Artificial intelligence-driven rational design of ionizable lipids for mRNA delivery

Area of Science:

• Biotechnology and Pharmaceutical Sciences
• Computational Chemistry and Drug Design
• Nanomedicine and Drug Delivery

Background:

• Lipid nanoparticles (LNPs) are crucial for mRNA delivery, as demonstrated by their success in COVID-19 vaccines.
• Traditional methods for optimizing ionizable lipids, the key LNP component, involve inefficient and expensive experimental screening.
• A need exists for accelerated, rational design strategies for novel ionizable lipids.

Purpose of the Study:

• To employ artificial intelligence (AI) and virtual screening for the rational design of ionizable lipids.
• To predict key LNP properties, including apparent pKa and mRNA delivery efficiency.
• To expedite the discovery of high-performance ionizable lipids for mRNA therapeutics.

Main Methods:

• Utilized AI-driven generation and virtual screening to evaluate approximately 20 million ionizable lipids across two iterations.
• Predicted apparent pKa and mRNA delivery efficiency for candidate ionizable lipids.
• Validated promising AI-identified lipids through in vivo mouse testing.

Main Results:

• Identified three novel ionizable lipids in the first AI iteration, with one showing performance comparable to DLin-MC3-DMA (MC3).
• Discovered six new ionizable lipids in the second iteration, all matching or exceeding MC3 performance.
• One lipid from the second iteration demonstrated efficacy similar to the superior control SM-102, and the AI model provided interpretable structure-activity relationships.

Conclusions:

• AI and virtual screening offer an efficient and effective approach for the rational design of ionizable lipids.
• The developed AI model successfully identified novel ionizable lipids with superior or comparable performance to existing standards.
• This AI-driven methodology significantly accelerates LNP development for mRNA delivery applications.