Construction and application of a multifunctional material intelligent screening platform for food-derived bioactive compounds: enhancing solubility, antioxidant, and antibacterial properties.

This study aimed to construct an intelligent screening platform for co-amorphous materials (CAMs) based on machine learning and large language models to improve the development efficiency of co-amorphous systems for food-derived bioactive compounds. The platform enabled formation-probability prediction and interpretable analysis, thereby supporting the efficient development of CAMs. Guided by this platform, three novel myricetin-based CAMs were successfully developed. Solid-state characterization confirms the successful amorphization of the three systems. Multiscale simulations reveal that multi-site noncovalent interactions are the main driving forces for amorphization. The CAMs markedly improved myricetin wettability and simulated digestive release. The CAMs enhanced antioxidant activity, with the lowest DPPH and ABTS IC50 values reaching 4.21 and 1.86 μM, respectively. CAMs increased the inhibition zone diameters against S. aureus and E. coli to approximately 1.14-1.26-fold and 2.12-3.15-fold those of free myricetin, respectively. Preservation studies further show that the three materials effectively extend the shelf life of tomatoes.