Identification of novel neuraminidase 1 modulators as potential therapeutics for Alzheimer’s disease using virtual screening and molecular dynamics simulations

Abstract

Alzheimer’s disease (AD) is a neurodegenerative disorder caused by the accumulation of β-amyloid (Aβ) plaques and neurofibrillary tangles, resulting in neuronal dysfunction and cognitive decline. The neuraminidase isoenzyme NEU1 is a most ubiquitous mammalian enzyme, involved in various cellular mechanisms. The deficiency of NEU1 has been implicated in the pathophysiology of AD, significantly in amyloid precursor protein (APP) metabolism and Aβ clearance. Despite extensive research, no potent NEU1 modulator has been developed to regulate its activity for therapeutic intervention in AD. The present work aims to identify potential NEU1 modulators from a library of seaweed Metabolites Database through molecular docking, ADMET analysis, and molecular dynamics (MD) simulations. A library of 1,077 seaweed metabolites was screened, identifying 20 active compounds, of which 4 met Lipinski’s Rule of Five criteria. ADMET profiling revealed favorable pharmacokinetic properties for BE003, BS032, and RG007, with good blood-brain barrier permeability and bioavailability. Molecular docking demonstrates that BE003, BS032, RG007, and BD039 metabolites exhibited the highest binding affinities for the NEU1 active site. Additionally, MD simulation and MM-GBSA validated the stability of the metabolite-protein complex, with BE003 demonstrating the most stable interactions. Comparative docking against a natural substrate (Neu5Ac) and a NEU1 inhibitor (17f) revealed that BE003 shares significant interaction, RMSD stability profiles with the substrate and loop conformational dynamics while differing from the inhibitor. Our findings emphasize the potential of this modulator as a novel therapeutic target against NEU1 in AD treatment. Further experimental validation and preclinical studies are needed to confirm its efficacy in modulating NEU1 activity.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-22955-7.

Introduction

Alzheimer’s disease (AD) is the leading cause of dementia, affecting millions of people worldwide. AD accounts for 60% to 70% of cases of emerging neurological impairment in elderly patients¹. Several studies have been conducted to discover therapeutic agents against Alzheimer’s disease (AD)². Despite the significant therapeutic approaches for managing symptoms being available, the effectiveness of current therapeutic drugs remains limited. AD is associated with a gradual decline of two important catabolic mechanisms in eukaryotic cells, the ubiquitin-proteasome system (UPS) and the autophagy lysosomal pathway (ALP)⁴. The UPS is responsible for the degradation of aberrant and nuclear proteins, such as short- and long-lived proteins⁵, while ALP contributes to the degrading of proteins accumulated collectively with damaged organelles⁶. The cytoskeleton-associated proteins Tau and Aβ are phosphorylated to form neurofibrillary tangles (NFTs), which lead to senile plaques (SP)⁷. These Aβ aggregates damage the hippocampal part, leading to dysfunctional neural activity and cell death⁸. Four types of neuraminidase isoenzymes have been identified in mammals: NEU1, NEU2, NEU3, and NEU4, which exhibit multiple subcellular localization, and varying preferences for glycan substrates⁹. The aberrant activity of human neuraminidases has been associated with various pathological conditions such as lysosomal storage disorders, metabolic disorders, cancers, and neurodegenerative diseases¹⁰. NEU4 has broad substrate specificity and is located on mitochondria, lysosomes, and the endoplasmic reticulum, which is significantly involved in the degradation of gangliosides¹¹. NEU3 works similar to the NEU4 at the outer or inner leaflet of the plasma membrane¹². NEU2 is the only sialidase with a known three-dimensional structure; however, its physiological substrates remain unidentified¹³.

The NUE1 mammalian protein sequence comprises a single domain 415 in length. The three-dimensional model of the NEU1 protein has 24 β-strands and 4 α-helices and adopts a β-propeller shape. The NEU1 is the Predominant homolog, shares minimal sequence similarity with its family members, and functions in lysosomes and on the plasma membrane¹⁴. Furthermore, ubiquitously expressed NEU1 enzyme mediates the removal of the sialic acid residues from glycoconjugates, modulating several pathological cellular functions¹⁵. NEU1 is closely linked to the implication of AD etiology, where it controls APP metabolism by desialylation¹⁶. This pathologic characteristic is associated with impaired autophagy macrophages, which qualitatively revealed impaired Aβ clearance and the increased aggregation of amyloid plaques in AD development¹⁷ (Fig. 1).

Fig. 1.

Mechanism of the NEU1 in implication of AD. In AD the low expression of NEU1 leads to lysosomal dysfunction, and sialic acid imbalance, resulting in Aβ accumulation and neuronal damage, a hallmark of AD. The potential NEU1 modulators are suggested to modulate NEU1 activity, which may help in preventing neurodegeneration.

Several studies have been conducted that revealed the involvement of NEU1 in AD. Sialidosis (Neu1-/-) mice showed a severe neurological phenotype with symptoms resembling AD ¹⁸. Within the lysosomes, APP undergoes processing into a soluble fragment and a beta C-terminal fragment (CTF). The CTF is cleaved to form Aβ peptides, which are released extracellularly through lysosomal exocytosis. This increased lysosomal exocytosis disrupts tissue integrity. In Neu1-/- neurons, the deposition of sialylated APP (Sia-APP) in lysosomes promotes the secretion of toxic Aβ peptides into the cerebrospinal fluid (CSF), activation of the early onset of amyloidosis, a hallmark of AD¹⁹. Another study demonstrates that NEU1 signaling can enhance macrophage polarization in the M2 state, lowering harmful oligomer production²⁰. Therefore, NEU1 is considered a target with therapeutic potential for treating AD. NEU modulates the sialylation state of glycans for cell surface glycoproteins and gangliosides, which is crucial in disease and homeostasis. Low NEU1 expression leads to the impaired processing of Aβ accumulation in the brain, which promotes neurodegeneration²¹. Guo et al.., (2018) identify potential inhibitors for human NEU1 and NEU2 isoenzymes, with 17f exhibiting the highest potency and 330-fold selectivity for NEU1. These inhibitors, particularly 17f, also revealed selective inhibition in mouse orthologs, which makes them useful tools for understanding the function of NEU enzymes in biological systems²². Moreover, the literature demonstrated many selective inhibitors of human NEU1 based on DANA scaffold alteration, such as the c9 amino analog of DANA²³. Despite the importance of NEU1 in AD pathophysiology, no modulator has been developed to target the NEU1 enzyme to manage AD. The elucidation of ligand binding interaction is necessary to develop more effective and potent drugs for this novel potential target. Seaweed metabolites ligands are widely abundant and number in the thousands²⁴; their potential uses in this regard must be investigated. A secondary metabolites library was used to find promising modulators against AD. Seaweed-derived metabolites interact with various proteins enzymatically and can modulate biological processes, making them promising targets for pharmaceutical applications²⁵.

