FALCON: Closed-Loop Multi-Objective Optimization of Lipid Nanoparticles for Cell-Selective mRNA Delivery.

Area of Science:

• Biotechnology
• Nanomedicine
• Molecular Biology

Background:

• Efficient and cell type-selective delivery of genetic materials is crucial for advancing gene and cell therapies.
• Current lipid nanoparticle (LNP) optimization relies on inefficient, brute-force screening methods.
• Developing targeted delivery systems requires overcoming challenges in payload delivery and LNP composition design.

Purpose of the Study:

• To introduce FALCON (Framework for Active Learning-driven Compositional Optimization of Nanoparticles), an AI-powered pipeline for accelerating LNP design.
• To demonstrate FALCON's ability to optimize LNP composition for enhanced gene delivery and cell selectivity.
• To validate FALCON's performance in preclinical models for B cell and myeloid cell targeting.

Main Methods:

• Utilized a closed-loop pipeline combining iterative screening, surrogate modeling, and multi-objective optimization.
• Employed active learning strategies to guide the compositional design of LNPs.
• Performed in vivo validation experiments targeting splenic B cells and myeloid cells in preclinical models.

Main Results:

• FALCON-optimized LNPs showed a 1.8-fold increase in splenic B cell transfection in vivo.
• Achieved an 84-fold improvement in selective splenic B cell transfection over off-target liver cells.
• Demonstrated enhanced myeloid cell-selective delivery and improved vaccine-induced immune responses (higher IgG2c titers, Th1 bias).

Conclusions:

• FALCON significantly accelerates the design and optimization of LNP compositions for targeted gene delivery.
• The FALCON framework enables data-driven optimization for improved therapeutic efficacy and reduced off-target effects.
• FALCON represents a powerful tool for developing precision gene delivery systems for various therapeutic applications.