Introduction

Neurodegenerative diseases are characterized by the aggregation of amyloid fibrils in specific regions of the brain, contributing to more than 30 human diseases, including Alzheimer’s disease (AD), Parkinson’s disease and type 2 diabetes1,2,3,4,5. Among these, AD is the most prevalent neurodegenerative disorder and is distinguished by the accumulation of amyloid fibril plaques along with numerous neurofibrillary tangles composed of filaments formed from highly phosphorylated microtubule-associated Tau proteins6. The aggregation of fibrillar plaques, which represents a critical event in the onset and progression of disease, is driven by the formation of highly ordered b-sheet-rich structures7,8. Amyloid-b (Aβ) peptides are generated through the proteolytic cleavage of the amyloid precursor protein by b- and g-secretases. The wild-type human Aβ (1–42) monomer sequence is D1AEFRHDSGYEVHHQKLVFFAEDVGSNKGAIIGLMVGGVVIA42, which contains two helical regions of amino acids 8 − 25 and 28 − 38 connected by a kink. While monomeric Aβxhibits minimal neurotoxicity, its aggregation into b-sheet-rich oligomers and fibrils is associated with synaptic dysfunction, mitochondrial impairment, and neuronal apoptosis through mechanisms involving membrane permeabilization and oxidative stress9.

Growing evidence suggests a direct association between the deposition of Aβ aggregates (such as dimers, trimers, and 12-mers oligomers) in the brains of AD patients and disease progression10. Consequently, there is significant interest in identifying small organic molecules or peptides capable of either preventing fibril formation or enabling early-stage detection of amyloid aggregation11,12,13,14. Alzhemed (tramiprosate) was one of the drugs found to bind to Aβ monomers and inhibit fibril formation, entering Phase III clinical trials. However, despite its initial success in fibril inhibition, Alzhemed failed to demonstrate cognitive improvement in AD patients, ultimately resulting in the discontinuation of its development. Another compound, ELND005 (also known as scyllo-inositol), has shown promise in inhibiting Aβ aggregation and is currently undergoing Phase III clinical trials. However, the efficacy of these compounds in altering the course of AD remains limited, underscoring the need for more effective therapeutic strategies. This highlights the need for atomic-resolution insights into Aβ-inhibitor interactions. Recent advances in structural biology, including cryo-electron microscopy (cryo-EM) studies of Aβ42 fibrils15, solid-state NMR characterization of oligomer interfaces16, and single-molecule force spectroscopy of assembly kinetics, have revolutionized our understanding of amyloid genesis. These studies reveal that effective inhibition requires compounds to either stabilize non-toxic monomeric conformations or disrupt critical b-strand interactions within oligomeric nuclei, offering valuable guidance for the rational design of novel therapeutic agents.

Natural products have emerged as a promising source of compounds for inhibiting Aβ aggregation. Among these, flavonoids represent a structurally privileged scaffold in amyloid inhibition, combining p-p stacking capacity with hydrogen-bonding functionality. Epigallocatechin gallate17, a major polyphenolic compound found in green tea, has been shown to effectively inhibit the formation of toxic Aβ oligomers and fibrils, while also mitigating the neurotoxic effects of Aβ aggregates. Structure-activity relationship studies by Choi et al.18 demonstrated that amentoflavone (AMF) exhibit significantly greater inhibition compared to their monomeric analogs. However, the molecular mechanisms underlying the inhibitory effects of AMF remain incompletely understood. In particular, the precise structural features responsible for their activity, as well as their specific binding interactions with Aβ monomers and aggregates, require further elucidation to enable rational drug design.

Experimental approaches, while valuable, are often time-consuming and resource-intensive, highlighting the need for more efficient computational tools to guide the identification and optimization of lead compounds. In this context, computational chemistry simulations offer an indispensable approach for investigating the molecular interactions between Aβ and potential inhibitors at the atomic level19,20,21,22. Chong and Ham23 utilized classical molecular dynamics (MD) simulations to investigate Aβ42 dimerization, revealing that water-driven intermolecular interactions serve as key stabilization factors. While classical MD simulations are effective for probing amyloid monomer/dimer conformations, their limited conformational sampling capacity may restrict a comprehensive assessment of aggregation mechanisms. To overcome these limitations, enhanced sampling methods such as replica exchange molecular dynamics (REMD)24 have been widely employed. REMD has emerged as a powerful tool for mapping the energetic landscape of amyloid aggregation25,26. Chakraborty and Das27 applied REMD simulations to investigate the effects of an N-terminal hexapeptide amyloid inhibitor on Aβ42 monomer conformations, providing key insights into its inhibitory mechanism. These computational approaches not only offer deeper insights into the molecular mechanisms of Aβ aggregation inhibition but also enable the rational design of new molecules with improved therapeutic potential.