The aim of this study is to identify and assess putative NEU1 modulator derived from seaweed metabolites for their ability to modulate NEU1 activity as a treatment approach against AD using in-silico approaches. In-silico approaches plays a pivotal role in modern drug discovery by significantly reducing the time and cost associated with traditional experimental approaches²⁶. To strengthen the assessment, the study includes a comparison with a natural NEU1 substrate (Neu5Ac) and a selective NEU1 inhibitor (17f)³⁹ as reference controls. Molecular docking was used to screen the seaweed metabolite pool against NEU1 and analyze binding interactions in comparison to the controls. Promising candidates were subsequently investigated using molecular dynamics simulations (RMSD analysis) and MM-GBSA binding energy calculations, enabling a detailed comparison of interaction stability and conformational behavior between the metabolite hits and reference compounds. modulator.

Methods

The methodology of this study consists of a three-stage screening approach to identify potential NEU1 modulators. Initially, High-Throughput Virtual Screening (HTVS) was conducted. AutoDock Vina⁴⁰ was employed for molecular docking studies to predict the binding affinity and interaction profiles of the compounds with NEU1.AutoDock tools⁴¹ were utilized to optimize the receptor and ligands. PyMol(https://www.pymol.org/) was used to visualize and check the interactions between ligands and NEU1. MD simulations were performed using GROMACS⁴², which provided insights into the conformational changes and binding stability over time. While, MM-GBSA (Molecular Mechanics/Generalized Born Surface Area) analysis were carried out using gmx_MMPBSA to estimate the binding free energy. The detailed workflow of the study is presented in Fig. 2.

Fig. 2.

Detailed Workflow of the study.

Receptor preparation

The crystallized structure of the target protein, NEU1 (neuraminidase 1) was obtained from AlphaFold (https://alphafold.ebi.ac.uk/) with the id Q99519. The structure was assessed to ensure its reliability and accuracy using PROCHECK server⁴³, which assess the stereochemical quality and structural integrity of structure.

To prepare the NEU1 structure for molecular docking simulations, the protein underwent a rigorous optimization procedure. Gasteiger charges and Polar hydrogens were added and the optimized structure was sin the .pdbqt format, which is compatible with Auto Dock Vina docking software. The whole optimization process, including the addition of charges and hydrogens, was carried out with AutoDock Tools⁴¹.

Seaweed metabolites library Preparation

In order to find putative NEU1 modulators, a library of 1,077 unique seaweed-derived metabolites was downloaded from the Seaweed Metabolite Database (SWMD) (https://www.swmd.co.in/About.html). The SWMD is a library that catalogs bioactive metabolites found in seaweed species, valuable resources for exploring novel compounds with pharmaceutical potential. Metabolites produced by live organisms are commonly exploited in drug development due to their diverse chemical structures and biological activities. Metabolites that were flexible with high number of rotatable bonds were removed, resulting in a refined set of 1,038 compounds used for docking.

Seaweed-derived metabolites are important because of their unique chemical diversity, which supports their survival in marine ecosystems. These chemicals possess antibacterial, antioxidant, anti-inflammatory, and enzyme-modulating characteristics, making them ideal candidates for medicinal uses⁴⁴. As modulators, these metabolites could potentially bind to and activate NEU1 (neuraminidase 1). Targeting NEU1 with particular modulators may assist to modify its activity, offering up new paths for therapeutic intervention in disorders associated with its dysregulation. The Metabolites library acquired from the SWMD was in 2D format. The metabolites were converted into 3D configurations using Openbabel. The metabolites were then optimized for virtual screening using AutoDock Tools, which included adding polar hydrogen and adjusting the Gastgier charges. The compounds were then stored in .pdbqt format for screening against NEU1.

Molecular docking

Molecular docking was performed using AutoDock Vina⁴⁰, a popular program for predicting the binding orientation and affinity of small compounds to their target proteins. To cover the complete active site of the target protein NEU1, grid box dimensions were adjusted to 45 Å × 45 Å × 45 Å for the x, y, and z axes. The catalytic site was selected based on reported key active-site residues. The grid box center was set at center_x = − 9.416, center_y = 1.970, and center_z = − 14.983 to encompass these essential residues and ensure accurate docking at the catalytic pocket.The exhaustiveness was set to 16 and the number of output binding modes was set to 10. To streamline the docking process, a bash script was used which automate docking runs for all compounds, reducing time and ensuring consistency across the analyses. AutoDock Vina assesses binding affinity of each ligand to the active site by computing binding energy (ΔG), which implies the strength of the interaction between the ligand and the target protein. The docked ligands were sorted according to their binding energy. The top twenty compounds with the best binding energy scores were chosen for further analysis. PyMol (https://www.pymol.org/) and Discovery Studio Visualizer (https://www.3ds.com/products/biovia/discovery-studio) were used to display and analyze these chosen compounds in order to better understand their binding modes and interactions.

ADMET analysis

ADMET (Absorption, distribution, metabolism, excretion, and toxicity) characteristics are critical to drug candidate success. Most medication candidates fail due to toxicity. Thus, rigorous ADME and toxicity analysis is critical in discovering prospective drug-like compounds. The pkCSM (https://biosig.lab.uq.edu.au/pkcsm/prediction)⁴⁶ and SwissADME servers(https://www.swissadme.ch/) ⁴⁷were used for ADME analysis, and the ProTox server (https://tox.charite.de/protox3/) was used to assess compound toxicity predictions. The ADMET properties of the compound were evaluated for its Bioavailability score, blood brain barrier permeability, Lipophilicity, and water solubility.

