensions of 7 × 7 × 9 nm were filled with 14,611 water molecules. Periodic boundary conditions (PBC) were applied in all directions to reproduce the bulk behavior. Simulation conditions were maintained at 1 atm pressure and 300 K using a Parinello-Rahman barostat and a Nose-Hoover thermostat, respectively^38,39. The systems were energy-minimized using the steepest descent algorithm to remove steric clashes, followed by 10 ns equilibration phases under constant volume and temperature (NVT) and subsequently 10 ns under constant pressure and temperature (NPT). For reproducibility, energy-time series were examined to assess convergence and stability. The MD simulations were conducted for 300 ns with a time step of 2 femtoseconds. Short-range non-bonded interactions cut-offs were set at 1 nm, and long-range electrostatics were calculated using the particle lattice Ewald (PME) method⁴⁰. The LINCS algorithm⁴¹ was applied to constrain bond lengths. Visualization and analysis of the pathways were performed using VMD software⁴², while BioViva Discovery Studio software⁴³ was utilized to analyze interactions between IMA and the COF framework.

Results and discussion

Molecular Docking of IMA to monolayer COF

The Docking study was first performed to obtain the best configuration of the drug on COF surface. The populations and binding energies of the obtained clusters for the IMA@COF complex are shown in Fig. 1. As can be seen, the binding energy for IMA varied between − 4.5 and − 6.76 kcal/mol, and the most populated cluster had a binding energy of −6.42 kcal/mol. Figure 2 shows that the interaction of the drug with COF is through heterocyclic rings (pyridine, benzamide, and pyrimidine moieties) and in the π-π mode. Based on these docking results, hydrophobic interactions between the drug and COF play an important role in the stabilization of the IMA@COF complex and drug adsorption. Five complexes with favorable population and binding energy (Table 1), obtained from the docking study, were selected for conducting a more detailed study of the adsorption energy using the DFT method in aqueous medium.

Fig. 1.

Cluster analysis between IMA and COF in molecular docking study (red bar graph indicate the most populated cluster).

Fig. 2.

Interaction between the most favorable configurations of IMA@COF in molecular docking study.

Table 1.

Representative docked complexes selected for DFT study and their calculated interaction energies.

Complex	Docking analysis		DFT study
	Binding energy (Kcal/mol)	Population	Eint (kcal/mol)	BSSE (kcal/mol)
1	−6.16	9	−24.18	11.22
2	−6.08	9	−20.49	10.46
3	−5.31	9	−22.71	9.26
4	−6.07	9	−26.00	10.46
5	−6.07	5	−27.60	9.83

DFT calculations

Since the molecular docking study was performed in the gas phase without considering water effects, the selected complexes were geometrically optimized using the DFT method in an aqueous environment, and the interaction energies between the IMA and the COF surface were calculated based on the BSSE correction. The calculations were performed using the ONIOM method. The selected QM and MM regions for complex No.5 are shown in Fig. 3. The same approach was applied to all the selected complexes. The corrected interaction energies obtained for each complex are summarized in Table 1. The obtained interaction energies ranged from − 20 to 33 − 27 kcal/mol, indicating a strong interaction energy between the IMA and the COF surface in the aqueous environment. Furthermore, all DFT-optimized complexes exhibited π-π stacking interactions between the aromatic moieties of IMA and COF, with a vertical distance less than 0.4 nm (Fig. 3). The computed adsorption energy exhibits strong drug-host interactions, which is in agreement with the experimentally observed high loading of the drug²¹. This concordance supports a direct correlation between binding strength and loading efficiency in the COF system, reinforcing the potential of the framework for high-capacity. Additionally, the interaction energy between the COF and the drug was initially estimated by docking, yielding an approximate binding energy of −6.43 kcal/mol. However, subsequent quantum chemical calculations at the DFT level, with BSSE correction and explicit solvent effects, predict a substantially stronger interaction of −27.6 kcal/mol. The large increase in magnitude is consistent with correcting for BSSE, which artificially lowers the energy in non- BSSE-corrected docking, and with solvent stabilization that is not captured in the docking score. Therefore, the docking value (−6.43 kcal/mol) should be viewed as a qualitative, screen-summary estimate, while the DFT-BSSE-solvent value (−27.6 kcal/mol) provides a more reliable quantitative measure of the binding interaction under realistic conditions. Among the selected complexes, complex No.5 with the lowest interaction energy of −27.6 kcal/mol showed the best optimal configuration, and this complex was selected for deeper molecular investigation through NBO and QTAIM.

