Machine learning-assisted microfluidic approach for broad-spectrum liposome size control.

Area of Science:

• Pharmaceutical Sciences
• Biotechnology
• Chemical Engineering

Background:

• Liposomes are crucial drug and vaccine carriers, with size significantly impacting biological effects.
• Microfluidics offers precise, reproducible, and scalable liposome preparation.
• Current research often overlooks the production of larger liposomes, focusing primarily on smaller sizes.

Purpose of the Study:

• To investigate variables influencing liposome size in microfluidic preparation.
• To develop a machine learning model for accurate liposome size prediction.
• To enable precise control over both small and large liposome production.

Main Methods:

• Utilized a staggered herringbone micromixer (SHM) chip for microfluidic liposome preparation.
• Investigated multiple variables affecting liposomal size and polydispersity index (PDI).
• Developed and validated machine learning models, including ensemble methods, for size and PDI prediction.

Main Results:

• Identified significant interrelationships between variables affecting liposomal size.
• Ensemble machine learning algorithms, specifically gradient boosting for size and random forest for PDI, demonstrated high predictive accuracy.
• Successfully produced uniform large (600 nm) and small (100 nm) liposomes using ML-optimized conditions.

Conclusions:

• Presents a robust methodology for precise control over liposome size distribution using machine learning.
• Provides valuable insights for optimizing microfluidic liposome preparation for medicinal research.
• Enables tailored production of liposomes for diverse drug delivery applications.