Molecular dynamics simulations

Further molecular dynamics simulations were performed on the selected ligand and NEU1 complexes were to assess their stability using GROMACS 2024 ⁴². MD simulations were performed for apo NEU1 and NEU1 complexed with three selected candidate ligands. For the protein topology, the charmm36-jul2022 ⁴⁸ was utilized, which provides accurate parameterization for proteins and biomolecules. For the ligand topologies, the CGENFF server⁴⁹ was employed, which generates force field parameters compatible with CHARMM for small molecules. The simulation systems were solvated using a TIP3P water model within a cubic box. The dimensions of the box were chosen such that the distance between the edges of the box and the protein on each side was 1.0 nm, ensuring sufficient space for water molecules and ions around the protein. To neutralize the systems, Na + ions were added, balancing the charge of the solute. Next, the systems underwent energy minimization using the steepest descent method for a total of 1,000 steps, ensuring the systems reached a local energy minimum. Following energy minimization, the systems were equilibrated to establish a stable temperature and pressure environment. The NVT ensemble was applied to stabilize the temperature of the systems, followed by the NPT ensemble to stabilize the pressure. These equilibration steps ensured the systems reached a thermodynamically stable state before the production run. Finally production run was performed for 100ns time. Upon completion of the simulations, the trajectories were analyzed using GROMACS utilities.

Principal component analysis (PCA)

PCA is one of the most effective strategies for reducing complexity and extracting intense movements in MD simulation analysis ⁵⁰. This method produced a matrix for all trajectories after eliminating rotational and translational motions. The essential dynamics technique was utilized to compute the eigenvalues, eigenvectors, and projections along the first two principal components (PC). By diagonalzing the matrix, eigenvectors and eigenvalues are calculated, where eigenvalues representing the magnitude of the eigenvectors. The covariance matrix was calculated and diagonalized using Gmx covar module, and the eigenvectors were analyzed using Gmx anaeig.

Binding free energy calculations

Following completion of the molecular dynamics simulations, binding free energy calculations (MM-GBSA) were used to assess the stability of the ligand-NEU1 complexes. This analysis is crucial for understanding the strength and nature of the interactions between the ligands and protein. The effective binding free energies were calculated using the gmx_MMPBSA module⁵¹. For effective binding energy calculations, 2500 frames were extracted from the final 25 ns of the 100 ns molecular dynamics trajectory at intervals of 10 ps. During the MM-GBSA calculations, every 5th frame was considered, resulting in a total of 500 frames used for the analysis.Gmx_MMPBSA calculate the binding free energy of the ligand-protein complex as follows:

1

The binding free energy is further broken down into various energetic components, expressed as:

2

Where:

3

4

is the molecular mechanics energy comprising bonded and non-bonded.

(electrostatics and van der waals) energy terms, is solvation energy comprising polar and non-polar solvation, and is the entropic contribution. The non-polar solvation contribution is computed as SASA while polar solvation is estimated as Poisson Boltzmann (P.B.) or Generalized Born (G.B.) models. This detailed analysis highlights not only the strength of ligand binding but also provides insights into the driving forces, such as hydrophobic interactions, electrostatic contributions, and entropy loss, that stabilize the complex.

Results

Virtual screening

To find potential candidates for NEU1 in AD, high throughput virtual screening was carried out against a library of 1,077 metabolites. The 3D structure of NEU1 was not present in the Protein Data Bank (PDB), the structure of NEU1 was retrieved from AlphaFold (ID: Q99519) which was then optimized through energy minimization. The quality of the structure was confirmed using PROCHECK to ensure its reliability for docking studies. The ramachandran plot of the structure is presented in Figure S1. Out of 366 residues, 263 residues were in the most preferred areas of the Ramachandran plot, showing that the vast majority of the residues adopt very stable and energetically favorable conformations, which is a strong sign of a well-refined and reliable protein structure. 42 residues were in additionally allowed regions, with just 2 residues being in generously allowed zones. There are no residues in banned areas, indicating that the protein has no steric conflicts or severe structural distortions. Additionally to assess the structural reliability of the AlphaFold-predicted model, we analyzed both the predicted Local Distance Difference Test (pLDDT) scores and the Predicted Aligned Error (PAE) map. As shown in Figure S3 and S4, the structured core of the protein exhibited very high confidence (pLDDT > 90), while only the N-terminal disordered region showed low confidence (pLDDT < 50). This flexible tail was excluded from the docking grid to avoid introducing noise.

Screening was performed on the metabolite library against the NEU1 structure. After screening the docked compounds were ranked according to their binding affinity (kcal/mol), with higher scores suggesting stronger interactions with the NEU1.

The top twenty docked compounds were chosen for further investigation. These compounds were initially evaluated for drug-likeness by counting the number of rule-of-five violations. The rule of five is a collection of parameters that predict whether a chemical is orally bioavailable, and a breach of these rules indicates poor drug-like qualities. Four of the top twenty compounds (BE003, BS032, RG007, and BD039) met the criteria for the rule of five, suggesting their potential as drug-like possibilities. The docked position and conformation of these four compounds are shown in Fig. 3. The details of these top twenty compounds, including their binding scores and drug-likeness properties, are provided in Table 1.

Fig. 3.

Docked conformation of the four selected metabolites (BE003, BS032, RG007, and BD039) based on the rule of five criteria.

Table 1.

The binding energies (kcal/mol) of the top 20 and number of ‘Rule of 5 Violations’, molecular weight, ALOGPS, and binomial name.

