Predicting the encapsulation efficiency of polymeric nanoparticles using machine learning approaches

Background and Aims: Encapsulation efficiency is a critical parameter in drug delivery systems, which is defined as the percentage of drug loaded in the carrier to the total amount of drug used in the formulation. Since determining encapsulation efficiency experimentally is time-consuming and costly, machine learning techniques could be used to predict this parameter. This study emphasized the development and comparison of predictive models for encapsulation efficiency prediction using Random Forest and Support Vector Machine algorithms. Methods: A data set of ۵۷ samples was collected from related articles, including variables such as particle size, polydispersity index, zeta potential, polymer properties (name, molecular weight, Log P, glass transition temperature), drug properties (name, solubility, molecular weight, Log P), and experimental conditions (polymer concentration, drug loading method, polymerization method, temperature, pH, drug-to-polymer ratio, incubation time, solvent type). Predictive models were built in RapidMiner ۱۰.۱.۰۰۲ using Random Forest and Support Vector Machine algorithms. Model performance was evaluated using Root Mean Squared Error and Mean Absolute Error. Correlation-based weights were used to measure feature importance. Results: The accuracy of Random Forest model was more than the Support Vector Machine model, with a Root Mean Squared Error of ۱۲.۴۹۸ and a Mean Absolute Error of ۱۲.۴۹۸. The Support Vector Machine model, for comparison purposes, shows a Root Mean Squared Error of ۱۷.۰۶۲ and a Mean Absolute Error of ۱۷.۰۶۲. Weight analysis using correlation determined that polydispersity index (۰.۵۶۹), particle size (۰.۵۴۷), and pH of the medium (۰.۴۵۷) were the most effective features of encapsulation efficiency. Other important features included zeta potential (۰.۳۱۹), polydispersity index after loading (۰.۳۰۳), and drug-to-polymer ratio (۰.۲۱۴). Features such as polymer name (۰.۰۰۵) and temperature (۰.۰۳۳) had no significant contribution. Conclusion: The study is a witness to the capability of machine

کلیدواژه ها:

Encapsulation efficiency ، polymeric nanoparticles ، machine learning ، Random Forest ، Support Vector Machine ، drug delivery systems