Neural-Network-Assisted Molecularly Imprinted Polymers for Green Extraction of Neohesperidin from Grapefruit Peel Waste

Neohesperidin (NHP), a flavanone compound widely distributed in citrus peels and traditional Chinese medicinal materials, such as immature and mature bitter orange (Aurantii fructus immaturus and Aurantii fructus), exhibits various pharmacological activities, including antioxidant, anti-inflammatory, and neuroprotective effects. The large quantity of peel waste generated as a byproduct of citrus processing is rich in NHP, representing an important source of this compound and highlighting the importance of its high-value utilization. However, conventional extraction methods suffer from low efficiency and significant solvent-related pollution. Therefore, the development of adsorbents featuring high efficiency, good reusability, and strong specificity toward NHP is of great significance. In this study, a neural network model was constructed for the first time to predict suitable functional monomers for molecularly imprinted polymers (MIPs). Based on the prediction results, NHP molecularly imprinted polymers (NHP/MIPs) were successfully synthesized. The experimental results demonstrated that the prepared NHP/MIPs exhibited high specificity and adsorption capacity for NHP, with an adsorption capacity of 35.58 mg/g and an imprinting factor (IF) of 1.89. After six adsorption–desorption cycles, the material maintained excellent regeneration stability, achieving a recovery rate of 74.62% to 77.99% for NHP in spiked samples. The neural-network-assisted functional monomer screening strategy proposed in this study significantly enhances the synthesis efficiency and molecular recognition specificity of MIPs, providing an approach for the selective enrichment of target molecules in complex samples. This research offers a promising adsorbent for the selective enrichment of NHP from citrus fruits and complex matrices.