Intelligence prediction of microfluidically prepared nanoparticles.

Developing poly(lactic-co-glycolic) acid (PLGA) nanoparticles with optimized drug encapsulation and loading is crucial for effective drug delivery. However, controlling the physicochemical properties of these nanoparticles remains challenging. In this study, we compiled a dataset of over 300 PLGA nanoparticle formulations from the literature, including 25 key features related to their preparation on microfluidic platforms. We applied various machine learning algorithms to predict encapsulation efficiency (EE) and drug loading (DL). The random forest model showed the best performance, achieving R² values of 0.93 and 0.96 for DL and EE predictions, respectively. Although EE and DL had minimal impact on each other's prediction, they provide distinct insights into nanoparticle formulation. Our results demonstrate that machine learning can effectively guide the design of drug delivery systems with desired properties, accelerating their development.