Machine Learning for the Prediction of Size and Encapsulation Efficiency of mRNA-Loaded Lipid Nanoparticles Following...

Area of Science:

• Nanomedicine
• Biotechnology
• Materials Science

Background:

• Lipid nanoparticles (LNPs) are crucial for mRNA delivery.
• Optimizing LNP formulation requires understanding lipid composition and preparation methods.
• Microfluidics offers precise control over nanoparticle formation.

Purpose of the Study:

• To investigate mRNA-LNPs development using microfluidic preformed vesicles (PFVs) and postencapsulation.
• To create and analyze a dataset of PFVs and mRNA-LNPs properties.
• To develop a predictive machine learning model for LNP characteristics.

Main Methods:

• Microfluidic production of PFVs with varying lipid types, sterol types, flow rates, and chip designs.
• Postencapsulation of mRNA into selected PFVs.
• Characterization of PFVs and mRNA-LNPs size, polydispersity index (PDI), and encapsulation efficiency (EE%).
• Training and validation of an XGBoost model with semisupervised learning.

Main Results:

• Chip design significantly impacts PFV size and PDI, with higher mixing efficiency yielding smaller PFVs.
• Postencapsulation increased nanoparticle size and decreased PDI.
• The XGBoost model accurately predicted LNP size and EE% based on formulation and process parameters.
• Semisupervised learning enhanced model performance by incorporating PFV data.

Conclusions:

• Microfluidics combined with machine learning accelerates LNP formulation development.
• The predictive model aids in optimizing mRNA-LNP design for efficient mRNA delivery.
• This integrated approach provides valuable insights with limited resources, advancing nanomedicine research.