Neural Network-based Optimization of Extract-loaded Chitosan Particles: Modeling, Preparation and Antioxidant Evaluation.

Background: Silymarin is a flavonolignan extracted from Silybum marianum with various therapeutic applications. Many studies have focused on improving the bioavailability of silymarin due to its wide range of efficacy and low bioavailability. Chitosan, a naturally occurring polymeric substance, has a strong reputation for increasing the solubility of poorly soluble compounds.

Objective: This study used artificial neural networks (ANNs) to measure the effects of pH, chitosan to silymarin ratio, chitosan to tripolyphosphate ratio, and stirring time on the loading efficiency of silymarin into chitosan particles.

Methods: A model was developed to investigate the interactions between input factors and silymarin loading efficiency. The DPPH method was utilized to determine the antioxidant activity of an optimized formula and pure raw materials.

Results: According to the outcome of the ANN model, pH and the chitosan to silymarin ratio demonstrated significant effects on loading efficiency. In addition, increased stirring time decreased silymarin loading, whereas the chitosan-to-tripolyphosphate ratio showed a negligible effect on loading efficiency.

Conclusion: Maximum loading efficiency occurred at a pH of approximately~5. Moreover, silymarin- loaded chitosan particles with a lower IC value (36.17 ± 0.02 ppm) than pure silymarin (165.04 ± 0.07 ppm) demonstrated greater antioxidant activity.