In silico discovery of CETP inhibitors from ZINC15 compounds using docking, ADMET filtering, and molecular dynamics simulations Authors: Kimia Mohammadi, Mojgan Nejabat, Forough Mirabi, Amirhossein Sahebkar, Farzin Hadizadeh Published: 2026 ABSTRACT Background and purpose:Cholesteryl ester transfer protein is a crucial regulator of lipid metabolism and is considered a promising target for treating atherosclerotic cardiovascular disease. Given the clinical limitations and adverse effects of previous cholesteryl ester transfer protein inhibitors, such as torcetrapib, this study aimed to identify safer, more effective alternatives through computational drug discovery.Experimental approach:A diverse set of natural compounds from the ZINC15 database was screened using a multi-step in silico workflow comprising drug-likeness assessment, ADMET profiling, molecular docking, and molecular dynamics simulations. This integrated process narrowed down a large compound library to a refined selection of candidates with favorable safety and pharmacokinetic properties.Findings/Results:Calebin, a bioactive compound derived from turmeric, emerged as the most promising cholesteryl ester transfer protein inhibitor. It demonstrated binding interactions with cholesteryl ester transfer protein (-8.4 Kcal/mol) that were comparable to curcumin (-8.5 kcal/mol) and lower than the reference x-ray inhibitor torcetrapib (-9.4 kcal/mol), along with similar molecular dynamic complex stability (RMSD of 2.515 ± 0.282 Å in comparison to torcetrapib 2.502 ± 0.332 Å).Conclusion and implications:These findings highlight the value of computational drug discovery for identifying natural inhibitors of cholesteryl ester transfer protein and provide a strong foundation for future experimental validation. Calebin A demonstrated a promising binding affinity and interaction stability with cholesteryl ester transfer protein, but proof-of-concept studies are warranted to confirm this potential activity. INTRODUCTION Cardiovascular diseases (CVDs) remain the leading cause of death worldwide, according to 2021 data from the World Health Organization (WHO). These conditions account for over one-third of all global deaths, with more than 17.9 million lives lost in 2019 alone. Heart attacks and strokes are the most prevalent and lethal manifestations, responsible for approximately 85% of all CVD-related fatalities(1). The rising burden of CVDs is closely associated with modern lifestyle factors and a range of behavioral and environmental risks, including smoking, poor diet, obesity, and physical inactivity. These challenges are particularly pressing in low- and middle-income countries, where access to preventive healthcare is often limited. At the biological level, CVDs frequently begin with arterial damage caused by elevated blood lipid levels. This process results in plaque buildup that narrows or obstructs blood vessels, ultimately precipitating serious events such as heart attacks and strokes(23). Identifying low-density lipoprotein cholesterol (LDL-C) as a primary contributor to plaque formation led to the widespread adoption of LDL-lowering agents, most notably statins, as the mainstay of treatment. While statins are the frontline of pharmacotherapy for atherosclerotic CVD owing to their LDL-lowering and pleiotropic actions(456789), a considerable residual risk remains after statin therapy that needs to be addressed(1011). Hence, newer agents need to be developed for addition on top of statins to achieve maximal CVD risk reduction(121314). Among the various agents that have been developed or are in the pipeline, natural products and phytochemicals have emerged as potential candidates owing to their natural origin, ethnopharmacological background, and their capacity to interact with various targets involved in lipoprotein metabolism(151617). Lipoprotein cholesterol concentrations are partly regulated by the lipid transfer protein family, with cholesteryl ester transfer protein (CETP) playing a central role. CETP is a 476-amino acid, hydrophobic plasma glycoprotein that facilitates the bidirectional transfer of cholesteryl esters (CE) and triglycerides (TG) between various lipoprotein particles(181920). Primarily synthesized in the liver, CETP activity increases in response to elevated cholesterol levels and binds to HDL-C in circulation. Through its exchange function, CETP promotes the efflux of CE from HDL in exchange for TG from apoB-containing lipoproteins, such as LDL-C and very low- density lipoprotein-cholesterol (VLDL-C). This transfer yields TG-rich, CE-depleted HDL-C particles that are rapidly catabolized. In addition to its peripheral effects, CETP activity has also been linked to cholesterol regulation in the central nervous system.