Polymer design via SHAP and Bayesian machine learning optimizes pDNA and CRISPR ribonucleoprotein delivery

Area of Science:

• Polymer Chemistry
• Biomaterials Science
• Gene Delivery Systems

Background:

• Optimizing nucleic acid delivery vehicles is crucial for advancing gene therapies.
• Current carriers often face challenges with efficiency, specificity, and safety.
• Systematic exploration of polymer properties is needed to design improved delivery systems.

Purpose of the Study:

• To synthesize and characterize a diverse library of clickable polymers.
• To identify structure-property relationships governing intracellular delivery of plasmid DNA (pDNA) and CRISPR-Cas9 ribonucleoprotein (RNP).
• To accelerate the design of next-generation nucleic acid delivery vehicles using machine learning.

Main Methods:

• Facile synthesis of a polymer library with systematic variations in length, composition, pKa, and hydrophobicity.
• Physicochemical characterization of polymer properties.
• Application of machine learning (explainable AI and Bayesian optimization) to analyze structure-property relationships and optimize formulations.
• In vitro and in vivo evaluation of polymer performance for pDNA and RNP delivery.

Main Results:

• Discovered distinct optimal design parameters for pDNA and RNP delivery.
• Lower polymer pKa and higher benzimidazole ethanethiol content enhanced pDNA delivery.
• Increased polymer length and specific captamine cation identity improved RNP delivery.
• Machine learning identified key quantitative structure-property relationships.
• Optimized polymers demonstrated enhanced transgene expression in vivo over 20 days.

Conclusions:

• A facile, coupled approach of synthesis, characterization, and machine analysis accelerates the development of nucleic acid delivery vehicles.
• Identified disparate polymer design principles for pDNA versus RNP delivery, offering new insights for carrier optimization.
• The developed polymer library and machine learning tools provide a powerful platform for designing efficient and targeted gene therapy vectors.