In this study, we employed REMD simulations to elucidate the microscopic mechanisms by which the natural product AMF inhibits early-stage Aβ oligomerization. Using multiple computational strategies, we systematically characterized AMF’s binding modes, and its ability to disrupt b-sheet formation—a structural hallmark of neurotoxic Aβ assemblies. Detailed energetic analyses revealed key binding hotspots, particularly within the hydrophobic core region 16KLVFFAEDV24, and demonstrated that AMF preferentially stabilizes disordered, non-b-sheet conformations that are less prone to self-assembly. Guided by these mechanistic insights, we further conducted high-throughput virtual screening and identified two candidate molecules with superior predicted binding affinities compared to AMF. Figure 1 shows a schematic of the workflow.

Fig. 1
figure 1

Schematic overview of the computational workflow used to investigate the inhibitory mechanism of the natural product AMF on Aβ oligomerization.

Full size image

Materials and methods

Molecular modeling

The initial structure of Aβ (1–42) [Protein Database Bank (PDB) ID: 1IYT28 were obtained from the PDB29. To study protein aggregation, it is common practice to perform high-temperature simulated annealing on the crystal structure to obtain a fully extended coil conformation30 (Fig. 2A). The 2D and 3D structures of the AMF molecule are shown in Fig. 2B. The geometry of AMF was optimized using the CP2K software package (version 2024.1; https://www.cp2k.org.)31, employing the M06-2X functional32. The valence electron wavefunction was expanded in a double-z plus polarization basis set33, and Goedecker–Teter–Hutter (GTH)34,35 pseudopod potentials for core electrons were applied. The Aβ protein was described using the CHARMM36m force field36, while water molecules were modeled with the TIP3P water model37. Three extended Aβ monomers were randomly placed in a cubic water box with dimensions of 80 Å × 80 Å × 80 Å, forming the Aβ-trimer system. To maintain charge neutrality, sodium ions were introduced into the system. Periodic boundary conditions were applied, with a 14 Å cutoff for non-bonded interactions. Long-range electrostatic interactions were treated using the particle-mesh Ewald (PME) algorithm38, and the SHAKE algorithm39 was used to constrain all hydrogen atoms to their equilibrium positions.

To better align with the AMF: Aβ molar ratio reported in reference18, three AMF molecules were randomly distributed within the simulation box, defining the low-concentration system (Aβ-trimer-AMF-Low). The corresponding initial configuration is shown in Fig. 2. To assess potential concentration-dependent effects, a high-concentration system (Aβ-trimer-AMF-High) was also constructed by introducing seven AMF molecules. While this setup does not correspond to a specific experimental condition, it enables the evaluation of AMF–Aβ interactions under elevated ligand ratios, thereby facilitating a mechanistic understanding of potential saturation effects. The force field parameters for AMF were derived from the CGenFF force field40.

Fig. 2
figure 2

(A) The crystal structure and extend coil conform for Aβ. (B) 2D and 3D structure of AMF. (C) Aβ-trimer system. (D) Aβ-trimer-AMF-Low system. (E) Aβ -trimer-AMF-High system.

Full size image

REMD simulations

To investigate the impact of AMF concentration on Aβ aggregation, each system underwent a systematic preparation protocol before being subjected to 200 ns of all-atom REMD simulations. Initially, energy minimization was performed using the steepest descent algorithm to remove steric clashes and ensure a physically stable starting structure. This was followed by a 500 ps equilibration phase in the canonical (NVT) ensemble at 300 K, allowing the system to reach thermal equilibrium. Subsequently, an additional 500 ps equilibration in the isothermal-isobaric (NPT) ensemble was conducted to stabilize pressure and density fluctuations. After equilibration, REMD simulations were performed using the GROMACS 2022.5 software package41, ensuring enhanced conformational sampling of Aβ aggregation dynamics. Each system consisted of 40 replicas, with temperatures logarithmically distributed between 310 K and 389 K to optimize the efficiency of temperature exchanges while maintaining sufficient overlap between adjacent replicas. The 40-temperature distribution used in the REMD simulations (Table S1) was generated according to the algorithm proposed by Patriksson42, which ensures approximately uniform exchange probabilities across adjacent replicas. Each replica was run for 200 ns, and the final 100 ns were selected for analysis. This setup strikes a balance between exhaustive sampling and computational feasibility, facilitating a comprehensive exploration of AMF’s inhibitory effects on Aβ aggregation. The exchange attempt time was set to 2 ps, yielding an acceptance ratio of approximately 25%.