(2122). Given its role in lipid modulation, pharmacological inhibition of CETP has been explored as a therapeutic strategy to increase HDL-C and influence LDL-C concentrations (Fig. 1). CETP mechanism of action and the effect of inhibitors on HDL and LDL composition. Blue and orange dots represent CE and TG, which exchange between HDL and LDL by CETP, resulting in an increase in CE and a decrease in TG in LDL while decreasing CE and increasing TG in HDL. CETP, Cholesteryl ester transfer protein; CE, cholesteryl ester; TG, triglycerides; HDL, high-density lipoprotein; LDL, low-density lipoprotein. Five CETP inhibitors, including torcetrapib, dalcetrapib, evacetrapib, anacetrapib, and the most recent agent, obicetrapib, have undergone clinical evaluation for cardiovascular risk reduction. Although torcetrapib and dalcetrapib (first-generation agents) produced significant increases in HDL-C (72.1% and 40%, respectively), it was withdrawn due to off-target effects that increased cardiovascular events and mortality, while dalcetrapib failed to demonstrate efficacy in outcome trials(2324). These outcomes are thought to result from differences in molecular binding and CETP inhibition mechanisms(25). Second-generation inhibitors, evacetrapib and anacetrapib, demonstrated dual efficacy in lowering LDL-C and raising HDL-C; however, neither showed a significant impact on all-cause mortality or major adverse cardiovascular events, leading to their discontinuation(24). Moreover, the prolonged adipose tissue accumulation of anacetrapib was another concern, prohibiting its further development(26). The emergence of obicetrapib has renewed interest in CETP inhibition. Early and ongoing trials suggest that obicetrapib effectively improves lipid profiles as a monotherapy and in combination with statins. Upcoming results are expected to clarify their clinical utility in CVD management(272829). Computational methods have become integral to modern drug discovery and development, offering cost-effective and efficient alternatives to traditional high-throughput screening of extensive compound libraries. These in silico techniques conserve time and resources by prioritizing potential drug candidates before experimental validation(3031). Among these approaches, structure-based drug design (SBDD) is particularly influential. SBDD utilizes detailed structural information of target biomolecules to identify and optimize therapeutic compounds. Virtual screening, a core component of SBDD, computationally evaluates large chemical libraries to predict ligand-receptor interactions. This process usually starts with molecular docking, which helps predict how strongly molecules bind together. Next, molecular dynamics (MD) simulations are used to see how molecules and the protein targets move and change shape. This progress allows researchers to explore new possibilities for finding potential drug candidates(3233). A crucial element in SBDD is comprehensive ligand libraries such as the ZINC database, which provides ready-to-dock, three-dimensional molecular structures of commercially available compounds. As a freely accessible and widely utilized platform, ZINC supports virtual screening, pharmacophore modeling, and AI-driven prioritization, making it a foundational tool in modern cheminformatics and computational drug discovery(3435). Hence, we developed our ZINC natural product library using the ZINC15 database, applying the ’standard ok’ filter (which includes substances with standard reactivity, encompassing PAINS and ZINC15 "clean" filters) and the ’in man’ filter (which includes substances that have been used in humans). Human placental CETP 3D-structure with PDB ID of 4EWS was obtained from the RCSB depository and resolution of 2.59 Å, and molecular screening was performed by the Schrodinger suite (www.schrodinger.com). MD simulation was performed for the most compatible compound (calebin) to understand the interactions better. Calebin A is used as a supplementary product with beneficial metabolic effects. It has structural similarities with curcumin, which has established pharmacological effects(3637383940414243). Pharmacokinetic and drug-likeness properties of selected compounds were also evaluated. MATERIALS AND METHODS Database preparation The ZINC15 database (https://zinc15.docking.org/substances/subsets/) serves as a widely recognized and freely accessible resource for virtual screening and structure-based drug discovery. In this study, we undertook a systematic filtration of the database to identify suitable compounds based on two specific criteria. We applied the "in-man" filter, which includes substances that have undergone testing in human subjects, and the "standard ok" filter, which selects compounds characterized by standard reactivity and that have successfully passed the PAINS and ZINC15 "clean" filters. To evaluate the drug-likeness of the selected compounds, we employed Lipinski’s rule of five via the ADMETLab online server (https://admet.scbdd.com/). This rule stipulates that an appropriate candidate should satisfy the following criteria: a molecular weight of less than 500 Da, fewer than five hydrogen-bond donors, fewer than ten hydrogen-bond acceptors, and a calculated octanol-water partition coefficient (log P) of less than 5. Furthermore, we assessed critical safety parameters, including mutagenicity, hepatotoxicity, and the non-inhibition of the human ether-a-go-go-related gene (hERG) channel, utilizing ADMETLab2 (https://admetmesh.scbdd.com/). Lastly, to ensure the suitability of the selected compound library for computational screening, we excluded P-glycoprotein (P-gp) substrates. Receptor identification The crystal structure of human CETP (PDB ID: 4EWS; in complex with inhibitors; resolution 2.59 Å) has been obtained from the Protein Data Bank (https://www.rcsb.org/). Molecular docking study The CETP structure was prepared before the docking study by removing water molecules and inhibitors and adding polar hydrogens to the PDB file using user-friendly Schrodinger Suites, Maestro (version 2021-3). The protein was processed, prepared, and minimized with the default force field (OPLS4) using the ’protein preparation wizard’ module under the default settings and pH 7 ± 0.5, making it ready for the docking study. Torcetrapib is the X-ray ligand in the 4EWS interacting with SER 230 and HIS 232. The grid box location and dimensions (66 Å × 66 Å × 66 Å) are centered on the Torcetrapib bond within the protein pocket. To ensure the robustness of our research, the ligand was re-docked in the location to validate the docking protocol(2144). Due to structural variation, 3D structures of the selected 107 ligands (out of 1,072 candidate molecules from the Zinc15 database) were prepared using Maestro’s ’Ligprep’ module at pH 7 ± 2, ensuring the protonation was acceptable. Torcetrapib was selected as the reference compound because it was the only CETP inhibitor for which a 3D crystal structure resolved by X-ray crystallography was accessible. The 2D structure of torcetrapib was drawn in ChemDraw Ultra 21.0.0 software (www.cambridgesoft.com) and prepared as mentioned earlier. The 3D structures were generated from 2D structures in ChemBio3D (www.cambridgesoft.com). Glide docking was performed on the ligand library and reference compound in standard precision (SP) mode, 10 poses per ligand, and flexible ligand sampling. The results were analyzed based on docking scores, hydrogen bonds, and hydrophobic interactions within the binding pocket. During docking libraries with Maestro, the SP mode is used as it operates within an acceptable time frame, whereas the XP (extra-precision) mode is more time-consuming. Quantum mechanics/molecular mechanics calculations Quantum mechanics (QM) is one of the pioneering methods to obtain the electronic level properties of molecules. The macromolecules like proteins/enzymes and deoxyribonucleic acid (DNA) are large in size and also contain more than a thousand atoms in a molecule. Hence, to obtain accurate electronic level information, treating these macromolecules quantum mechanically is quite important. However, it is very difficult to solve Schrodinger’s equation for a large system. Therefore, the QM/molecular mechanics (MM) analysis of proteins has been interrogated into two different approaches; one is the major QM region where the ligand interacts with the proximal active site amino acids of the protein. Relatively, the QM region of calebin/CETP includes Thr 27 and Arg 201. The remaining amino acids present in the complex were treated as the MM region. Therefore, the number of atoms present in the QM region is treated by the default DFT-B3LYP implemented in the Jagular/Qsite module of Schrödinger 2021. The remaining part of the protein is considered the MM region, which excludes the QM region, and this MM region