Computational Evaluation of Oleuropein Interactions with Alzheimer's Disease-Related Proteins Using Molecular Docking and Molecular Dynamics

Summary

This study presents an integrated molecular docking and molecular dynamics workflow to computationally evaluate the interactions of oleuropein with key Alzheimer’s disease-related proteins, revealing predicted binding interactions and preliminary complex stability across multiple targets.

Abstract

Alzheimer’s disease (AD) is a multifactorial neurodegenerative disorder associated with amyloid aggregation, tau pathology, and neuroinflammation. In this study, an integrated computational workflow combining ADMET prediction, PASS-based activity screening, molecular docking, and molecular dynamics (MD) simulations was employed to evaluate the interactions of oleuropein with five AD-related protein targets, including β-amyloid, tau, apolipoprotein E4 (ApoE4), triggering receptor expressed on myeloid cells 2 (TREM2), and complement protein C1q. ADMET and PASS analyses predicted pharmacokinetic properties and potential biological activities associated with neurodegenerative disease-related pathways. Docking analysis predicted favorable binding affinities across the investigated targets, with the strongest predicted interaction observed for the C1q protein (-7.5 kcal/mol). Molecular dynamics (MD) trajectory analyses, including root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), radius of gyration (Rg), dynamic cross-correlation matrix (DCCM), principal component analysis (PCA), and distance-based metrics, were used to evaluate the dynamic behavior of the protein-ligand complexes during the simulations. The computational analyses suggested relatively stable interactions for tau, C1q, and TREM2 complexes, whereas β-amyloid and ApoE4 exhibited comparatively higher conformational variability during portions of the simulations. These findings provide a preliminary computational assessment of oleuropein interactions with AD-related proteins and may support future experimental studies investigating its potential biological relevance in neurodegenerative disease models.

Introduction

AD represents the most common type of dementia that generates serious and escalating challenges worldwide^1,2. AD arises from a combination of genetic and environmental factors. Multiple pathogenic hypotheses have been proposed to explain AD, including cholinergic, amyloid, tau, inflammatory, oxidative stress, metal ion, excitotoxicity, microbiota-gut-brain axis, and autophagy-related mechanisms³. AD is a multifactorial disorder involving β-amyloid deposition, tau pathology, neuroinflammation, and genetic risk factors such as ApoE4 and TREM2^4,5. Due to the multifactorial nature of AD, multiple target-directed ligands (MTDLs) have emerged as promising therapeutic strategies based on polypharmacological interactions with Aβ, tau, ApoE4, TREM2-mediated microglial function, and complement-driven neuroinflammation^3,6,7. β-amyloid aggregation is a key pathological hallmark of AD that disrupts neuronal function and is modulated by ApoE4 isoforms and immune responses^8,9.

Hyperphosphorylation and other post-translational modifications of tau, including truncation, acetylation, ubiquitination, and glycosylation, contribute significantly to the formation of neurofibrillary tangles (NFTs), one of the major pathological hallmarks of AD. These modifications reduce the affinity of tau for microtubules, leading to microtubule destabilization, impaired axonal transport, synaptic dysfunction, mitochondrial abnormalities, and progressive neuronal degeneration. In addition, abnormal tau aggregation has been strongly associated with cognitive decline and disease severity in AD, often correlating more closely with neurodegeneration than amyloid burden itself. Furthermore, increasing evidence suggests that tau pathology interacts closely with neuroinflammatory signaling pathways, where activated microglia and inflammatory mediators may further exacerbate tau propagation and neuronal damage^10,11,12.

TREM2-mediated microglial activation has been associated with neuroinflammatory responses during AD progression¹³. Complement proteins such as C1q and C3 have been implicated in synaptic loss, complement-mediated neuroinflammation, and inflammatory signaling associated with AD pathology^14,15. Notch signaling, oxidative stress, inflammatory pathways, and ApoE4 further accelerate AD progression^9,16.

Olive leaf extract (OLE) is a bioactive phytocomplex rich in secoiridoids, flavonoids, triterpenes, and phenolic acids, as summarized in Table 1. These constituents have been associated with antioxidant and anti-inflammatory biological activities in previous studies¹⁷. Oleuropein is the principal secoiridoid identified in olive leaves, as it constitutes the primary phenolic component in these leaves. This renders oleuropein the most abundant phenolic secoiridoid identified in olive leaves, and its chemical structure is illustrated in Figure 1.