Accession number	Binding energy (Kcal/mol)	# of Rule of 5 violations	Molecular weight (g/mol)	ALOGPS	Binomial name
BE004	-10.9	1	602.45	41.5	Ecklonia stolonifera
BS070	-10.8	2	554.71	7.33	Stypopodium flabelliforme
RL322	-10.6	1	492.73	4.97	Laurencia obtuse
BE009	-10	1	602.45	41.4	Ecklonia cava
BE010	-10	1	496.37	3.60	Ecklonia cava
RL332	-10	2	647.67	5.05	Laurencia venusta
BS057	-9.9	1	454.64	6.45	Stypopodium flabelliforme
RL354	-9.6	2	588.10	8.15	Laurencia tristicha
BS052	-9.5	1	426.58	6.50	Sargassum tortile
RL015	-9.5	2	588.41	8.15	Laurencia microcladia
BS062	-9.4	1	426.67	7.50	Sargassum sp.
RC025	-9.4	2	566.36	7.01	Callophycus serratus
BE003	-9.3	0	372.28	2.71	Ecklonia stolonifera
BS032	-9.3	0	412.60	5.0	Stypopodium flabelliforme
BS055	-9.3	1	304.46	3.50	Stypopodium flabelliforme
BS072	-9.3	1	510.66	5.57	Stypopodium flabelliforme
RG007	-9.3	0	414.62	4.56	Galaxaura marginata
RL513	-9.3	1	603.62	4.67	Laurencia viridis
BS051	-9.2	1	426.58	5.37	Sargassum tortile
BD039	-9.1	0	420.53	2.56	Dictoyota dichotoma

Interaction analysis of selected metabolites with NEU1

Molecular docking is an in silico approach that aids in Predicting the binding affinity and conformation of molecule to a target protein. The current study involved screening a seaweed metabolite library against the NEU1 protein. The active site of NEU1 is a deep groove-like structure that binds sialic acid-containing substrates. The sialic acid is positioned in the active site via interactions with several residues around the groove, allowing the enzyme to cleave the glycosidic bond between the sialic acid and its sugar chain³⁶. The metabolite BE003 interacts with several key amino acid residues within the binding pocket of NEU1 with a binding score of -9.3 kcal/mol. Strong hydrogen bond interactions were observed between BE003 metabolite and GLN-282, HIS-220, LEU-139, TYR-370, and GLY-221 stabilizing its binding conformation. Additionally, GLU-264 and ARG-280 engage in ionic interactions with the aromatic core of the BE003 metabolite while the ILE-79 and PRO-200 formed Pi alkyl interactions (Fig. 4A). The metabolite compound exhibits a Pi anion interaction with glutamate residue and Pi-cation interaction with ARG-280, contributing to the overall stability of the metabolite compound-protein complex.

Fig. 4.

Metabolite and NEU1 interactions and binding conformation of (A) BE003, (B) BS032, (C) RG007, and (D) BD039.Residues forming hydrogen bonds are shown as sticks and labels. The hydrogen bonds are shown in yellow dotted lines.

BS032 and RG007 metabolites showcase a significant binding affinity towards the NUE1 protein with binding energy recorded at -9.3 kcal/mol. Even slight variations in binding energies can lead to substantial differences in binding stabilities. BS032 exhibits a distinct binding conformation compared to BE003. BS032 metabolite interacts with ARG-280, GLU-264, and TYR-370 by strong hydrogen bonds while the LYS-255, LEU-223, and HIS-160 formed Pi alkyl interactions (Fig. 4B). RG007 metabolite establishes a hydrogen bond with GLN-282, ARG-280, and HIS-220. Moreover, the other residues such as ASP-103, SER-102, and GLU-104 contribute additional hydrogen bonds, making RG007 one of the most extensively interacting metabolites among the four metabolic compounds (Fig. 4C). Moreover, the HIS-160 formed Pi alkyl interaction with the compound.

BD0039 binds to the protein with a marginally enhanced binding energy of -9.1 kcal/mol. BD0039 forms a strong interaction with GLN-282 and ARG-280, similar to RG007 metabolite. Furthermore, the residues ARG-341 and TYR-370 were involved in hydrogen bonding and reinforcing ligand orientation within the pocket (Fig. 4D). Moreover, HIS-160, ALA-159, HIS-220, and ALA-197 facilitate the Pi alkyl interaction which enhances the binding affinity of the compound. The docking results depicted in Fig. 4; Table 2, revealed the hydrogen bonding interactions between the selected metabolites and the NEU1. Each metabolic compound interacts distinctly with the NEU1 protein. The varied binding energies and the specific residues involved in these interactions revealed the potential of these metabolites to modulate the protein’s function, leading to unique therapeutic outcomes.

Table 2.

Detail of top four Docking statistics of metabolites with target protein.

Docking protocol validation and predictive performance (ROC-AUC)

To validate the docking protocol followed in the study, we self docked the DANA inhibitor complexed with NEU2 (PDB ID: 1VCU), a known inhibitor of both NEU1 and NEU2. The inhibitor DANA was re-docked into the active site of NEU2. The re-docked conformation was then superimposed on the original co-crystallized conformation to evaluate the accuracy of the docking procedure. DANA attained the exact binding conformation as the crystallographic conformation with an RMSD of 0.241 Å indicating the accuracy and precision of docking method (Figure S2). To further validate the virtual screening protocol, DANA inhibitor and the candidate compound along with decoy compounds (15) generated using the DUD-E ⁵² ((based on the candidate ligands and the known inhibitor), were docked into the NEU2 active site (File S1.csv). The candidate compound and the DANA inhibitor were labeled as 1 (actives) and the decoys were labeled as 0 (Inactives). ROC-AUC and enrichment factors were calculated based on the docking scores using scikit-learn python library. The ROC-AUC value of 0.90 and high enrichment factors (EF5% = 4.0, EF10% = 4.0) indicated that the docking protocol reliably prioritized active compounds over decoys, demonstrating its predictive power (Figure S5).

ADMET analysis

ADMET properties of the four selected metabolites were estimated using pkCSM and SwissADME servers. Table 2 describes the ADMET properties of the four metabolite compounds. Among the four compounds, all demonstrated no toxicity in terms of organ and end-point toxicity, indicating that they are non-toxic. Based on the predicted logBB values, BS032 (–0.088), RG007 (0.292), and BD039 (–0.086) showed favorable BBB permeability, while BE003 (–1.944) was predicted to have relatively lower passive permeability across the BBB. All compounds show high intestinal absorption, with values of 100%. BS032 has 93.58% intestinal absorption. All compounds have the same bioavailability score of 0.54, suggesting that their overall bioavailability is similar. Three compounds (BE003, BS032, and R007) exhibited good solubility while the compound BD039 has low water solubility (-3.28), indicating that it may face challenges in formulation. Moreover, this compound also has less lipophilicity which might reduce its ability to cross lipid membranes as efficiently (Table 3). All compounds showed no hERG I inhibition, indicating a low risk of cardiotoxic effects and supporting their safety profile for further development. Notably, none of the selected compounds were predicted to inhibit CYP3A4, suggesting a low risk of metabolic interference. This favorable prediction supports the metabolic safety of the compounds. Additionally none of the compounds are likely to be mutagenic, which supports their toxicological safety. Based on these observations three compounds BE003, BS032, and Rg007 were selected for further validation studies.

