Analysis of an electrically responsive drug delivery system for ibuprofen on-demand release using a machine learning...

This study aims to optimize ibuprofen-based Drug Delivery Systems (DDSs) to address their short half-life and enhance controlled release. Advanced machine learning techniques, including Artificial Neural Networks, Random Forest, and CatBoost, were employed to model and predict the behavior of DDSs under various experimental conditions. Ibuprofen was loaded into a conducting polymer matrix based on PEDOT, and drug release was achieved through programmable electrical stimuli. Data preprocessing included the normalization of numerical variables and One-Hot Encoding for categorical variables. The predictive models, optimized via Optuna for hyperparameter tuning, demonstrated robust performance, achieving high R² values for the released concentration and kinetic parameters of release. SHAP (SHapley Additive exPlanations) analysis provided insights into the influence of key variables, such as diffusion time, stimulus type, and electrode type, that significantly impacted drug release behavior. The results demonstrated the potential of machine learning to optimize DDSs, improving predictive accuracy and enabling personalized and efficient approaches. This study contributes to the design of more advanced and sophisticated pharmaceutical formulations aligned with the principles of personalized medicine and on-demand delivery.