could be treated with OPLS-2005. MD simulation To better understand the dynamic behavior of molecules and the flexibility and stability of protein-ligand complexes, an MD simulation was performed under physiological conditions (pH 7 and pressure of 1 atm and NaCl concentration of 0.15 M) using Schrödinger Suite, Maestro, and Desmond (version 2021-3). The simulation system was personalized to set up the environment with Transferable Itermolecular Potential with 3 Points (TIP3P) water and neutralized with the addition of 0.15 M NaCl to resemble the physiological situation. Prepared and pre-docked selected protein-ligand complexes undergo a 100 ns simulation at 300 K, 1 atm, after equilibration, recording an interval of 100 ps(45). Torcetrapib was simulated primarily inside the active site of the prepared enzyme under the mentioned conditions to validate the method(4647). Root mean square fluctuation (RMSF), root mean square deviations (RMSD), and the radius of gyration (Rg) were obtained for further analysis. RESULTS Database screening result This study extracted a subset of natural compounds from the ZINC15 database using the "standard ok" and "in man" filters, resulting in 1,072 candidate molecules. To evaluate their drug-likeness, Lipinski’s rule of five (no more than 5 hydrogen bond donors, no more than 10 hydrogen bond acceptors, molecular weight less than 500, logP of less than 5)(48) was applied through the ADMETLab server, ensuring that the selected compounds met essential physicochemical criteria, including a molecular weight lower than 500 Da, lipophilicity with a logP of less than 5, and acceptable limits for hydrogen bond donors and acceptors. The resulting molecules were then screened with ADMETLab2, focusing on critical safety and pharmacokinetic parameters such as mutagenicity, hepatotoxicity, and hERG channel inhibition (Table 1). Compounds identified as non-substrates of P-gp were selected for subsequent analysis(34). After this multi-criteria filtering process, a refined set of 107 molecules with favorable drug-like and safety profiles was identified. Table 1Predicted pharmacokinetic properties of the top 3 compounds through the ADMETLab server.CompoundMW1Log P2H-bond acceptor3H-bond donor4Surface area5Caco2 permeability6AMES toxicity7hERG inhibitor8Hepato- toxicity Curcumin 368.1302.5966293.060-4.852NoNoYes Calebin 384.1202.49372102.290-4.834YesYesYes Torcetrapib 598.1906.7195042.010-4.772YesYesYesMW, Molecular weight; 1, less than 500 acceptable; 2, less than 5 acceptable; 3, should be no more than 10; 4, should not be more than 5; 5, less than 140 Å3 intestinal good absorption; 6, more than 8 × 10-6 cm/s good intestinal absorption; 7, "Yes" (mutagenic); 8, "Yes" (potential cardiotoxic); hERG, human ether-à-go-go-related gene. Predicted pharmacokinetic properties of the top 3 compounds through the ADMETLab server. MW, Molecular weight; 1, less than 500 acceptable; 2, less than 5 acceptable; 3, should be no more than 10; 4, should not be more than 5; 5, less than 140 Å3 intestinal good absorption; 6, more than 8 × 10-6 cm/s good intestinal absorption; 7, "Yes" (mutagenic); 8, "Yes" (potential cardiotoxic); hERG, human ether-à-go-go-related gene. While this approach ensured that only compounds with favorable ADMET profiles were advanced, we acknowledge that the use of strict criteria in a single stage may have excluded ligands with moderate docking scores or alternative beneficial properties. This represents a limitation of the current study, as multi-stage screening pipelines that integrate docking and broader ADMET profiling could preserve greater chemical diversity and reduce the risk of discarding potentially valuable candidates. Ligand-protein docking method validation To gain a deeper understanding of the mechanism by which CETP is inhibited, we extracted the torcetrapib molecule from the CETP crystal structure and prepared various conformations without constraints using the Ligprep module of Schrödinger. These conformations were then docked freely into the enzyme’s binding pocket using Schrödinger 2021. The best docking result was obtained from the torcetrapib conformer-CETP complex, which exhibited the lowest docking score of -9.40 kcal/mol. This result closely resembled the conformation found in the crystal X-ray structure. Additionally, RMSD for this complex was less than 2 Å, demonstrating strong conformational similarity to the torcetrapib binding observed in the crystallography and validating the docking protocol used. Ligand library docking