Analysis methods

The binding free energies of Aβ with AMF were calculated using the molecular mechanics/Poisson–Boltzmann surface area (MM/PBSA) method, as implemented in the GROMACS 2022.5 package. MM/PBSA provides a computationally efficient approach to estimate binding affinities by combining molecular mechanics energies with solvation contributions from both polar (Poisson–Boltzmann equation) and nonpolar (solvent-accessible surface area) components. MM-PBSA is one of the efficient ways to quantitatively estimate the host–guest interactions. The methodological details of this approach have been described extensively in our previous study43. To gain deeper insights into the conformational dynamics of Aβ and its interactions with AMF, the simulation trajectories were analyzed using a range of GROMACS built-in tools. The secondary structure evolution was characterized with gmx dssp, providing insights into b-sheet disruption and conformational transitions.

Results

Aggregation behavior of the Aβ-trimer system

To assess the convergence of the REMD simulations, we calculated the standard deviations of several key structural parameters over three consecutive time intervals (100–133ns, 133–166 ns and 166–200 ns) as shown in Figure S1. Consistent results across these windows suggest that the system had reached a stable and representative sampling regime. While the standard deviation of the coil probability was relatively large, possibly reflecting the intrinsic flexibility, the deviations observed for other secondary structure remained within an acceptable range. We also evaluated the convergence of the REMD simulations by tracking the time evolution of temperature exchange for a representative replica in temperature space. As illustrated in Figure S2, the representative replicas of the Aβ-trimer, Aβ-trimer-AMF-Low and Aβ-trimer-AMF-High systems traversed the entire temperature range multiple times during the 200 ns simulation, indicating that the replicas were not trapped in one single temperature. Taken together, these results support that REMD simulations achieved reasonable convergence within 200 ns.

Aβ trimers are highly toxic intermediates in amyloid aggregation, and understanding their structural and dynamic properties is essential for the rational design of inhibitors that can mitigate their neurotoxicity44. REMD simulations of the Aβ-trimer system reveal a pronounced tendency for the peptides to transition from an initially fully extended coil conformation to compact structures enriched in b-sheet elements. During the early stages of the simulation, extended monomeric states rapidly reorganize into stable oligomeric assemblies, characterized by well-defined b-sheet motifs. Secondary structure analysis using the DSSP algorithm demonstrates a progressive increase in b-sheet content over time, with the most significant formation occurring within the 3EFRHDSG9, 16KLVFFAEDV24, 31IIGLMVG37, and 40VIA42 sequences (Fig. 3). Notably, the 16KLVFFAEDV24 region exhibits the highest propensity for b-sheet formation, consistent with previous studies identifying this hydrophobic core as a key nucleation site for Aβ aggregation45.

Fig. 3
figure 3

Averaged probabilities of secondary structure for each residue in Aβ-trimer system during the REMD simulations.

Full size image

To quantitatively characterize the structural evolution of Aβ trimers, we monitored the temporal progression of b-sheet content throughout the simulations. The b-sheet fraction increased from an initial value of 0% in the fully extended state to 13.6 ± 1.8% as the system approached equilibrium, indicating a spontaneous transition toward aggregation-prone conformations. To further dissect the conformational landscape associated with Aβ trimerization, we constructed free energy surfaces using principal component analysis derived from the simulation trajectories. Cluster analysis revealed several predominant conformational states, with the most populated clusters exhibiting extensive inter-strand hydrogen bonding networks. Structurally, the Aβ trimer predominantly adopts either a three-stranded antiparallel b-sheet or a two-layered architecture with an antiparallel arrangement (Fig. 4), stabilized by key turn regions at residues 12 − 14, 25 − 27, and 38 − 40. These findings provide atomic-level insights into the early-stage oligomerization of Aβ The observed structural patterns are consistent with experimental solid-state NMR data46, reinforcing the reliability of our computational approach in capturing the fundamental mechanisms underlying Aβ assembly.

Fig. 4
figure 4

Typical cluster conformation of Aβ-trimer.

