Optimization of poly(lactic-co-glycolic acid) particle properties for biomedical applications

In addressing modern health challenges, the strategic use of biomaterials stands at the forefront of scientific innovation and progression. These materials have become essential tools in the search for treatments and cures against many diseases. Poly(lactic- co-glycolic acid) (PLGA) is one such material that has exceptional biocompatibility and versatility and has been approved by the Food and Drug Administration (FDA) in terms of utilization in biomedical applications. Different health applications require particles with specific sizes and properties; for instance, drug delivery applications generally require nano-sized particles while cell-based therapies rely mostly on micro-sized particles to house the cells within. Consequently, the tailoring of PLGA particles for specific purposes requires a comprehensive optimization study. Here, by using a computational approach, the effect of individual synthesis parameters in the final size on the PLGA particles has been identified. According to the effects of these individual parameters, an artificial neural network (ANN) model is developed to predict the particle size accurately and robustly. Development of an ANN model required its optimization, in terms of model performance. Since the training data of the model comes from experimental results, the model lacked a high number of data points. Thus, any kind of generated model carries a risk of overfitting, where the model specializes in predicting the training set and performs poorly on presented unknown data. To be able to understand whether a model overfits, several cross-validation methods like leave-one-out cross-validation (LOOCV) and leave-p-out cross-validation (LPOCV) have been utilized. As a result, the Bayesian Regularization (BR) backpropagation method with 5-1-5 multiple hidden layer size network architecture was found to be robustly predicting the final size with any given parameters. Utilizing this ANN model, particles with desired sizes have been synthesized and used in several applications such as i- macroporous PLGA particles that carry cells and ii- PLGA particles that form an aggregate with spheroids to improve the viability. The first application required extensive optimization regarding the pore size of the particles which is essential for cell penetration. As a result, NIH-3T3 fibroblast cells were successfully pushed inside the PLGA particles via centrifugal cell immobilization without having any negative impacts on the viability of these cells. As for the second application, negatively charged PLGA particles were unsuccessful in forming aggregates with the fibroblast spheroids. Nevertheless, coating PLGA particles with chitosan showed promising results without impairing the viability of the cells and the release kinetics from the particles. In conclusion, this study offers a method to estimate particle size prior to synthesis and opens the way for further innovations in biomedical applications, leading to the development of more effective and customized therapeutic solutions.