Design and development of a microfluidic-based high-throughput platform for formulation screening and optimization of lipid nanomedicines for gene therapy

Nanomedicine is rapidly transforming modern biopharmaceuticals, with lipid nanoparticles (LNPs) emerging as the leading non-viral carriers for nucleic acid therapeutics. However, their clinical translation remains limited by inefficient formulation screening, challenges in scale-up, and the lack of predictive design strategies. This thesis develops an integrated microfluidic and machine learning framework to overcome these bottlenecks, advancing both the engineering foundations and therapeutic applications of LNPs. Chapter 1 reviews the state-of-the-art in nanoparticle-based biopharmaceuticals, highlighting knowledge gaps in formulation, scale-up, and characterization. Chapter 2 presents a tilted baffle–structured microfluidic chips, optimized via computational fluid dynamics (CFD) and validated experimentally, to enhance chaotic mixing and produce uniform mRNA-LNPs with reproducible performance. Chapter 3 introduces a unique aerofoil-structure channel, incorporated into the MiNANO-form platform, which enables high-throughput parallel formulation screening through eight-channel microfluidic cartridges. This system achieves robust LNP preparation across flow rates from 0.1 to 4 mL/min using minimal reagent consumption, providing an efficient and reproducible route for systematic formulation exploration. Chapters 4 and 5 integrate these platforms with artificial intelligence (AI) to address therapeutic applications. A library of over 864 LNP formulations was generated and analyzed to build predictive models linking composition to size, stability, and cellular uptake. In Chapter 4, this workflow identified lead formulations for siRNA delivery, achieving effective S100P gene silencing in A549 lung cancer cells and favourable in vivo biodistribution. Chapter 5 applied the same strategy to cystic fibrosis (CF), where machine learning (ML)–guided screening pinpointed SM-102–based LNPs capable of restoring CFTR protein expression in bronchial epithelial cells following mRNA delivery. Overall, this thesis establishes a systematic and generalizable framework for LNP design by combining advanced microfluidic engineering with AI-driven optimization. While demonstrated here in siRNA-based cancer therapy and mutation-agnostic CFTR mRNA delivery, the principles and methodologies developed are broadly applicable, offering a foundation for accelerating the translation of next-generation RNA therapeutics.