Table 3.

ADME properties of the selected four compounds, (water solubility logs (ESOL): log S is the logarithm of the water solubility (S), log P_o/w (WLOGP): is the logarithm of the partition coefficient (P) between water and an organic solvent).

ADME properties	BE003	BS032	RG007	BD039
Water solubility Log S (ESOL) (predicted by SwissADME)	-4.06	-6.75	-5.47	-3.28
Intestinal absorption (human) Numeric (% absorbed) (predicted by PkCSM)	100	93.586	100	100
BBB permeability Numeric (log BB) (predicted by PkCSM)	-1.944	-0.088	0.292	-0.086
hERG I inhibitor Categorical (yes/no) (predicted by PkCSM)	No	No	No	No
CYP3A4 inhibitor CYP2D6 inhibitor CYP2C19 inhibitor Categorical (yes/no) (predicted by PkCSM)	No	No	No	No
Renal OCT2 substrate Categorical (yes/no) (predicted by PkCSM)	No	No	No	No
Log Po/w (WLOGP) (predicted by SwissADME)	3.61	6.02	5.70	3.45
Bioavailability score (predicted by SwissADME)	0.55	0.55	0.55	0.55
Hepatotoxicity Neurotoxcity Cytotixicity Carcinogeneity Mutagenecity (predicted by ProTox-3.0)	Inactive	Inactive	Inactive	Inactive

The selected lead compounds were additionally screened for structural alerts using PAINS and Brenk alerts (Table S1). None of the compounds triggered PAINS alerts, indicating the absence of common assay interference motifs. BE003 and BS032 were completely free of Brenk alerts, while RG007 showed one alert related to a three-membered heterocycle and BD039 exhibited two Brenk alerts (isolated alkene and multiple esters).

Docking of neuraminidase substrate sialic acid (Neu5Ac) and selective inhibitor (1f)

In order to identify the binding orientation and interaction pattern of natural substrate sialic acid (Neu5Ac) and selective inhibitor 17f were docked to the neu1 active site (Table 4). The substrate, a sialic acid (Fig. 5B), engages NEU1’s active site through hydrogen bonds with GLU-264, TYR-370, HIS-220, and ASN-138, key residues essential for catalysis. Conversely, the inhibitor 17f (Fig. 5A) interacted with HIS-220, ASP-103, LEU-139, and ARG-280. Notably, BE003 and BS032 both exhibited favorable alignment with the substrate-binding pose. BE003 interacted with TYR-370, GLN-282, and HIS-220, while BS032 formed hydrogen bonds with GLU-264, TYR-370, and ARG-280. These interactions strongly mirror those seen in substrate docking, particularly the engagement of GLU-264 and TYR-370, which are critical for catalytic function. However, RG007 showed strong overlap with the inhibitor 17f, forming interactions with HIS-220, ASP-103, and ARG-280, which are typically involved in catalytic suppression. This binding mode is less consistent with modulatoric behavior and may indicate residual inhibitory potential. Overall, while absolute docking scores differ between substrate, inhibitor, and hits due to chemotype variation, the interaction patterns provide stronger evidence of functional similarity.

Table 4.

Binding scores and interacting residues of natural selective inhibitor (17f) and substrate of Neu1.

Compound	Binding score (Kcal/mol)	Interactions
17f Inhibitor	-6.7	Asp103, Leu139, His160, His220, Arg280
Sialic acid (Neu5Ac) substrate	-6.6	Asn138, His220, Glu264, Tyr370

Fig. 5.

Docked conformation and interaction profile of (A) natural selective inhibitor (17f) and (B) substrate of Neu1.

Molecular dynamics simulations

Root Mean Square Deviation (RMSD) is an important metric in molecular dynamics (MD) simulations that assesses the structural stability of a protein and conformational changes over time. All of the systems had RMSD values lower than 0.3 nm and the distribution probability of the simulated systems exhibited a similar pattern (Fig. 6A). For RG007, the system demonstrated stability for the first 40 ns, with a constant RMSD. However, fluctuations occurred between 40 and 80 ns, reaching 0.25 nm, indicating structural flexibility during this time. After 80 ns, the RMSD values decreased to 0.2 nm, indicating that the system stabilized and remained stable for the rest of the simulation.

Fig. 6.

Post simulation analysis of the Apo NEU1 and complexes. (A) Left: RMSD of all the systems plotted over the time of 100 ns. Right: The distribution of the RMSD values over the time of 100ns. (B) Left: RMSF of the residues with X-axis the residues numbers and Y-axis RMSF values. Right: the probability distribution of the RMSF over the time of 100 ns.

BE003 showed a small rise in RMSD after 20 ns, reaching around 0.25 nm, indicating initial structural adjustments. However, after 30 ns, the RMSD dropped to 0.2 nm and remained steady till the end of the simulation. Apo and RG007 showed the same RMSD pattern, indicating comparable stability and structural behavior. In the case of BS032 complex minor changes in BS032 were noticed between 50 and 70 ns, indicating structural dynamics. However, after 80 ns, the system converged and the RMSD stabilized, indicating that the structure had achieved a steady-state conformation. These results emphasize that all systems finally acquired stable RMSD values, suggesting structural stability. However, the fluctuations in RMSD at different time points indicate varying degrees of conformational adjustments before reaching stability.

