The researchers began with Jowiseungchungtang, or JWS, a formulation containing 14 medicinal herbs. A search of public databases identified 2,253 entries associated with those herbs. Removal of duplicates left 1,751 unique molecular structures, represented by Simplified Molecular Input Line Entry System strings obtained from PubChem and TM-MC 2.0. The library was then filtered using two simultaneous criteria: compliance with Lipinski’s rule of five for basic oral drug-likeness and predicted blood-brain barrier permeability under the BOILED-Egg model. These filters excluded 1,185 compounds and retained 566 candidates considered suitable for additional analysis.

Molecular docking assessed whether the 566 retained compounds might interact with aggregation-prone regions of amyloid-beta, or Aβ. The researchers used three Protein Data Bank structures: the alpha-helical monomer 1Z0Q, the beta-sheet-rich monomer 2NAO and the pentameric assembly 2BEG. Compounds were ranked separately for each structure, and the top 35 from every model were compared. Thirteen compounds consistently ranked within those groups, after which five were prioritized because their predicted poses contacted one or more central hydrophobic residues—Leu17, Val18, Phe19, Phe20 and Ala21. The five candidates advanced to testing were 1-monopalmitin, isobutyl octyl oxalate, pentadecanoic acid, dimethyl azelate and dibutyl sebacate.

The predicted binding patterns differed among the structural states. Against the alpha-helical monomer, 1-monopalmitin had a CDOCKER energy of −39.3 kilocalories per mole and was predicted to form hydrogen bonds with Phe19 and Asp23 alongside hydrophobic contacts involving His14 and Val18. In the beta-sheet-rich monomer, pentadecanoic acid had the most favorable score among the four compounds later found to inhibit aggregation, at −30.6 kilocalories per mole. In the pentameric model, hydrophobic contacts were more prominent than hydrogen bonds, with interactions involving residues including Leu17, Phe19, Val40 and Ala42. These values are computational pose-ranking scores, not experimentally measured binding affinities.

Thioflavin T fluorescence assays supplied the principal experimental validation for the computational predictions. Synthetic Aβ42 was incubated for 48 hours at 37 degrees Celsius with each candidate at 2.5, 25 or 250 micromolar; morin at 100 micromolar served as a positive control. Four compounds significantly reduced fluorescence associated with Aβ42 aggregation. The concentration-dependent effect was clearest for 1-monopalmitin, isobutyl octyl oxalate and dimethyl azelate. At 250 micromolar, they reduced fluorescence to 67.08%, 73.03% and 80.68% of the vehicle level, respectively. Pentadecanoic acid significantly inhibited aggregation only at 25 micromolar, where fluorescence fell to 85.63% of vehicle. Dibutyl sebacate showed no significant inhibition at any tested concentration.

A second assay examined whether the candidates could reduce aggregates that had already formed. Researchers pre-aggregated Aβ42 for 24 hours and then incubated the resulting fibrils with each compound for another 48 hours. Isobutyl octyl oxalate significantly reduced fluorescence from preformed aggregates at 25 micromolar, to 60.10% of the vehicle level. Pentadecanoic acid produced a larger reduction at 250 micromolar, reaching 25.59% of vehicle fluorescence. Neither 1-monopalmitin nor dimethyl azelate significantly disaggregated preformed aggregates under the tested conditions, and dibutyl sebacate again showed no significant effect. Crude JWS extract also reduced fluorescence, consistent with activity arising from multiple constituents, although the extract was more potent than any tested individual compound.

The four compounds with anti-aggregation activity were examined in atomistic molecular dynamics simulations of an Aβ17–42 dimer derived from 2BEG. The researchers ran triplicate 10,000-picosecond simulations in explicit water with 0.145 molar sodium chloride for the ligand-free dimer and each ligand-bound complex. Representative trajectories showed interface rearrangement, partial chain separation, bending, twisting and expansion of coil or turn segments in several compound-bound systems. Ligands frequently changed position and orientation, indicating transient contacts rather than persistent stabilization of a single bound pose. Because the simulations covered only 10 nanoseconds, they could not sample rare unbinding events, higher-order oligomerization or a complete pathway of fibril disassembly.

Backbone root mean square deviation analyses did not show that any compound uniquely stabilized the original dimer geometry. Over the 1,000–10,000-picosecond analysis window, the ligand-free dimer had a mean whole-dimer RMSD of 11.91 plus or minus 2.16 angstroms. Mean values for complexes containing 1-monopalmitin, isobutyl octyl oxalate, pentadecanoic acid or dimethyl azelate ranged from 8.98 to 12.39 angstroms. These between-system differences were smaller than the variation among independent simulation runs. The simulations therefore support a hypothesis that the molecules can transiently perturb dimer packing, but they do not demonstrate equilibrium binding, stronger affinity, irreversible disruption of amyloid assemblies or compound-specific dimer destabilization.

Pharmacokinetic predictions added a further stage of prioritization while revealing potential liabilities. Predicted human intestinal absorption ranged from 91.41% for 1-monopalmitin to 98.11% for pentadecanoic acid. Predicted blood-brain barrier penetration was higher for 1-monopalmitin and pentadecanoic acid, with scores of 6.48 and 6.64, respectively, than for dimethyl azelate at 0.99 and isobutyl octyl oxalate at 0.28. All four compounds were predicted to bind plasma proteins at 100%, which could limit the freely available fraction, and all were predicted to inhibit CYP2C19 while not inhibiting CYP2D6. The predicted CYP2C19 inhibition indicates a possible drug-interaction concern that requires experimental testing rather than demonstrating that such interactions occur in patients.

Important limitations constrain the conclusions. Docking used a restricted set of Aβ structures that cannot represent the peptide’s full conformational diversity, and the experimental work did not include cells, animals or humans. Neuronal toxicity, intracellular uptake, actual brain exposure, pharmacokinetics, safety and efficacy in Alzheimer’s disease models remain unevaluated. The researchers also did not quantify the four active compounds in the parent JWS extract, so it is unknown whether they are present at concentrations sufficient to account for the formulation’s previously reported anti-amyloid activity. Direct biophysical binding measurements, quantitative chemical analysis, cellular experiments, toxicology studies and in vivo work are still needed. The results identify preliminary chemical probes and possible development starting points, not established Alzheimer’s treatments.

Subject of Research: Computational screening and cell-free testing of herbal small molecules that inhibit Aβ42 aggregation and disaggregation

Article Title: Integrated computational screening and biochemical validation identify natural small molecules targeting Aβ aggregation and dissociation

Article References: Lee, H., Kumar, V., Shin, S. J., Nam, Y., Jang, S., Kim, H. S., Lee, M., Song, G. Y., Lim, E.-K., Chae, H.-J., & Moon, M. (2026). Integrated computational screening and biochemical validation identify natural small molecules targeting Aβ aggregation and dissociation. Journal of Pharmacy & Pharmaceutical Sciences, 29, Article 17216. https://doi.org/10.3389/jpps.2026.17216

Image Credits: AI Generated

DOI: 10.3389/jpps.2026.17216

Keywords: amyloid-beta 42, Alzheimer’s disease, Jowiseungchungtang, natural small molecules, computational screening, molecular docking, thioflavin T assay, molecular dynamics, blood-brain barrier, drug-likeness, Integrated, computational