Effect of various parameters on encapsulation efficiency of mPEG-PLGA nanoparticles: artificial neural network

In this work we prepared curcumin loaded mPEG-PLGA nanoparticles using precipitation technique and investigated the effect of various parameters such as polyvinyl alcohol (PVA), curcumin concentrations and stirrer time on encapsulation efficiency (EE) of curcumin into mPEG-PLGA nanoparticles. Artificial neural networks (ANN) were used to model the data in order to find an ideal model which can fit the data and predict the EE with the lowest error and highest linear regression. The different samples of nanoparticles were prepared as training and testing datasets using the k-fold cross validation procedure. The best ANN design comprised 2 hidden layers with 8 and 1 nodes in each layer, respectively. Levenberg-Marquardt back propagation with log-sigmoid transfer function was the best model for our datasets. The mean square error and correlation coefficient between the observed and the predicted EE of curcumin into mPEG-PLGA nanoparticles were 0.1609 and 0.9209, respectively. In addition, three-dimensional correlation graphs showed that the most important pairs of variables which had a greater impact on EE were mPEG-PLGA/curcumin concentration and mPEG-PLGA/PVA concentration.