Previous studies have reported biological activities of oleuropein associated with amyloid aggregation, tau-related pathology, oxidative stress, and inflammatory signaling pathways^18,19. Additional studies have also suggested potential interactions of oleuropein with pathways related to ApoE4, TREM2-mediated microglial activation, and complement-associated neuroinflammation^20,21.

Despite the growing interest in oleuropein, its interactions with multiple AD-related protein targets have not been comprehensively evaluated within a unified computational framework. Most previous studies have primarily focused on its antioxidant properties or its potential effects on individual pathological pathways, particularly amyloid aggregation and tau-related toxicity. Therefore, the present study employed molecular docking and molecular dynamics simulations to computationally assess the predicted interactions and dynamic behavior of oleuropein against five AD-related protein targets, including β-amyloid, tau, ApoE4, TREM2, and C1q. These targets were selected because they represent distinct yet interconnected pathological mechanisms involved in AD progression. β-amyloid plays a central role in extracellular plaque formation, synaptic dysfunction, oxidative stress, and neuronal toxicity, whereas tau pathology is closely associated with neurofibrillary tangle formation, cytoskeletal disruption, and progressive cognitive decline²². ApoE4 represents the strongest genetic risk factor for sporadic AD and contributes to impaired amyloid clearance, enhanced neurotoxicity, lipid dysregulation, and increased neuroinflammatory responses²³. TREM2 is a key regulator of microglial activation and neuroimmune signaling and has been implicated in amyloid clearance, inflammatory regulation, and disease-associated microglial responses during AD progression²⁴. In addition, complement protein C1q plays a critical role in complement-mediated neuroinflammation, aberrant synaptic pruning, and microglia-associated synaptic degeneration in AD²⁵.

The overall in silico workflow used to evaluate oleuropein against the selected Alzheimer’s disease-related protein targets is summarized in Figure 2.

Protocol

1. ADMET prediction

The pharmacokinetic and toxicity properties of oleuropein were evaluated using a computational ADMET prediction platform. Predicted parameters included absorption, distribution, metabolism, excretion, BBB permeability, central nervous system (CNS) permeability, and toxicity-related properties. ADMET analysis was performed to assess the predicted pharmacokinetic profile and potential drug-likeness properties of oleuropein prior to molecular docking and MD simulations²⁶.

2. Prediction of anti-Alzheimer and neuroprotective activity

The canonical SMILES representation of oleuropein was used for computational prediction of anti-Alzheimer and neuroprotective activities²⁷. Probability of activity (Pa) and probability of inactivity (Pi) values were recorded, and only activities with Pa values greater than Pi values were considered potentially relevant for subsequent computational analyses.

3. Oleuropein ligand structure preparation

The three-dimensional (3D) structure of oleuropein was retrieved from a public chemical database in SDF format. Ligand geometry optimization and energy minimization were performed using the MMFF94 force field until convergence was achieved. The minimized ligand structure was converted into PDB format and subsequently prepared for molecular docking through assignment of polar hydrogen atoms and partial charges prior to conversion into PDBQT format.

Oleuropein was modeled in its predominant neutral form, and all ionizable residues were assumed to be protonated under physiological conditions (pH 7.4) during molecular docking and molecular dynamics simulations.

4. Protein retrieval and preparation

Protein structures associated with AD pathology were retrieved from the Protein Data Bank, including β-amyloid (2BEG), tau (5O3L), ApoE4 (1B68), TREM2 (5ELI), and C1q (2JG9). Target-specific structural information, including chain selection and experimental structure type, is summarized in Table 2.

Water molecules, ions, co-crystallized ligands, and non-essential heteroatoms were removed prior to receptor preparation. Structures were visually inspected for incomplete residues, missing atoms, and structural discontinuities before docking and simulation setup. Polar hydrogen atoms and Kollman charges were subsequently assigned, and receptor structures were converted into PDBQT format for docking studies. Default protonation states corresponding to physiological conditions (pH 7.4) were applied during receptor preparation, and all ionizable residues were assumed to adopt their predominant protonation states under physiological pH conditions. For β-amyloid, Model 1 from the deposited NMR ensemble was selected, and all chains within the deposited oligomeric assembly (chains A-E) were retained during receptor preparation. For the tau protein, all chains within the deposited cryo-EM filament assembly (chains A-J) were retained during receptor preparation. Similarly, all chains within the multimeric C1q assembly (chains A-F) were retained to preserve the deposited quaternary structure. For ApoE4 and TREM2, the deposited biologically relevant chains were retained during receptor preparation.