Fig. 3.

A: QM: MM regions for geometric optimization of IMA@COF using M062X/6-311G(2d, p):UFF, B: Important interactions of IMA on COF.

NBO analysis

A deeper investigation of the drug interaction with COF was performed using natural bond orbital (NBO) analysis. This method provides a localized Lewis representation of the electronic structure, allowing the investigation of electron density distribution, which offers detailed insights into donor-acceptor interactions. The stabilization energy obtained from the electron density distribution between the occupied donor orbitals NBO (Lewis type) and the unoccupied acceptor orbitals NBO (non-Lewis type) is quantified using the following equation:

where (q_i) is the donor orbital occupancy, (ε_j) and (ε_i) are acceptor and donor orbital energies, and (F_ij) is the off-diagonal NBO Fock matrix element⁴⁴.

The calculated E⁽²⁾ values for the different donor-acceptor interactions in this complex are summarized in Table S1. The stabilization energy E⁽²⁾ of the IMA@COF complex was calculated to be 27.4 kcal/mol, indicating strong electronic interactions between the two components. Significant charge transfer was observed from the lone pair (LP) of N₅ in COF to Bonding Descriptor(BD) (N₉₈-H₉₉) orbital in IMA, with an E⁽²⁾ value of 5.01 kcal/mol. In addition to that, LP(2) N₅ in COF contributed to the same acceptor orbital in IMA with an E⁽²⁾ of 2.76 kcal/mol, indicating a secondary stabilization effect. Another key interaction involved LP O₁₉ in COF, which transferred charge to the BD(C₁₀₇-H₁₃₁) orbital in IMA, contributing 1.23 kcal/mol to the overall stabilization. Moreover, the BD(2) C₄-C₁₁ orbital in COF donated electron density to the BD(C₁₀₃-C₁₀₄) orbital in IMA, with an associated E⁽²⁾ value of 0.42 kcal/mol. Conversely, significant charge transfer from IMA to COF was also observed, including LP(1) N₉₈ in IMA donating to the BD(O₁₉-C₂₁) orbital in COF, with E(2) of 0.10 kcal/mol, and LP(1) N₉₈ in IMA interacting with the BD(O₁₉-C₄₃) orbital in COF, with E⁽²⁾ of 0.13 kcal/mol. Additional interactions included the BD(2) C₉₁-C₉₂ orbital in IMA donating electron density to the BD(C₂₁-C₂₃) orbital in COF, with E(2) of 0.30 kcal/mol, and BD(1) N₉₈-H₉₉ in IMA transferring charge to the RY(1) N5 orbital in COF, with E⁽²⁾ of 0.13 kcal/mol. These findings highlight the cooperative nature of charge transfer in the IMA@COF complex, which plays an important role in stabilizing the interaction. This confirms the satisfactory affinity of IMA for adsorption onto COF and highlights the key electronic factors underlying the strong interaction between IMA and the COF framework.