The Root Mean Square Fluctuation (RMSF) analysis reveals residue-specific dynamics and flexibility, with notable patterns in loop regions (Fig. 6B). These fluctuations indicate that both systems maintain a degree of flexibility, particularly in loop regions. In contrast, BE003 and BS032 exhibit fewer residues with RMSF values above 0.3 nm, indicating reduced flexibility compared to Apo-NEU1 and RG007. This suggests that the binding of BE003 and BS032 stabilizes the structure to some extent, particularly outside of loop regions. The loop regions are quite flexible in all the simulated systems. Particularly, regions K161-Q165, H120-E224, V246-N261, and G378-E383 exhibit higher RMSF values in Apo-NEU1 and the three complexes. These areas show elevated RMSF values due to their inherent structural mobility. The active site of the NEU1 is surrounded by flexible loops that tend to adjust their conformation depending on the presence and orientation of the substrate. These loops undergo significant conformational changes when binding to the compounds, which might helps for effective catalysis. The flexibility of these loops is critical for the enzyme’s ability to adapt to different substrates and inhibitors^53,54. The distribution probability of RMSF values confirms this observation. Apo-NEU1 and RG007 have sharper peaks at higher RMSF ranges, reflecting the presence of more flexible regions, while in case of BE003 and BS032 high probability distribution occurs at small values of RMSF consistent with overall reduced fluctuations.

Solvent accessible surface area (SASA) and radius of gyration (Rg)

SASA is an important parameter for understanding the protein folding and solvation properties of the protein. SASA is a measure of the area of protein that is exposed to solvent. The time-dependent behavior of SASA is shown in Fig. 7A. The average SASA value of the Apo-NEU1, RG007, BE003, and BS032 were 161.627 nm², 158.187 nm², 163.35 nm², and 161.871 nm² respectively. Initially, the SASA values for the RG007 system were somewhat higher. However, at 40 ns, the SASA of RG007 steadily reduced, indicating that the protein attained a more compact structure. The SASA values of Apo-NEU1, BE003, and BS032 were consistent throughout the simulation and more stabilized after 40 ns. This implies that these complexes achieved a rather stable conformational state.

Fig. 7.

(A) solvent accessible surface area and (B) Radius of Gyration of the Apo-NEU1 and complexes.

Additionally, we also performed a radius of gyration analysis to observe the compactness of complexes. The results showed that the RG values of Apo-NEU1 and complexes remained stable throughout the simulation (Fig. 7B). The comparative Rg results revealed stable folding behavior of NEU1 after binding with compounds indicating high compactness of the NEU1 structure. The Rg values for all the systems were below 2.00(nm). The Apo-NEU1 system showed modest Rg values, similar to RG007, indicating that the unbound protein retained its compactness. After 40 ns, the Rg values in all systems were reasonably constant, indicating that the proteins had achieved a stable conformational state.

Inter-molecular hydrogen bonds

To evaluate the binding affinity of compounds towards NEU1, molecular dynamics (MD) trajectories were analyzed to monitor hydrogen bond formation between the receptor-ligand complexes. Hydrogen bonds play a crucial role in stabilizing these interactions and contribute significantly to binding strength. Compound RG007 exhibited a fluctuating number of hydrogen bonds, ranging between 0 and 3, indicating less interaction stability. Compound BE003 demonstrated a stronger interaction, forming up to 6 hydrogen bonds, suggesting enhanced binding affinity. Compound BS032 also formed hydrogen bonds ranging between 0 and 3, indicating weaker interactions. Among all compound BE003 formed the highest number of hydrogen bonds, suggesting the strongest potential for stabilizing the ligand-protein complex. These results indicate that BE003 exhibits the strongest hydrogen bonding interactions with NEU1 which may contribute to its higher binding affinity and stability within the active site. The hydrogen bond profile of the three compounds plotted over the time of 100ns is presented in Fig. 8 and Figure S6.

Fig. 8.

Time-dependent behavior of the number of hydrogen bonds between NE1 and the three compounds.

Principal component analysis (PCA)

PCA was used to analyze the conformational dynamics of protein-ligand complexes by evaluating the covariance matrix of atomic fluctuations. The trace of the covariance matrix, which reflects the total variance in the system, remained constant before and after diagonalization, demonstrating that the transformation preserved total variance (Table 5). Apo-NEU1, RG007, BE003, and BS032 have trace covariance matrix values of 161.555, 161.026, 152.246, and 177.027, respectively. BE003 causes the greatest reduction in protein flexibility with trace covariance matrix values 152.246), stabilizing the protein structure the most. RG007 has a modest decrease in trace covariance matrix values (161.026) compared to the Apo state, indicating moderate stability. BS032 enhances protein flexibility (177.027), which might contribute to a more dynamic and unstable structure. Figure 9of Principal Component Analysis (PCA) illustrates the essential motions of the protein in different simulated systems. The distribution of points in each plot represents the conformational space explored by the protein during the simulation. The apo-NEU1, BE003, and BS032 have similar V shape patterns of the data points (Fig. 9). BE003 and BS032 show moderate flexibility, positioning them between the two extremes. RG007, on the other hand, shows a different curved distribution of the data points.

Table 5.

Trace values of the covariance matrix for Apo-NEU1 and NEU1 bound with compounds.

Trace of covariance matrix	Apo	RG007	BE003	BS032
Before diagonalization	161.555	161.026	152.246	177.027
After diagonalization	161.555	161.026	152.246	177.027

Fig. 9.

Principal component analyses of Apo-NEU1 and complexes in a two-dimensional space defined by Principal Component 1 (PC1) and Principal Component 2 (PC2).

Replicate run of molecular dynamics (MD) simulation

To evaluate structural stability and reproducibility, a second 100 ns run of MD simulation was performed for each complex (Fig. 10). For all the complexes, the RMSD profiles from the second run closely overlapped with those from the first run throughout the simulation period. The similarity in the RMSD patterns indicates that the systems reached comparable conformational states in both replicates, with no major deviations between runs. These findings confirm the reproducibility of the MD simulations, ensuring that the observed stability patterns are consistent.

Fig. 10.

Replicate MD simulation run of all the candidate compounds.