Full size image

Inhibitory effects of AMF on Aβ aggregation

To elucidate how AMF perturbs the early aggregation behavior of Aβ oligomers, we performed REMD simulations using the same protocol as the Aβ-trimer control system, but in the presence of either low (three AMF molecules) or high (seven AMF molecules) concentrations of AMF. Comparative analysis of secondary structure propensities reveals that AMF exerts a marked inhibitory effect on b-sheet formation. In the absence of AMF, Aβ trimers undergo a spontaneous transition from disordered coil states to b-sheet-rich conformations, achieving a final b-sheet content of approximately 13.6 ± 1.8% (Fig. 3). By contrast, systems containing AMF exhibit a dramatic suppression of b-sheet content, reduced to 4.9 ± 0.9% and 4.7 ± 0.8% in the low- and high-concentration systems, respectively (Figs. 5 and 6). Notably, the additional AMF molecules in the high-concentration system do not further decrease b-sheet formation significantly, suggesting that a saturation threshold may exist for AMF’s inhibitory effect. Interestingly, the presence of AMF also induces a distinct structural redistribution within the Aβ-trimer ensemble. Whereas helix conformations are virtually absent in the AMF-free system, the introduction of AMF promotes helix formation, with the helical content increasing alongside AMF concentration as shown in Fig. 6. This shift suggests that AMF not only prevents b-sheet formation but also stabilizes alternative, non-aggregative conformations, possibly by imposing local torsional constraints or shielding key backbone hydrogen bond.

Fig. 5
figure 5

Averaged probabilities of secondary structure for each residue in Aβ-trimer-AMF-Low system during the REMD simulations.

Full size image

Representative conformations extracted from the REMD simulations clearly demonstrate that in the Aβ - trimer system, the peptides predominantly adopt b-sheet-rich structures (Fig. 4). However, upon introducing AMF molecules—whether at low concentration (three AMF molecules, Aβ-trimer-AMF-Low system) or high concentration (seven AMF molecules, Aβ-trimer-AMF-High system)—the conformational ensemble shifts significantly, with the Aβ-trimer adopting predominantly disordered coil-like structures. Only small, isolated segments of b-sheet structure persist, as shown in Fig. 7A and B. These findings indicate that AMF interacts with key aggregation-prone regions of Aβ disrupting intermolecular interactions essential for b-sheet stabilization.

To further elucidate the molecular interactions underlying the aggregation behavior, we constructed both inter-chain and intra-chain residue–residue contact maps averaged over the final 100 ns of the REMD simulations, as shown in Fig. 8. This approach is consistent with methodologies employed in recent studies47,48,49. For the Aβ system, prominent inter-chain contacts were observed between residues in the regions 16KLVFFAEDV24 and 31IIGLMVGGVVIA42, suggesting the presence of stable intermolecular b-sheet interactions that contribute to fibril formation. In addition, strong intra-chain contacts were also detected within certain segment. Upon the introduction of low concentrations of AMF, a general attenuation of intra-chain interactions was observed, accompanied by a pronounced disruption of inter-chain contacts specifically within the 16KLVFFAEDV24 region. This region is known to play a critical role in b-sheet stacking and fibril formation. At higher concentrations of AMF, both intra- and inter-chain residue contacts were markedly diminished, and the interaction patterns became more diffuse. These findings indicate that AMF molecules infiltrate the Aβ oligomeric assemblies and effectively interfere with key residue–residue contacts, thereby disrupting the structural integrity required for stable aggregation.

Fig. 6
figure 6

Averaged probabilities of secondary structure for each residue in Aβ-trimer-AMF-High system during the REMD simulations.

Full size image

Given the pronounced structural disruption observed in the Aβ-trimer-AMF-Low system, we focus our subsequent analyses on this system to gain deeper insights into the molecular mechanisms by which AMF inhibits Aβ aggregation. To further elucidate the molecular basis underlying the inhibitory effects of AMF on Aβ aggregation, performed extensive 200 ns all-atom MD simulations on the dominant conformational clusters (Cluster-1) extracted from both the Aβ-trimer system and the Aβ-trimer-AMF-Low system for subsequent MM/PBSA analysis. This approach enabled us to quantify the binding free energies and decompose the contributions from the polar (△Eele+△Gsol, GB)and non-polar (△EvdW+△Gsol, np) terms, as summarized in Table 1.