QTAIM analysis

QTAIM was used to characterize the non-covalent interactions between IMA and COF by examining electron-density features at bond critical points (BCPs)^45,46. Each BCP was characterized by the electron density ρ, its Laplacian ∇²ρ, and the local total energy density H(r) parameters, the value of which determine the nature of interactions (Table 2). In IMA@COF complex, all observed BCPs exhibit ρ < 0.1 a.u., ∇²ρ > 0, and H(r) > 0, indicating predominantly non-covalent interactions with a dispersion-dominated character. Notably, a BCP corresponding to a hydrogen-bond–like contact between the IMA hydroxyl group and a COF nitrogen atom was identified. The hydrogen-bond strength, estimated with the Espinosa equation (E_HB ≈ 1/2 V(r)), was calculated as ≈ − 5.4 kcal/mol. The total number of BCPs for the IMA@COF complex is 20, with a cumulative ρ ≈ 0.13 a.u., suggesting a robust interaction network driven mainly by π–π stacking between the heterocyclic rings of IMA (pyridine and benzylamine moieties) and the COF framework (Fig. 4). Topological parameters of the IMA@COF complex obtained from QTAIM analysis was reported in Table S2.

Table 2.

QTAIM parameters and their interaction at BCPs.

Interaction type	Representative BCP descriptor (typical ranges)
Hydrogen-bond–like	ρ < 0.1 a.u.; ∇²ρ > 0; H(r) > 0; EHB ≈ − 5 to − 6 kcal/mol
π–π/π-donor–acceptor	ρ around 0.05–0.1 a.u.; ∇²ρ > 0; H(r) > 0; clear π-π channel
vdW-dominated contacts	ρ < 0.05 a.u.; ∇²ρ > 0; H(r) > 0; dispersed network

Fig. 4.

Critical bonding path (yellow) between IMA and COF.

Molecular docking of IMA in a ten-layer COF

To investigate whether the drug is more stable on the surface (outside) or inside the COF cavity, docking studies were performed in two different approaches. In the first approach, the entire COF cavity was defined as the grid box, whereas in the second, the interior of the COF cavity was considered (Fig. 5). For this purpose, a ten-layer COF structure with an interlayer distance of 5.64 Å was constructed using VMD software. Based on the results, when the entire COF cavity was considered as the grid box, 40 clusters were obtained, in which the drug molecule exhibited the lowest binding energy on the outer surface of the COF. However, in the subsequent clusters, the drug molecule was located within the COF cavity. It is noteworthy that the difference in binding energy for the drug outside and inside the cavity is only about 1 kcal/mol. This indicates that the drug showed good stability in both positions. When the entire cavity was considered as the grid box, in all the obtained clusters, the drug remained in the cavity, and the configuration with the highest cluster population (15 docking positions) and binding energy of −12.36 kcal/mol was the best obtained complex. The results confirmed that the drug can be adsorbed and stabilized on both the surface (outside) and inside of the COF cavity.

Fig. 5.

Grid box and orientation of IMA on the ten-layer COF in the docking study (a) the entire cavity and (b) inside the COF cavity as the grid box.

Molecular dynamics simulation of ten-layer COF and IMA@10COF complex

In the next steps, the stability of the ten-layer COF and IMA@10COF structures in the aqueous environment was investigated through molecular dynamics simulation. The most stable confirmations of the drug molecule, inside (IMA@10COF-I) and outside (IMA@10COF-O) of the COF cavity, obtained from the molecular docking study, were used as input for the molecular dynamic simulations. The molecular dynamics simulations were performed for 300 ns. The stability and equilibrium of the systems were evaluated by calculating the root mean square deviation (RMSD) of all atoms from their initial positions. As shown in Fig. 6, the RMSD curve stabilized after about 20 ns and all systems reached equilibrium.

Fig. 6.

RMSD curves of the 10 layer COF, IMA@10COF-I, and IMA@10COF-O after 300 ns.

In all three systems, it was found that during the 300 ns simulation, the COF maintained its structural integrity, and no interlayer separation was observed. During the simulation, the distance between layers of COF decreased from 5.6 Å to 4.3 Å, and after reaching the equilibrium, the average distance between layers remained constant (Table 3). This can be attributed to the broad and flat π-electron system of the aromatic building blocks of the COF, which facilitates strong π-π stacking interactions. Figure 7 shows the initial and final images.

Table 3.

Average distance (nm) between the center of mass of each layer in the ten-layer COF and IMA@10COF systems.