5. Molecular docking

1. Binding site prediction and grid box definition
   Potential ligand-binding pockets were predicted using a structure-based binding-site prediction workflow that evaluates cavity geometry, pocket volume, depth, hydrophobicity, and physicochemical descriptors associated with ligand-binding potential and druggability²⁸.
   Predicted pockets were ranked according to druggability score and geometric characteristics. The highest-ranked solvent-accessible pocket for each target was selected for molecular docking analyses. Residues corresponding to the selected binding pocket were extracted from the generated pocket-residue output files and subsequently visualized together with the corresponding protein structures using molecular visualization software.
   Center-of-mass coordinates of the selected binding-pocket residues were calculated and used for docking grid-box generation. Grid-box dimensions were defined to fully encompass the predicted binding region for each target protein. Exact docking grid-box center coordinates and dimensions for all targets are summarized in Table 3.
   NOTE: The complete executable command-line workflow and scripts used for center-of-mass calculations are provided in Supplementary Coding File 1.
2. Docking parameters and execution
   Molecular docking calculations were performed using AutoDock Vina 4.2.6 with an exhaustiveness value of 8 and 10 output binding modes for each protein-ligand complex. The binding pose with the lowest predicted binding affinity was selected for subsequent analyses. Protein-ligand interactions were visualized using molecular visualization software.
   NOTE: The command structure is adapted for each protein depending on the receptor file name and grid box coordinates. The docking output files were split into individual binding modes and converted to PDB format using command-line tools. The full commands are provided in Supplementary Coding File 1.

6. Molecular dynamics (MD) simulations

The MD workflow was executed within a cloud-based Conda/Miniforge computational environment configured for biomolecular simulation and trajectory analysis. GPU-accelerated runtime resources were used for all production simulations. Complete executable environment setup scripts, dependency installation commands, and simulation workflows are provided in Supplementary Coding File 2.

Protein-ligand complexes generated from molecular docking were used as starting structures for MD simulations. Docked ligand poses corresponding to the lowest predicted binding-affinity conformations were selected for topology generation and system assembly. Protein structures were prepared after removal of crystallographic solvent molecules and non-essential heteroatoms prior to simulation setup.

Topology generation and system assembly were performed using the ff19SB protein force field and the GAFF2 ligand force field. Complexes were solvated using the TIP3P explicit water model and neutralized with NaCl ions at a final ionic concentration of 0.15 M under periodic boundary conditions.

NOTE: All molecular dynamics simulations were executed within a cloud-based notebook environment configured with Python 3 runtime, T4 GPU hardware acceleration, and the latest available runtime version at the time of analysis.

1. MD equilibration parameters and Energy minimization
   Following topology generation, energy minimization was performed for 50,000 steps prior to equilibration. NPT equilibration simulations were subsequently conducted for 5 ns using positional restraints applied to protein backbone atoms under the following conditions: (1) integration timestep = 2 fs; (2) temperature = 298 K; (3) pressure = 1 bar; (4) harmonic restraint force constant = 700 Kj.mol⁻¹.nm⁻²; (5) trajectory frames saved every 10 ps; and (6) log files recorded every 10 ps.
   NOTE: The complete executable equilibration scripts and simulation commands are provided in Supplementary Coding File 2.
2. MD production parameters and run NPT production
   Production MD simulations were performed under NPT conditions for a total analyzed production trajectory time of 10 ns using the following parameters: (1) integration timestep = 2 fs; (2) temperature = 298 K; (3) pressure = 1 bar; (4) trajectory frames saved every 10 ps; and (5) log files recorded every 10 ps.
   NOTE: The complete executable MD simulation scripts are provided in Supplementary Coding File 2.
3. Trajectory concatenation and alignment
   Production trajectories generated from multiple simulation strides were concatenated into continuous trajectories after imaging under periodic boundary conditions. Periodic boundary artifacts, solvent molecules, and ions were removed before structural alignment and downstream analyses.
   Trajectory alignment was performed using protein Cα backbone atoms relative to the equilibrated reference structure to ensure consistent orientation throughout trajectory analyses.
4. Protein-ligand interaction analysis
   Protein-ligand interaction fingerprints and interaction persistence throughout the MD trajectories were analyzed using trajectory-based interaction profiling approaches. Interaction-network analyses were performed using processed production trajectories following solvent removal and structural alignment.
   NOTE: The complete executable interaction-analysis workflows are provided in Supplementary Coding File 2.
5. MD trajectory analyses
   Trajectory analyses were performed using aligned production trajectories following solvent removal, concatenation, imaging under periodic boundary conditions, and structural alignment. All analyses were performed on the 10 ns production trajectories after removal of solvent molecules and ions. Frames were analyzed at 10 ps intervals. Unless otherwise stated, structural analyses were performed after least-squares alignment of protein Cα atoms to the equilibrated starting structure of each corresponding complex.
   RMSD, RMSF, and radius of gyration (Rg) analyses were calculated using protein Cα atoms to evaluate structural stability and conformational dynamics throughout the simulations. RMSD was calculated for protein Cα atoms relative to the equilibrated reference structure. RMSF was calculated for protein Cα atoms over the aligned production trajectory. Rg was calculated using protein Cα atoms to monitor global compactness.
   Linear Interaction Energy (LIE), intermolecular distance measurements, principal component analysis (PCA), and dynamic cross-correlation matrix (DCCM) analyses were subsequently performed using trajectory frames sampled at 10 ps intervals throughout the analyzed production trajectories. Interaction-energy analysis was calculated between oleuropein and the protein atoms of each complex over the production trajectory. Center-of-mass distance was calculated between oleuropein, and the selected binding-pocket residues used for docking-grid generation, whereas minimum distance was calculated as the shortest atom-to-atom distance between oleuropein and the corresponding protein binding-pocket residues in each frame. PCA was performed using protein Cα coordinates after trajectory alignment. DCCM analysis was calculated from protein Cα atomic fluctuations over the aligned production trajectory.
   NOTE: Each molecular dynamics simulation and trajectory-analysis step is provided as a separate executable code section in Supplementary Coding File 2. Supplementary Coding File 2 includes the scripts and commands used for system preparation, production simulation, trajectory processing, and downstream analyses. The complete workflow version corresponding to the submitted manuscript is archived in Zenodo with https://doi.org/10.5281/zenodo.20134270, and the corresponding GitHub release is available at: https://github.com/Elshekh00/molecular-dynamics-protein-ligand/releases/tag/v1.0.0.

