Background/Objectives: The accurate prediction of drug release profiles from Poly (lactic-co-glycolic acid) (PLGA)-based drug delivery systems is a critical challenge in pharmaceutical research. Traditional methods, such as the Korsmeyer-Peppas and Weibull models, have been widely used to describe in vitro drug release kinetics. However, these models are limited by their reliance on fixed mathematical forms, which may not capture the complex and nonlinear nature of drug release behavior in diverse PLGA-based systems. Method: In response to these limitations, we propose a novel approach—DrugNet, a data-driven model based on a multilayer perceptron (MLP) neural network, aiming to predict the drug release data at unknown time points by fitting release curves using the key physicochemical characteristics of PLGA carriers and drug molecules, as well as in vitro drug release data. We establish a dataset through a literature review, and the model is trained and validated to determine its effectiveness in predicting different drug release curves. Results: Compared to the traditional Korsmeyer–Peppas and Weibull semiempirical models, the MSE of DrugNet decreases by 20.994 and 1.561, respectively, and (R2) increases by 0.036 and 0.005. Conclusions: These results demonstrate that DrugNet has a stronger ability to fit drug release curves and better capture nonlinear relationships in drug release data. It can deal with the nonlinear change of data better, has stronger adaptability and advantages than traditional models, and overcomes the limitations of the mathematical expressions in traditional models.
Loading....