Speeding Up the Design of Siloxane-Based Ionizable Lipids for Lipid Nanoparticles Using Data-Efficient Kolmogorov-Arnold Networks

A KAN model achieved a predictive accuracy of 0.710 for mRNA delivery efficiency using only 36 training samples.

• The KAN model outperformed traditional machine learning models by an average absolute improvement of 0.627 in cross-validation.
• The framework identified three novel siloxane-based ionizable lipid candidates with superior predicted performance.
• Molecular dynamics simulations showed that the optimal candidate lipid had a binding free energy minimum of -3.011 kcal/mol, a 187% reduction compared to the best experimental control.
• The findings suggest a strong correlation between the predicted delivery efficiency and the binding affinity of the candidate lipids to the endosomal membrane.

Simplified