Related Experiment Video
Updated: Jun 11, 2025
13:27
High-Throughput DNA Plasmid Multiplexing and Transfection Using Acoustic Nanodispensing Technology
Published on: August 8, 2019
8.9K
Machine Learning Elucidates Design Features of Plasmid Deoxyribonucleic Acid Lipid Nanoparticles for Cell
Leonardo Cheng1,2,3, Yining Zhu1,2,3, Jingyao Ma1,3,4
1Institute for NanoBioTechnology, Johns Hopkins University, Baltimore, Maryland 21218, United States.
ACS Nano
|October 7, 2024
Summary
Machine learning enhances lipid nanoparticle (LNP) design for cell and gene therapies. This approach identifies key LNP composition rules for efficient, cell type-specific gene delivery, improving therapeutic accessibility.
Area of Science:
- Biotechnology
- Gene Therapy Delivery
- Nanomedicine
Background:
- Nonviral gene carriers like lipid nanoparticles (LNPs) are crucial for cell and gene therapy accessibility.
- High-throughput screening (HTS) accelerates LNP discovery but is costly and lacks clear design rules.
- Existing methods often fail to provide actionable insights into LNP formulation parameters.
Purpose of the Study:
- To develop a machine learning (ML) workflow for extracting compositional and chemical insights from LNP transfection data.
- To identify key composition-function relationships governing cell type-preferential LNP transfection efficiency.
- To provide design rules for optimizing LNP formulations for targeted gene delivery.
Main Methods:
- Utilized a machine learning workflow with curated plasmid DNA LNP transfection data across six cell types.
- Applied SHapley Additive exPlanations (SHAP) to ML models to interpret composition-function relationships.
- Analyzed high-throughput screening data of diverse LNP libraries to elucidate lipid component interactions.
Main Results:
- Achieved prediction errors between 5-10% for LNP transfection efficiency across different cell types.
- Identified consistent LNP composition parameters enhancing in vitro transfection, including 9-50% charged helper lipid and cationic/zwitterionic helper lipids.
- Discovered parameters modulating cell type-preferentiality, such as ionizable/helper lipid percentages, N/P ratio, PEGylated lipid percentage, and helper lipid hydrophobicity.
Conclusions:
- ML analysis of HTS data provides valuable insights into LNP formulation design.
- Identified specific LNP compositional parameters that enhance transfection efficiency and cell type-specificity.
- This approach facilitates the rational design of LNP carriers for improved cell and gene therapy applications.