MM-GBSA calculations

To evaluate the binding strength of the docked complexes the last 25 ns of the molecular dynamics (MD) trajectory was extracted for MM-GBSA analysis. The effective binding free energy of three selected compounds and substrate is presented in (Table 6). For MM-GBSA calculation, a total of 2500 frames were considered, with the interval set to every 5th frame. MM-GBSA reveals that the BE003 and BS032 complexes exhibit strong binding affinity. BE003 has the strongest binding affinity among all with a total energy of -28.51 kcal/mol. It has significant van der Waals (-31.34 kcal/mol) and electrostatic − 33.66 kcal/mol) contributions, suggesting strong hydrophobic and electrostatic interactions with NEU1. BS032 shows moderate binding affinity compared to BE003 with a total effective binding energy of -18.64 kcal/mol. A notable van der Waals contribution indicates strong hydrophobic interactions. Electrostatic interactions (-6.91 kcal/mol) are weaker than BE003 but still contribute to binding. In contrast, the RG007 complex demonstrated weak binding affinity with a total effective binding score of -3.56 kcal/mol. Overall, both binding strength and the spatial fluctuations of key residues are crucial factors affecting protein stability. Both van der Waals (-6.02 kcal/mol) and electrostatic (-2.14 kcal/mol) contributions are weak, meaning minimal interactions with the protein. In addition to the total effective binding free energy estimation, per-residue energy decomposition analysis was also performed to gain residue-level insights into ligand–protein interactions. The detailed per-residue energy decomposition for BE003 and the natural substrate are provided in the supplementary files (File S1-BE003.csv and File S2-Substarte.csv) for each frame. It includes the energy contribution of each active-site residue located within 4.5 Å of the ligand, calculated across all frames. Such decomposition allows the identification of key hotspot residues responsible for stabilizing ligand binding and highlights interaction similarities and differences between BE003 and the natural substrate.

Table 6.

MM-GBSA calculations of the three complexes. ΔEvdw: Van der Waals energy, ΔEEL: electrostatic energy, ΔEGB: Polar solvation energy, ΔESURF: non-polar solvation energy and ΔTOTAL is the total effective binding free energy.

Compound	Energy (kcal/mol)
	ΔEvdw	ΔEEL	ΔEGB	ΔESURF	ΔGGAS	ΔGSOLV	ΔTOTAL
BE003	-31.34 ± 2.64	-33.66 ± 4.75	40.45 ± 3.74	-3.97 ± 0.16	-65.00 ± 4.97	36.48 ± 3.69	-28.51± 2.62
RG007	-6.02 ± 8.03	-2.14 ± 4.08	5.41 ± 7.11	-0.81 ± 1.10	-8.16 ± 11.10	4.60 ± 6.13	-3.56 ± 5.46
BS032	-27.19 ± 4.10	-6.91 ± 4.09	19.12 ± 3.73	-3.67 ± 0.55	-34.09 ± 6.68	15.45 ± 3.37	-18.64 ± 4.18
Substrate	-25.03 ± 2.54	-14.36 ± 6.89	26.40 ± 5.88	-3.38 ± 0.40	-39.39 ± 7.21	23.02 ± 5.64	-16.37 ± 2.68

Comparison of candidate compound BE003 with substrate and inhibitor (controls)

To evaluate the functional role of BE003 in NEU1 modulation, we performed molecular dynamics (MD) simulations in comparison with a natural substrate (Neu5Ac) and a known NEU1 inhibitor (compound 17f). The RMSD plot revealed that the substrate-bound NEU1 complex maintained the lowest overall deviation, reflecting its stabilizing interaction with the catalytic pocket and minimal disruption to the protein structure. Conversely, the inhibitor complex exhibited the highest RMSD values, particularly after 40 ns, indicating significant conformational fluctuations. The BE003-bound complex showed intermediate RMSD behavior higher than the substrate but consistently lower than the inhibitor. This suggests that BE003 induces moderate structural fluctuations in NEU1, more closely resembling the behavior of the substrate rather than the inhibitor (Fig. 11A).

Fig. 11.

Post-simulation structural and dynamic comparison of NEU1 in complex with BE003, natural substrate, and inhibitor. (A) RMSD plot over 100 ns simulation, BE003 shows a dynamic profile closer to the substrate. (B) Superimposed protein structures at 100 ns showing key conformational differences among the complexes in some loop regions. Loops 98–106 and 252–261 are highlighted, indicating distinct structural rearrangements. Cyan is inhibitor-bound complex, Pink is substrate-bound complex, yellow is BE003-bound complex, and green is the initial structure.

To assess the conformational impact of ligand binding on NEU1, we performed structural superposition of the final 100 ns MD simulation snapshots for each complex: inhibitor (17f), substrate (Neu5Ac), and candidate compound BE003, along with the original apo structure. Notable deviations were observed in two flexible loop regions, residues 98–106 and 252–261, both of which contribute to the stability of the catalytic cleft. The conformational snapshots indicated that the overall structural dynamics and loop arrangements in the BE003-bound complex resembled those observed in the substrate-bound state rather than the inhibited form. These findings suggest that BE003 displays a functional behavior more similar to the natural substrate, providing supportive computational evidence of its potential activating role on NEU1 (Fig. 11B).

Discussion

AD is a progressive neurological disorder marked by cognitive decline and memory loss⁵⁵. Recent research has emphasized the significance of NEU1 (neuraminidase 1) in neuroinflammation and lysosomal dysfunction, both of which are major pathogenic hallmarks of AD. Targeting NEU1 with modulators is a viable therapeutic method for restoring lysosomal function and slowing neurodegeneration⁵⁶. In this study, we identified sea metabolites as NEU1 modulators, indicating their potential to modulate disease-related pathways.