To further elucidate the binding preference of AMF toward the Aβ peptide, we calculated the contact frequencies between AMF and individual residues of full-length Aβ. This analysis provides insights into the specific regions where AMF preferentially associates. As shown in Figure S3, AMF exhibits the highest contact probability with residues within the 16KLVFFAEDV24 region. Notably, the non-polar contribution to binding energy (△Enon−polar) is substantially more favorable when AMF interacts with 16KLVFFAEDV24 (-49.19 ± 2.28 kJ/mol) compared to the self-association of 16KLVFFAEDV24 (-24.55 ± 1.76 kJ/mol). This nearly twofold increase highlights the strong hydrophobic interactions between AMF and the core aggregation motif of Aβ Given that 16KLVFFAEDV24 is a central hydrophobic cluster critical for b-sheet formation, the preferential binding of AMF suggests that it effectively shields this segment from participating in interpeptide interactions necessary for fibrillization. The disruption of these hydrophobic contacts is a well-established strategy for amyloid inhibition, reinforcing the potency of AMF as an aggregation suppressor.

Table 1 Calculated binding energies obtained using the MM/PBSA method.
Full size table
Fig. 7
figure 7

Representative conformations of the top five most-populated clusters (A) Aβ-trimer-AMF-Low and (B) Aβ-trimer-AMF-High systems.

Full size image
Fig. 8
figure 8

The intra- and inter-peptide residue-residue contact probability maps for the Aβ trimer in the absence and presence of AMFs (low and high concentration).

Full size image

Further decomposition of the binding energy reveals that Leu-17 (-4.15 ± 0.63 kJ/mol), Phe-20 (-4.67 ± 0.94 kJ/mol), and Val-24 (-3.60 ± 0.45 kJ/mol) contribute most significantly to the interaction energy. These residues are located within the hydrophobic core of Aβ and are essential for stabilizing b-sheet formation in amyloid fibrils. The preferential binding of AMF to these sites strongly suggests that AMF disrupts key aggregation hotspots, sterically and energetically impeding the formation of stable interpeptide contacts. The structural interactions between AMF and these critical residues were further analyzed, and representative binding configurations extracted from the final equilibrated structures are shown in Fig. 9A. These snapshots provide direct structural evidence of AMF engaging with the 16KLVFFAEDV24 sequence, supporting the computational binding energy results. The electrostatic contribution to binding energy (ΔEelectrostatic) is positive in both cases but significantly higher in the AMF-KLVFFAEDV interaction (17.19 ± 0.71 kJ/mol) compared to KLVFFAEDV self-association (3.24 ± 0.46 kJ/mol). This suggests that while AMF engages in strong hydrophobic interactions, its electrostatic contributions introduce a degree of destabilization. The increased positive electrostatic energy could arise from partial desolvation effects or repulsive interactions within the AMF-Aβ complex.

Fig. 9
figure 9

(A) Representative binding conformations of AMF interacting with the 16KLVFFAEDV24 region of Aβ, extracted from the final equilibrated structures of the Aβ-trimer-AMF-Low system. Key interacting residues, Leu-17, Phe-20, and Val-24, are highlighted. (B) Representative two-layered architecture formed by the 16KLVFFAEDV24 region of Aβ, exhibiting an antiparallel b-sheet arrangement.

Full size image

The overall binding energy (ΔEbinding) further underscores AMF’s inhibitory role, with AMF binding to 16KLVFFAEDV24 yielding a more favorable interaction (-32.00 ± 3.29 kJ/mol) than 16KLVFFAEDV24 self-association (-21.31 ± 2.89 kJ/mol). This suggests that AMF effectively outcompetes native Aβ-Aβ interactions, thereby preventing b-sheet formation. Given that hydrophobic interactions are the dominant force in amyloid aggregation, AMF’s preferential targeting of key hydrophobic residues may provide a mechanistic basis for its inhibitory effect. Collectively, these results provide compelling evidence that AMF disrupts Aβ aggregation by binding preferentially to the 16KLVFFAEDV24 region, which serves as a nucleation site for fibrillization. The strong hydrophobic interactions, particularly with Leu-17, Phe-20, and Val-24, prevent 16KLVFFAEDV24 from engaging in interpeptide b-sheet formation (Fig. 9B). These findings are consistent with our REMD results analyses, which indicate that AMF maintains Aβ in a more dynamic and disordered state, effectively preventing the transition into aggregation-prone conformations.

Virtual screening of natural compounds

To complement our mechanistic findings on the inhibitory effects of AMF and to explore additional potential aggregation inhibitors, we performed a structure-based virtual screening targeting the key aggregation-prone region 16KLVFFAEDV24. This segment, identified through our REMD and MM/PBSA analyses, plays a central role in mediating hydrophobic interpeptide contacts and b-sheet stabilization in early-stage Aβ oligomerization.

