A novel machine learning framework-based rapid screening of ionizable lipids in LNPs for highly-efficient mRNA...

Area of Science:

• Biotechnology
• Materials Science
• Computational Biology

Background:

• Ionizable lipids are crucial for lipid nanoparticles (LNPs) used in mRNA therapeutics.
• Traditional screening of ionizable lipids is labor-intensive, costly, and time-consuming.
• There is a need for efficient methods to discover novel ionizable lipids for improved mRNA delivery.

Purpose of the Study:

• To develop a machine learning framework, LipidAI, for rapid evaluation of novel ionizable lipids.
• To enhance data availability and model accuracy using the Methyl Tail Augmentation (MTA) strategy.
• To improve predictive accuracy by integrating multiple algorithms via Ensemble Stacking Learning (ESL).

Main Methods:

• Developed the Methyl Tail Augmentation (MTA) strategy to increase ionizable lipid data.
• Implemented an Ensemble Stacking Learning (ESL) algorithm for enhanced predictive power.
• Validated LipidAI predictions against in vivo Luc-mRNA expression data.

Main Results:

• The MTA strategy effectively increased the dataset size for ionizable lipids.
• The ESL algorithm demonstrated superior predictive accuracy compared to single algorithms.
• LipidAI predictions showed high consistency with experimental in vivo results.

Conclusions:

• LipidAI offers a rapid and efficient method for screening ionizable lipids.
• This AI-driven approach overcomes limitations of traditional lipid development.
• LipidAI has the potential to accelerate the development of LNP-based mRNA nano-drugs.