Layers	Distance in ten-layer COF (nm)	Distance in IMA@10COF –I (nm)	Distance in IMA@10COF –O (nm)
1 and 2	0.41981	0.448235	0.41981
2 and 3	0.41380	0.43973	0.41380
3 and 4	0.41548	0.435939	0.41548
4 and 5	0.44061	0.434872	0.44061
5 and 6	0.42806	0.430476	0.42806
6 and 7	0.42680	0.436416	0.42680
7 and 8	0.42943	0.439728	0.42943
8 and 9	0.43741	0.433011	0.43741
9 and 10	0.44066	0.447742	0.44066

Fig. 7.

Initial (a) and final (b) structures of the 10-layer COF after 300 ns of molecular dynamics simulation.

In the simulation of the IMA@10COF-I system, it was observed that the drug remained stable inside the cavity of COF through the entire simulation duration. In Fig. 8, the initial and final positions of the simulated IMA@10COF-I system are shown. Also, as for the IMA@10COF-O system, it was observed that the drug entered the cavity during the simulation and remained stable in the intra-cavity region. This observation confirms the spontaneous adsorption and diffusion of IMA into the COF structure. Figure 9 shows the different configurations of the IMA@10COF-O complex during the molecular dynamic simulation.

Fig. 8.

Initial (a, b) and final (c, d) structures of the IMA@10COF-I system after 300 ns molecular dynamics simulation.

Fig. 9.

Initial (a, b) and final (c, d) structures of the IMA@10COF-O system after 300 ns molecular dynamics simulation.

To determine the drug adsorption behavior into the system, interaction analysis was performed.

The results showed that in both IMA@10COF systems, the main contribution was from vdW forces, while the role of electrostatic interactions was minimal. The vdW interaction energy showed significant fluctuations, while the electrostatic component was relatively smaller and more stable than the vdW interaction. These findings indicate that vdW forces are the dominant factor in driving the interaction between IMA and the COF cavity. Figure 10 shows the vdW and electrostatic interactions of the IMA@10COF systems, confirming the importance of vdW interactions in the stability of the complex.

Fig. 10.

VdW and electrostatic interactions involved in the IMA@10COF systems.

To investigate which part of the IMA structure contributes most to vdW bonds, the number of contacts (NC) between IMA and the COF (at a threshold distance of less than 0.6 nm) was monitored over time. The obtained results showed that the benzamide moiety exhibited a higher NCs, which might be due to the formation of π-π interactions with the COF (Fig. 11). Hydrogen bonding analysis also revealed a limited contribution of H-bond to the overall stability of the drug on COF. IMA formed only one hydrogen bond with the COF cavity during the simulation through the nitrogen of the benzamide moiety of IMA as a hydrogen bond donor and the oxygen of the vanillin as a hydrogen bond acceptor.

Fig. 11.

Number of contacts between different IMA segments with 10-layer COF.

Drug diffusion in water and ten-layer COF nanopores

The diffusion of drug molecules into the COF was investigated according to previous study⁴⁷. It has been claimed that diffusion is significantly affected by the geometric constraints of the pore structure, which can limit molecular mobility and reduce the diffusion rate compared to bulk water. For this study, the diffusion coefficient (D) of an IMA molecule in bulk water was first obtained using the three-dimensional Einstein Equation (n = 3), where r² represents the MSD of a drug as a function of time.

Since the determination of D for a molecule in water requires long time-scale simulations to achieve convergence, multiple independent MD runs were performed. For this purpose, after 300 ns MD simulation 20 independent 3 ns runs with different random seeds was performed and the obtained MSDs were averaged over time. The results showed that the diffusion coefficient (D) for IMA in water was 6.88 × 10⁻¹⁰ m²/s (Fig. 12). The linear MSD profile (R² = 0.9956) confirmed its free translational motion, consistent with its small molecular size and hydrophilic nature.

Fig. 12.