For this purpose library of sea weed metabolites was retrieved and optimized to find potential modulator for NEU1. Seaweed-derived metabolites, rich in bioactive compounds like polysaccharides, phenolic, and alkaloids, are being explored for their potential in AD therapeutics due to their ability to modulate key neurodegeneration pathways. High throughput virtual screening (HTVS) was performed using Autodock Vina against NEU1 (File S3-S4). The selection of potential NEU1 modulators was carried out through a systematic screening process. Initially, compounds with the strongest binding affinity were shortlisted for further evaluation. The compounds with the strong affinity were picked and further assessed based on Lipinski’s rule of five. Out of twenty compounds only four compounds RG007, BE003, BS032, and BD032 fulfilled this screening criterion. Subsequently ADMET properties were calculated of these four compounds. Out of these compounds, three compounds RG007, BE003, and BS032 exhibited good solubility, toxicity profile, and BBB permeability, which is crucial for potential neuroprotective therapeutics targeting AD. The selected compounds formed significant hydrogen bond interactions with NEU1 amino acid residuesArg280, Asp103, Glu264, Arg341, and Tyr370. These residues are critical for substrate recognition and catalytic activity, making them essential for NEU1 enzymatic function⁵⁷. The ability of these compounds to interact with these critical residues implies that they may be beneficial in modulating NEU1 function. All selected compounds were deeply embedded within the active site groove of NEU1, exhibiting strong binding affinities. To evaluate the stability and binding affinity, these three final compounds were validated in silico using MD simulations and MM-GBSA (File S5-S7). These computational tools assisted in determining the stability of compound-NEU1 interactions over time and gave insights into binding free energy, validating their potential as effective NEU1 modulators. The RMSD and RMSF analyses were performed to evaluate the stability and flexibility of the NEU1-metabolite complexes during MD simulations. The RMSD values of all systems remained below 0.3 nm throughout the simulations, indicating that the complexes were structurally stable with minimal deviations from their initial conformations. The RMSF analysis further provided insights into the flexibility of different regions within the NEU1 structure. The results revealed that the loop regions of NEU1 exhibited noticeable fluctuations, indicating their inherent flexibility. This is consistent with previous studies, which have reported that NEU1’s loops and active site possess a dynamic nature, allowing the enzyme to undergo conformational changes upon substrate binding. These structural adaptations are crucial for the enzyme’s catalytic function, as they facilitate substrate recognition and proper positioning within the active site^53,54. Additionally, Solvent Accessible Surface Area (SASA) and Radius of Gyration (Rg) analyses were conducted to assess the compactness and solvent exposure of the protein-ligand complexes. Both SASA and Rg values remained stable throughout the simulation, further confirming the overall structural integrity of NEU1 in the complex with the selected compounds. These conformational changes are essential for the enzyme’s ability to perform its function. Understanding these dynamics can aid in the development of more potent modulators. Hydrogen bond analysis (Fig. 7) that among all selected compounds, BE003 exhibited the highest number of hydrogen bonds with NEU1. A larger number of hydrogen bonds indicates stronger and more persistent interactions, suggesting that BE003 has a high affinity for the NEU1. Further validation was carried out utilizing MM-GBSA analysis, which yields an estimate of the binding free energy. The results showed that among the three compounds, only BE003 had a considerably higher binding affinity, with a binding energy score of -28.51 kcal/mol. These results are consistent with the hydrogen bond analysis and other simulation data, indicating that BE003 is the most stable and powerful NEU1 modulator of the three chemicals tested. Moreover the post simulation structural dynamics analysis revealed a distinct transition of key catalytic loop regions into an open conformation upon BE003 binding (Fig. 12). This structural transition is functionally significant, as the open conformation is associated with enhanced substrate accessibility and catalytic activation in sialidase like NEU1. Such a conformational shift suggests that BE003 may not merely bind passively but could actively stabilize the catalytically competent state of the enzyme further supporting its classification as a putative NEU1 modulator^57,58.

Fig. 12.

Superimposition of the starting structure of BE003 complex and final structure (last frame) extracted at 100ns. The starting structure is shown as green and final structure is shown as pink. The final (simulated) structure shows several loops, especially around the active site, in an open conformation, while the starting structure remains more compact.

Further assessment was carried out to more rigorously evaluate BE003’s role as a modulator. For this purpose, a natural NEU1 substrate and a selective NEU1 inhibitor were used as reference controls. Comparative docking revealed that BE003’s interaction profile more closely resembled that of the substrate than the inhibitor. MD RMSD analysis confirmed similar stability trends to the substrate, particularly in maintaining an open catalytic loop conformation, whereas the inhibitor-bound complex showed greater conformational restriction. However, MM-GBSA binding energy analysis demonstrated that BE003 did not fully replicate the substrate’s energetic profile, indicating only partial substrate-like engagement. BE003, as a putative NEU1 modulator, can activate the enzyme and restore its activity. NEU1 activation is thought to improve lysosomal degradation, reduce Aβ buildup, and diminish neuroinflammatory responses, all of which help to delay the course of AD. The findings of this study provide a foundation for further experimental validation of BE003 as a possible therapeutic modulator for AD. However, it is important to emphasize that the findings of this study are solely based on computational predictions. The structural of NEU1 used for docking was obtained from AlphaFold. Although generally highly accurate, AlphaFold model may omit important dynamic conformations, potentially affecting the reliability of predicted binding interactions. Further experimental validation is required to confirm the predicted binding interactions, the proposed mechanism of NEU1 activation, and the potential therapeutic effects of BE003. A limitation of the present study is that all docking and molecular dynamics simulations were performed at neutral pH (7.0), whereas NEU1 is a lysosomal enzyme whose physiological microenvironment is acidic, with a pH of approximately 4.5–5.0. Such conditions can influence the protonation states of key active-site residues, potentially affecting ligand binding and catalytic activity. The use of neutral pH was chosen to maintain consistency with standard docking and MD protocols, which are well-validated under these conditions, and to ensure computational robustness and comparability across simulations. However, it is acknowledged that this may not fully capture lysosomal microstates. In future, pH-aware protonation schemes (e.g., constant-pH MD or residue-specific adjustments) will be incorporated to better represent the lysosomal environment. This extension will enable a more realistic evaluation of ligand binding poses and binding free energies under physiologically relevant conditions, thereby enhancing the biological relevance of the findings.

Conclusion

The purpose of this study was to identify the modulator to modulate the activity of NEU1. High throughput screening, rule of five violation, and ADMET screening ruled out four potential metabolites BE003, RG007, BD039, and BS032. Subsequently, MD simulations and MM-GBSA identified BE003 as a potential NEU1 modulator due to its high binding affinity, stable interactions, and strong hydrogen bonding network across molecular dynamics simulations. Further the observed partial substrate-like engagement further supports its classification as a modulator. Such modulation may help restore NEU1 function, enhance lysosomal degradation, and ultimately mitigate Alzheimer’s disease pathology. These findings offer a solid foundation for further in vitro and in vivo validation of BE003 as a possible therapeutic option for targeting NEU1 in AD treatment.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

S.A. performed all the analysis and wrote the manuscript.

Funding

This study was funded by King Salman center For Disability Research Research Group no KSRG-2024-340.

Data availability

Data is provided within the manuscript or supplementary information files. All raw data is publicly available at Zenodo: https://doi.org/10.5281/zenodo.16796929.

Declarations

Competing interests

The authors declare no competing interests.