Study on the formation, release and stability mechanisms of self-assembled nanoparticles based on machine learning, explainable artificial intelligence and multiscale simulations.

As a frontier strategy in nanomedicine, self-assembled nanoparticles provide an important carrier-free platform for delivering natural bioactive compounds. However, their formation, stability, and release mechanisms remain insufficiently understood, which severely limits their further development and translational application. In this study, berberine-mangiferin self-assembled nanoparticles (BM-NPs) were used as a model system, and machine learning algorithms were integrated with multiscale simulations to elucidate their formation, stability, and release mechanisms from two complementary levels: macroscopic formulation regulation and microscopic molecular interactions. Machine learning and explainable analysis were used to identify the key factors governing nanoparticle formation at the macroscopic level. Spectroscopic, solid-state, and morphological characterizations confirmed that nanoparticle formation was accompanied by intermolecular association and structural rearrangement. Multiscale simulations showed that the self-assembly process was cooperatively driven by electrostatic attraction, hydrogen bonding, hydrophobic interactions, and π-π stacking. In vitro release studies showed that BM-NPs exhibited significant sustained-release behavior and pH dependence compared with the free drugs. The corresponding dynamic simulations indicated that this release behavior was mainly regulated by pH-dependent intermolecular association, solvent exposure, and structural fluctuation. Stability studies and corresponding dynamic simulations further revealed that the stability of BM-NPs is closely related to aggregate compactness, solvent exposure, and intermolecular cohesion.