analysis The library consisted of 107 top molecules that passed Lipinski’s rule of five(48) and a toxicity study, along with other specified criteria. These molecules were docked into the CETP binding pocket using a validated protocol as described above, enabling us to obtain the inhibitory profile and analyze ligand-protein interactions. The binding energy of the compounds varied based on their affinity for the protein, ranging from -9.44 to -1.58 kcal/mol. The docking scores for the compounds were provided in Table 2. Table 2Screened structures of the ZINC database and their docking results, including obicetrapib, torcetrapib, and curcumin for comparison.No.Zinc codeSimplified molecular-input line-entry system (SMILES)Docking score (Kcal/mol) 1 (Obicetrapib) 209849193CC[C@@H]1C[C@@H](C2=C(N1C(=O)OCC)C=CC(=C2)C(F)(F)F)N(CC3=CC(=CC(=C3)C(F)(F)F)C(F)(F)F)C4=NC=C(C=N4)OCCCC(=O)O-9.44 2 (Torecetrapib) 8214714O=C(OC)N(CC1=CC(C(F)(F)F)=CC(C(F)(F)F)=C1)C2CC(CC)N(C(OCC)=O)C3=CC=C(C(F)(F)F)C=C32-9.4 3 (Curcumin) 899824O=C(CC(/C=C/C1=CC(OC)=C(O)C=C1)=O)/C=C/C2=CC=C(O)C(OC)=C2-8.5 4 (Calebin) 1908332O=C(OCC(/C=C/C1=CC(OC)=C(O)C=C1)=O)/C=C/C2=CC(OC)=C(O)C=C2-8.4 5 33805081C=C1C[C@](O[C@@]2([C@@H]1CC[C@H]2C)C/3)(O)C3=C(C)/C-7.55 6 1490O=C(NC1=C(C)C=CC=C1C)C2=CC=CC=C2C(O)=O-7.29 7 968326CCC(C=C1)=CN=C1CCOC2=CC=C(C[C@H]3C(NC(S3)=O)=O)C=C2-6.98 8 4529568O=C(C(C(F)(F)F)=CN1[C@H]2C[C@@H](O)[C@@H](CO)O2)NC1=O-6.59 9 14855479OC(C=C1)=CC=C1O[C@@H]2[C@H](O)[C@H](O)[C@@H](O)[C@@H](CO)O2-6.42 10 100830651C[C@H]1C=C[C@H](C2(C)C)[C@@H]2C1-6.24 11 100080480CC(C)[C@]12[C@@H](C2)[C@](C)(O)CC1-6.18 12 97700O=C(O)C1=CC=CC=C1C(OCC2=CC=CC=C2)=O-6.18 13 138219289NC1=NC=NC2=C1N=CN2[C@H]3[C@H](O)[C@H](O)[C@@H](CO[P@](O)(OP(O)(O)=O)=O)O3-6.16 14 895350O[C@@H]1O[C@H](CO)[C@H](O)[C@@H](O)[C@H]1O-5.94 15 3831420O[C@H]1[C@@H](CO)O[C@H](N2N=C(C(N)=O)N=C2)[C@@H]1O-5.87 16 1614204O=C(N[C@@H]1CCC(NC1=O)=O)CC2=CC=CC=C2-5.71 17 87525O=C1NC(N([C@H]2C[C@@H](O)[C@@H](CO)O2)C=C1Br)=O-5.54 18 1586772CCCC[C@@H](O)C1=CC=CC=C1-5.47 19 90750OC(C1=CC=CC=C1C(O)=O)=O-5.26 20 105469579O[C@@H]([C@@H]1O)CN2C1[C@H](O)CCC2-5.26 21 391885O=C(OC)C1=CC=CC=C1C(OC)=O-5.25 22 100028042C[C@]1(CC2)[C@H](OC(C)=O)C[C@@H]2C1(C)C-5.16 23 135822240O[C@@H]1[C@@H]2[C@@H]([C@@H](O)C(O2)=O)O[C@H]1O-5.09 24 1648289OC1=CC(CC)=CC=C1-5.05 25 1570209C12=CC=CC=CC1=CC=C2-5 26 2242633CC1=C[C@H]2[C@H](C2(C)C)CC1-4.97 27 41206201CC1=CC=C(C[C@@](C)(N)C(O)=O)C=C1-4.81 28 154800OC1=CC=C([C@@H](O)[C@H]([N+])C)C=C1-4.79 29 388179NC1=CC=CC(C(O)=O)=C1-4.78 30 3586101O=c1[nH]c(=O)n([C@@H]2C[C@H](O)[C@H](CO)O2)cc1I-4.69 31 202771077O[C@]1(CC2)[C@H](O)[C@H](O)[C@H](O)[C@@H]2N1-4.69 32 100193236C[C@@H]1CCC2=C1C[C@@H]3C(C)(C)[C@@]2(C)CC3-4.68 33 407096CC1=CCC(C(O)(C)C)CC1O-4.65 34 4027061N=C1C=CN2C(O[C@H]3[C@H]2O[C@H](CO)[C@H]3O)=N1-4.37 35 12153391O[C@H]1[C@H](O)[C@@H](O)[C@@H](CO)O1-4.12 36 391815O=C(OCC)C1=CN=CC=C1-3.99 37 38212689FC1=CNC(NC1=O)=O-3.89 38 100830723C[C@]12[C@H](O)C[C@H](CC2)C(C)(C)O1-3.88 39 158116CC(C)CC(N)=O-3.84 40 1451OC1=CC(OC)=CC(O)=C1-3.83 41 1673034CC1(C)C([C@@H]2C[C@@H]1CC2)=C-3.79 42 100017211O=C1OC(C)=CC(O)=C1C(C)=O-3.65 43 896546FC1=CNC(N=C1N)=O-3.58 44 3995575[O-]C([C@H]1[C@H](CC([O-])=O)[C@H](C(C)=C)C[N+]1)=O-3.37 45 3872953O=C(O)C1=CCNCC1-3.34 46 967521CC1(C)[C@]2(C)C(C[C@@H]1CC2)=O-3.07 47 403075CN1CCC[C@@H]1C(O)=O-2.83 48 1679741CCOC(C(C)=O)=O-2.55 49 18044615O=C(NC(N1)=O)NC1=O-1.58 Screened structures of the ZINC database and their docking results, including obicetrapib, torcetrapib, and curcumin for comparison. Given the reported effects of curcumin on dyslipidemia(495051), this research aimed to explore the lesser-known compound calebin A and assess its potential role in managing hyperlipidemia through modulation of CETP activity. Calebin A, a polyphenolic compound derived from Curcuma longa, has demonstrated broad therapeutic potential in areas such as cancer, inflammation, dermatology, and metabolic disorders(5253). Structurally similar to curcumin, calebin A inhibits nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) signaling promotes apoptosis, enhances chemosensitivity in cancer cells, and reduces inflammation through modulation of key pathways. Additionally, it suppresses melanogenesis, inhibits adipogenesis, activates AMP-activated protein kinase (AMPK), and exhibits neuroprotective properties. Preclinical studies suggest that the compound has low toxicity, but its oral bioavailability remains limited(5455). More recently, clinical trials with calebin A have shown its safety and lack of adverse events, plus positive effects on