Results

Prediction of ADMET profile

The predicted pharmacokinetic and toxicity profile of oleuropein was evaluated using an in silico ADMET prediction workflow. The predicted ADMET properties of oleuropein, including AMES toxicity, hepatotoxicity, hERG I and II inhibition, as well as oral rat acute and chronic toxicity, are listed in Table 4. Oleuropein was predicted to be non-mutagenic (AMES negative). Oleuropein was predicted to exhibit low predicted hepato...

Discussion

The primary objective of this integrated in silico study was to evaluate the predicted multi-target interaction potential of oleuropein against selected pathological proteins implicated in AD. The findings from ADMET prediction, PASS analysis, molecular docking, and MD simulations provide a computational basis for considering oleuropein as a molecule with predicted multi-target interaction potential against selected AD-related proteins. Given the multifactorial nature of AD, which involves amyloid aggregation, t...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
AmberTools 24.8	AmberMD development team	N/A; https://ambermd.org/AmberTools.php	Used for topology preparation and molecular simulation file handling. Installed from conda-forge. RRID: SCR_018497
AutoDock Tools 1.5.7	Scripps Research Institute	N/A; https://autodocksuite.scripps.edu/adt/	Used for receptor and ligand preparation and PDBQT file generation. RRID: SCR_012746
AutoDock Vina 4.2.6	Scripps Research Institute	N/A; https://vina.scripps.edu/	Docking engine used for molecular docking calculations. Version should be reported according to the executed workflow environment. RRID: SCR_011958
Avogadro 2.0.0	Avogadro Chemistry Software	N/A; https://avogadro.cc/	Used for ligand visualization, force-field assignment, and energy minimization. RRID: SCR_015983
Conda 25.3.1	Anaconda, Inc. / conda-forge community	N/A; https://docs.conda.io/	Package and environment manager used in the computational workflow. RRID: SCR_018317
Discovery Studio Visualizer 2025	Dassault Systèmes	N/A; https://discover.3ds.com/discovery-studio-visualizer-download	Used for protein visualization and preprocessing. RRID: SCR_008398
DoGSiteScorer / ProteinsPlus	ZBH Center for Bioinformatics	N/A; https://proteins.plus/	Used for binding-site prediction and pocket druggability scoring. RRID: Not available
ff19SB force field	AMBER force-field family	N/A; https://ambermd.org/	Protein force field used for molecular dynamics system preparation. RRID: Not available
GAFF2 force field	AMBER force-field family	N/A; https://ambermd.org/	Ligand force field used for molecular dynamics system preparation. RRID: Not available
Google Colab	Google	N/A; https://colab.research.google.com/	Cloud notebook environment used for workflow execution. Runtime type: Python 3; hardware accelerator: T4 GPU. RRID: SCR_018009
Mamba 2.1.1	QuantStack / conda-forge community	N/A; https://mamba.readthedocs.io/	Fast package manager used for installing Conda packages. RRID: Not available
Matplotlib 3.10.0	Matplotlib development team	N/A; https://matplotlib.org/	Used for generation of molecular dynamics analysis plots. RRID: SCR_008624
MDAnalysis 2.8.0	MDAnalysis development team	N/A; https://www.mdanalysis.org/	Used for trajectory processing and structural analysis. RRID: SCR_025610
MDTraj 1.11.1	MDTraj development team	N/A; https://www.mdtraj.org/	Used for molecular dynamics trajectory handling and analysis. RRID: Not available