For the virtual screening, we employed AutoDock Vina50, a widely used molecular docking engine known for its speed and accuracy in high-throughput ligand-receptor interaction prediction. The docking pocket was constructed based on the representative b-sheet conformation of the 16KLVFFAEDV24 segment extracted from the Aβ-trimer system. As a compound library, we selected the Taiwan Traditional Chinese Medicine (TCM) Database51, which comprises over 20,000 natural compounds derived from 453 herbal species. This database has been increasingly recognized as a valuable source of bioactive small molecules for drug discovery, including anti-amyloid and neuroprotective applications. Its structural diversity and pharmacological relevance make it particularly suitable for identifying lead-like scaffolds capable of modulating amyloidogenic protein-protein interactions.

Fig. 10
figure 10

Chemical structures of the top two natural products identified through structure-based virtual screening. Representative binding pose of the two candidates with the Aβ trimer.

Full size image

AMF was used as a reference compound in our molecular docking campaign, exhibiting a docking score of -41.92 kJ/mol. Following virtual screening, all candidate molecules were ranked based on their predicted binding affinities. Among the top hits, two compounds demonstrated stronger binding than AMF: a flavonoid-like natural product (ZINC000299817537) and ligustroflavone (ZINC000169724085). These two molecules, with docking scores of -45.27 kJ/mol and − 43.05 kJ/mol respectively, were the top-ranking hits in the screened library, highlighting their theoretical potential to interfere with Aβ aggregation more effectively than AMF (Fig. 10).

A detailed analysis of the binding modes showed that both compounds occupy a hydrophobic pocket similar to that of AMF, primarily interacting with residues Leu-17, Phe-20, and Val-24 within the Aβ trimer. These interactions involve extensive hydrophobic contacts, which are recognized as key contributors to the stabilization of b-sheet-rich oligomeric structures. The ligands share common pharmacophoric features, including planar aromatic scaffolds and multiple hydroxyl groups, which suggest that they may inhibit aggregation through a mechanism comparable to that of AMF. In addition to these hydrophobic interactions, both candidates also form hydrogen bonds and potential electrostatic interactions with Arg-5, facilitated by their hydroxyl-rich functionalities. This supplementary interaction network may further enhance binding stability and broaden their inhibitory effect by interfering with additional aggregation-prone regions. Together, these findings support the potential of ZINC000299817537 and ZINC000169724085 as promising inhibitors of Aβ self-assembly.

While natural small molecules have been widely studied for their ability to inhibit amyloid aggregation through direct interactions with hydrophobic cores or b-sheet motifs, they represent only one facet of a broader therapeutic landscape. Beyond natural small molecules, recent studies have highlighted alternative inhibitors with distinct mechanisms and therapeutic potential. Peptide-based inhibitors and nanoparticles have been developed to tau protein and Aβ aggregation at various stages52,53. Moreover, emerging insights into oligomer formation via liquid–liquid phase separation and b-barrel intermediates suggest novel intervention points for early-stage amyloid toxicity54. These findings collectively expand the landscape of anti-aggregation strategies.

Conclusions

In this study, our integrated computational investigation elucidates the molecular mechanism by which AMF inhibits Aβ aggregation. By combining REMD, all-atom MD simulations, and MM/PBSA energy decomposition analyses, we demonstrate that AMF preferentially targets the critical 16KLVFFAEDV24 segment—a key nucleation site in Aβ fibrillization. AMF binding disrupts the hydrophobic interactions that normally drive the formation of ordered b-sheet structures by engaging crucial residues such as Leu-17, Phe-20, and Val-24. This competitive binding not only outcompetes native Aβ–Aβ interactions but also stabilizes the peptide in a predominantly disordered, non-fibrillogenic state, as evidenced by a significant reduction in b-sheet content. The more favorable overall binding energy observed for the AMF–Aβ interaction further underscores AMF’s capacity to impede the structural transitions essential for toxic oligomer and fibril formation. These findings provide atomic-level insights into the early events of Aβ aggregation and establish a robust mechanistic framework for the design of next-generation inhibitors targeting amyloid pathology. Building on this mechanistic understanding, we performed a structure-based virtual screening campaign targeting the same motif and identified two promising small molecules from the TCM compound database. Future studies incorporating advanced machine learning techniques and experimental validation will be critical to refine these insights and translate them into therapeutic strategies for Alzheimer’s disease.