several cardiometabolic indices(5657). Both calebin A and curcumin have been shown not only to lack hepatotoxic effects(58) but also to provide numerous hepatoprotective benefits(59606162) (Fig. 2). Chemical structures of the top 3 cholesteryl ester transfer protein inhibitors. Ligand interactions in the CETP binding pocket In order to fully comprehend ligand-protein interactions and their effect on protein structure and inhibition mechanisms, it is essential to have a deep understanding of the protein’s three-dimensional structure and the characteristics of its substrate binding site. CETP has an elongated, banana-shaped structure featuring a long hydrophobic tunnel, approximately 60 Å in length, which facilitates the transfer of CEs and triglycerides between lipoproteins(6364) This tunnel is formed by the N- and C-terminal β-barrel domains and a central β-sheet domain, and it is lined with key hydrophobic residues, including ILE 15, LEU 23, ALA 202, ILE 205, LEU 206, PHE 263, PHE 265, and MET 433. A notable constriction at the "neck" region, which includes residues like PHE 263, plays a crucial role in regulating lipid movement. Inhibitors such as torcetrapib bind within this tunnel, making interactions with critical residues like GLN 199, SER 230, and HIS 232, effectively blocking lipid transport. Furthermore, the flexibility of the tunnel ends is important for lipid exchange, indicating that targeting these dynamic regions could improve inhibitor design(6364). The interaction profile of torcetrapib within the CETP binding pocket reveals a complex network of hydrophobic, polar, and π-π stacking interactions that stabilize the inhibitor within the central tunnel. Hydrophobic contacts are predominantly established with residues ILE 15, ALA 195, MET 194, VAL 198, ILE 215, LEU 217, LEU 261, PHE 263, PHE 265, LEU 467, VAL 438, ILE 443, and PHE 461. These residues, lining the hydrophobic tunnel, create a favorable environment that tightly accommodates the nonpolar regions of torcetrapib. In addition to hydrophobic stabilization, polar interactions are observed: GLN 199 and HIS 232 form hydrogen bonds with torcetrapib, anchoring the ligand deeper into the binding pocket. SER 230 also contributes to ligand stabilization through polar contacts, while ARG 201 is positioned nearby, potentially forming additional electrostatic interactions. Notably, aromatic stacking interactions are evident between the aromatic moieties of torcetrapib and the side chains of PHE 263 and PHE 265, enhancing binding affinity through π-π interactions(65). Additionally, the fluorine atoms help stabilize torcetrapib’s position within the tunnel through interactions with residues such as MET 459. The blockage of the CETP tunnel results in significant disruption of the normal lipid transfer process. This underscores the significant role of torcetrapib in the molecular basis of its inhibitory action, highlighting the importance of this research (Fig. 3). 2D visualization of torcetrapib placement and interactions in the cholesteryl ester transfer protein binding pocket. Analysis of calebin’s binding mode within the CETP tunnel revealed a distinct interaction profile compared to torcetrapib. The interaction analysis of calebin A and torcetrapib with the CETP enzyme reveals distinct binding profiles, despite both ligands occupying overlapping regions within the enzyme’s active site. Calebin A displayed a multi-modal interaction network, forming numerous hydrogen bonds with residues such as ARG 94, ASN 96, SER 101, and ASN 178 in the active site of CETP, in addition to engaging in strong π-π stacking interactions with aromatic residues TYR 132, PHE 174, and TRP 180. These aromatic contacts significantly enhance the stability of the complex. Furthermore, calebin is stabilized by salt bridges (with ASP 164 and LYS 176), π-cation interaction (with HIS 135), and hydrophobic contacts with residues like ILE 168 and LEU 171. These differences suggest that calebin and torcetrapib block the CETP tunnel through distinct interaction networks, which may translate into variations in their inhibitory potency and mechanism of action (Fig. 4 and Table 3). 