Miniforge	conda-forge community	N/A; https://github.com/conda-forge/miniforge	Conda-based environment distribution used to install scientific dependencies. RRID: Not available
NumPy 2.0.2	NumPy development team	N/A; https://numpy.org/	Numerical computing dependency used in the computational workflow. RRID: SCR_008633
Open Babel 3.1.0	Open Babel Project	N/A; https://openbabel.org/	Used for molecular file format conversion. RRID: SCR_014920
OpenMM 8.5.1	OpenMM development team	N/A; https://openmm.org/	Molecular dynamics simulation engine used for system setup, equilibration, and production simulations. RRID: SCR_000436
pandas 2.3.3	pandas development team	N/A; https://pandas.pydata.org/	Used for tabular data handling and analysis outputs. RRID: SCR_018214
ParmEd 4.3.1	ParmEd development team	N/A; https://parmed.github.io/ParmEd/html/index.html	Used for topology and parameter file handling. RRID: Not available
PASS-Way2Drug Server	Way2Drug	N/A; https://www.way2drug.com/passonline/	Used for predicted biological activity profiling. RRID: SCR_001971
PDBFixer 1.12	OpenMM development team	N/A; https://github.com/openmm/pdbfixer	Used for protein structure repair and preparation when required. Installed from conda-forge. RRID: Not available
pkCSM Server	University of Melbourne	N/A; https://biosig.lab.uq.edu.au/pkcsm/	Used for predicted ADMET profiling. RRID: Not available
ProLIF	ProLIF development team	N/A; https://prolif.readthedocs.io/	Used for protein-ligand interaction fingerprint analysis. Runtime import reported the version as 0+unknown. RRID: Not available
PubChem Database	National Center for Biotechnology Information / National Institutes of Health	N/A; https://pubchem.ncbi.nlm.nih.gov/	Used for ligand structure retrieval. RRID: SCR_004284
py3Dmol 2.5.4	py3Dmol development team	N/A; https://3dmol.csb.pitt.edu/	Used for molecular visualization in the notebook workflow. RRID: Not available
PyMOL 3.1.6.1	Schrödinger, LLC	N/A; https://pymol.org/	Used for molecular visualization and binding-pocket coordinate inspection. RRID: SCR_000305
Python 3.12.13	Python Software Foundation	N/A; https://www.python.org/	Programming language used to execute the computational workflow. RRID: SCR_008394
pytraj 2.0.6	AmberMD / pytraj development team	N/A; https://amber-md.github.io/pytraj/latest/index.html	Used for trajectory processing and analysis. RRID: Not available
RCSB Protein Data Bank	Research Collaboratory for Structural Bioinformatics	N/A; https://www.rcsb.org/	Used for protein structure retrieval. RRID: SCR_012820
RDKit 2025.03.1	RDKit development team	N/A; https://www.rdkit.org/	Used for cheminformatics processing of ligand structures. RRID: SCR_014274
SciPy 1.17.1	SciPy development team	N/A; https://scipy.org/	Used for statistical analysis, clustering, interpolation, and distance-matrix handling in trajectory analyses. RRID: SCR_008058
seaborn 0.13.2	seaborn development team	N/A; https://seaborn.pydata.org/	Used for plotting and visualization of analysis outputs. RRID: SCR_018132
SWISS-MODEL	Swiss Institute of Bioinformatics	N/A; https://swissmodel.expasy.org/	Used for structure modeling or reconstruction when required. RRID: SCR_018123
TIP3P water model	Original TIP3P water model / AMBER-compatible implementation	N/A; https://ambermd.org/	Water model used for molecular dynamics solvation. RRID: Not available