リポソームの薬物送達応用におけるマイクロ流体および人工知能支援設計の進歩

Liposomes are used as a vehicle in targeted drug delivery due to inherent biocompatibility and encapsulative potential for diverse bioactive agents. However, consistent production remains challenging. With the help of microfluidic systems, liposome synthesis can be precisely controlled, conferring consistency and scalability. This review focuses on the integration of innovative microfluidic-based liposome synthesis and artificial intelligence (AI), specifically machine learning (ML), for the design of optimal critical quality attributes (CQAs), such as vesicle size, size distribution and encapsulation efficiency. ML models such as artificial neural networks (ANN), support vector machines (SVM), and random forest (RF) can predict liposome quality with limited input data. Together with design-of-experiment (DoE), these paradigms provide data-driven optimization. However, the interpretability of the model is enhanced by explainable AI (XAI) tools such as SHAP, which identify determinants for every prediction. The integration of these methods makes an advance toward robust and scalable liposome design. We have also summarized the challenges and future perspectives of scaling AI-guided microfluidic platforms into mainstream pharmaceutical development.