2D visualization of calebin placement and interactions in the cholesteryl ester transfer protein. Table 3Residues involved in the Cholesteryl ester transfer protein interaction with torcetrapib and calebin, and the type of interaction.Interaction TypeTorcetrapibCalebin Hydrogen bonding GLN 199, SER 230, HIS 232ARG 94, ASN 96, SER 101, ASN 178, GLU 175 Hydrophobic contacts ILE 11, CYS 13, ILE 15, VAL 136, VAL 198, ILE 205, LEU 217, ALA 195, LEU 261, PHE 263, PHE 265, LEU 457, VAL 438, ILE 443, PHE 441, PHE 461, MET 459, PRO 464, PHE 463, LEU 206ILE 168, LEU 171, LEU 217, VAL 198 π-π Stacking PHE 263TYR 132, PHE 174, TRP 180 Polar interactions GLN 199, SER 230, HIS 232HIS 135, LYS 176, ASP 164 Residues involved in the Cholesteryl ester transfer protein interaction with torcetrapib and calebin, and the type of interaction. QM/MM studies The QM/MM energy minimization was carried out with the maximum number of minimization cycles set to 1000 steps, using the truncated Newton method. The convergence criterion was “energy and gradient”. The energy change and gradient criteria were set to 0.1 and 0.01, respectively. Long-range forces, which play an important role in the interaction between ligands and receptors and also between macromolecules, were updated every 10 steps. After convergence, there was a hydrogen bond interaction between THR 27 and the curcumin phenolic group, another interaction was between the phenolate anion of calebin and positively charged ARG 201 (Fig. 5A). It was found that calebin occupied the same position as CETP substrate CE (Fig. 5B). (A) Calebin-CETP QM/MM result after 100 ns; (B) simultaneous representation of calebin (red), torcetrapib (gray), and cholesteryl ester substrate (yellow) in the active site of CETP. CETP, Cholesteryl ester transfer protein. MD simulations MD simulations help us understand how protein-ligand complexes behave over time, which can help us more accurately comprehend the conformational flexibility, stability, and interaction mechanisms. Interactions that occur more than 30.0% of the simulation time between the ligand and the protein were shown in Fig. 6A and B for torcetrapib and calebin. A schematic of detailed ligand atom interactions with the protein residues (interactions that occur more than 30.0% of the simulation time) for (A) torcertrapib and (B) calebin Through MD simulations, important parameters such as RMSD, RMSF, and Rg can be assessed to evaluate the system’s stability and flexibility (Fig. 7A-C). RMSD quantifies the average displacement of atomic positions from the initial structure, indicating the complex’s overall stability. Generally, RMSD values stabilizing within 1-3 Å are considered acceptable for stable protein-ligand systems, as reported in studies(66). (A) RMSD, (B) RMSF, and (C) Rg plots of the CETP backbone of the apo-form, complexed with torcetrapib and calebin after 100 ns simulation. CETP, Cholesteryl ester transfer protein; RMSD, root mean square deviation; RMSF, root mean square fluctuation; Rg, radius of gyration. The simulation reached the plateau state according to apo CETP RMSD fluctuations, with an average of 2.032 ± 0.283 A. During 100 ns simulations, calebin demonstrated a similar RMSD (average of 2.515 ± 0.282 Å) in comparison with torcetrapib (average of 2.502 ± 0.332 Å). It can be concluded that the stability of the calebin-CETP complex was similar to that of the torcetrapib-protein complex during the simulation. RMSF provides information on residue-specific flexibility, identifying regions with high mobility such as loop regions, terminal residues, or adaptable binding sites. For well-folded, stable proteins, fluctuations typically range from 0.5 to 2 Å(67). The RMSF analysis showed that the calebin-CETP complex exhibited a similar pattern to tocetrapib-CETP and apo-CETP. Hence, the structure of CETP was not disturbed through binding to calebin. Rg quantifies the RMSD of atoms, indicating the compactness of a protein. Stable Rg values, generally fluctuating within ± 0.5 A, signify that the protein maintains compactness during simulation(68). In the calebin-CETP complex, Rg values remained steady at 34.259 ± 0.375 Å, which is slightly lower than the 34.413 ± 0.474 Å observed for Torcetrapib and higher than the 34.038 ± 0.236 Å for Apo-CETP. This indicates that calebin binding helped to preserve or enhance the protein’s structural compactness without significant conformational changes. Molecular mechanics-generalized born surface area (MMGBSA) calculation was performed by the Prime module of Schrödinger, and binding energies (ΔG) between CETP with calebin and torcetrapib were found to be -48.91 and -77.71 kcal/mol, respectively. DISCUSSION CETP plays a vital role in exchanging CE and TG among lipoproteins, making it an important target for treating atherosclerotic cardiovascular disease. Clinical failure and adverse effects related to previous CETP inhibitor