Average MSD curves of an IMA molecule (a) in water and (b) in the presence of COF.

When placed on the outer surface, IMA rapidly penetrated the COF pore within the first 50 ns and remained encapsulated throughout the simulation. The RMSD stabilized at around 0.25 nm after 10 ns, indicating the formation of a stable complex. Similarly, when IMA was initially positioned inside the pore, the RMSD stabilized at approximately 0.25 nm, and the drug showed no tendency to exit, suggesting strong retention within the COF structure.

Next, a similar approach was taken to calculate D in the presence of COF. The COF structure consisted of ten stacked layers, each separated by a distance of 5.64 Å, which ensured a sufficient cavity length for the analysis of limited drug diffusion. The drug molecule remained in the cavity during the 300 ns simulation, and no signs of escape were observed. After that, 20 independent 3 ns simulations were performed with different random seeds to refine the analysis. The one-dimensional Einstein equation (n = 1) was used to calculate the diffusion coefficient along the pore axis. For IMA, the diffusion coefficient in the COF pore was obtained 4.0 × 10⁻¹¹ m²/s. This reduction in the diffusion rate compared to water indicates a significant mobility limitation imposed by the nanopore (Fig. 12). The lower mobility of IMA within the COF pore is consistent with the higher number of contacts (NC) with the pore wall, as discussed earlier. This reduction in the diffusion coefficient of IMA through the COF pore is in good agreement earlier similar reports^47–49 and highlights the effectiveness of COF as a nanoparticle carrier for drug delivery. The ability of COF to restrict molecular mobility while maintaining strong binding interactions suggests its potential as a controlled release platform, especially for anticancer drugs such as IMA.

The demonstrated compatibility between multiple computational methods and experimental observables not only reinforces the reliability of COFs as versatile platforms for drug delivery but also highlights their potential to guide rational design and optimization of loading, stability, and release profiles.

Conclusions

This study used advanced computational techniques to investigate the potential of a triazine-based COF for the delivery of IMA at molecular level. Molecular docking results showed satisfactory binding affinity between IMA and the COF. DFT calculations, taking into account the electronic structure, provided a more accurate and correct interaction energy between the drug and COF. VdW interactions, especially π-π stacking between the aromatic moieties of IMA and the COF, were the most important interactions of the drug with COF. The randomization of the electron density and its distribution using NBO and AIM analysis provided strong and accurate electronic insights into the non-bonding interactions of the drug and highlighted the role of vdW in the drug-COF interactions. The stability of IMA on COF was investigated through molecular dynamics simulations of IMA@COF complexes in aqueous medium, which were investigated at different drug positions on COF. The results showed that the drug remained stable inside and outside of the COF, and vdW interactions played an important role in the stability of the complex. The molecular dynamics simulations also provided insights into the diffusion dynamics of IMA within the COF framework. MSD analysis revealed a decrease in the diffusion coefficient of IMA within the COF cavities compared to its behavior in water. The obtained results strongly confirmed the ability of COF as an efficient drug carrier and also showed that the synthesized COF can stabilize the drug on its surface through hydrophobic interactions. The dominant vdW and π–π interactions observed between IMA and the COF suggest that tuning the aromaticity, planarity, and stacking propensity of the COF backbone could enhance loading efficiency and stabilization for similar hydrophobic or aromatic drugs. This can be used in selecting drugs for this system or even modifying the structure of COF for use with hydrophilic drugs. The workflow combining docking, DFT, NBO/AIM insights, and MD can be extended to other drug-COF systems, and moreover, addressing aspects such as force-field validation, the dimensions of the simulation box, and the molecular behavior governing drug release would enhance the comprehensiveness and accuracy of investigations on these systems.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

M.O and M.S Software, Visualization, Writing—Original draft preparation, Formal analysis, and Investigation; G.K. Validation, Resources, and Writing—Review and Editing; P.A. Supervision, Project administration, Conceptualization, Validation, Resources, and Writing—Review and Editing. All authors reviewed the manuscript.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.