medications express the need for further studies for safer and more efficient alternatives. In this study, 1,072 natural compounds were initially retrieved from the ZINC15 database using the ’standard ok’ and ’in man’ filters. These were screened for drug-likeness using Lipinski’s rule of five via ADMETLab, followed by toxicity and pharmacokinetic profiling through ADMETLab2, with particular attention to mutagenicity, hepatotoxicity, hERG channel inhibition, and P-gp interaction. This rigorous in silico filtering resulted in a refined library of 107 candidate molecules with favorable safety and drug-likeness characteristics. Calebin A demonstrated a promising binding affinity and interaction stability with CETP, outperforming torcetrapib in key aspects. Comparative binding interaction analysis revealed that calebin formed more stabilizing hydrogen bonds (ARG 201, SER 474, ASP 470) and hydrophobic contacts (ILE 15, ALA 195, VAL 196) within the CETP tunnel while also engaging π-π stacking interactions (PHE 463), similarly to torcetrapib. MD simulations further supported calebin’s potential, showing lower average RMSD and RMSF values than torcetrapib and the apoprotein, indicating a more stable protein-ligand complex with reduced residue-level fluctuations. Rg analysis also confirmed that the calebin-CETP complex maintained compactness over time, comparable to or superior to the native protein structure. The interaction of calebin A with the activity of CETP may also be indirectly inferred from changes in lipid profile observed in clinical studies. According to Majeed et al., 90-day supplementation with calebin A (50 mg/day) in individuals with obesity led to statistically significant and clinically relevant reductions in serum LDL-C and elevations in HDL-C concentrations(57). These findings were confirmed in a later study using the same dosage and duration of supplementation in subjects with metabolic syndrome(56). Since the elevation of HDL-C and reduction in LDL-C are hallmarks of CETP activity inhibition, these findings may imply that the favorable effects of calebin A supplementation on lipid profile could be mediated, at least in part, by inhibition of CETP. The present study had some limitations that deserve acknowledgment. It must be noted that the computational nature of analyses precludes any robust judgment on the pharmacological effect, which necessitates validation in experimental studies and models of dyslipidemia. Additionally, although computational analysis predicted some toxicity for calebin A, this is unlikely to be of relevance, as toxicity studies in rodents, as well as clinical trials, have shown that calebin A is both safe and well-tolerated(2656). CONCLUSION In summary, the integration of docking studies, MD simulations, and ADMET profiling presents a positive perspective for calebin A as a potential natural CETP inhibitor that shows minimal toxicity. Although the computational analyses provide promising initial evidence, it is essential to perform additional in vitro and in vivo studies to validate calebin A’s therapeutic efficacy and safety for use. To this end, cell-based models combined with CETP activity assays and experimental models of dyslipidemia could be useful in determining whether calebin A effectively inhibits CETP. Experimental studies could assess whether this inhibition results in a reduction of LDL-C and an elevation of HDL-C, which are established outcomes of CETP inhibition. Authors’ contributions F. Hadizadeh and A. Sahebkar conceptualized the study; F. Hadizadeh and M. Nejabat supervised the study; F. Mirabi performed the experimental parts of the study; K. Mohammadi analyzed the data; M. Nejabat and K. Mohammadi wrote the manuscript. All authors have read and approved the finalized article. Each author has fulfilled the authorship criteria and affirmed that this article represents honest and original work. Conflict of interest statement The authors declared no conflict of interest in this study. AI declaration The authors used AI-assisted technologies in the preparation of the graphical abstract of this manuscript. Data availability The data supporting the findings of this study are documented and available from the corresponding author upon reasonable request. Researchers interested in accessing the data should contact the corresponding author to discuss the terms and conditions of access. Ethics approval All procedures were approved by the Ethics Committee of the Mashhad University of Medical Sciences (IR.MUMS.PHARMACY.REC. 1403.042).