An Overlooked Formulation Variable: Solvent Reshapes DNA Delivery

Nonviral polymeric vectors offer a tunable platform for nucleic acid delivery, yet formulation variables beyond polymer structure remain underexplored. Here, we systematically investigate the role of stock solvent in shaping gene delivery performance using end-capped highly branched poly(amino ester) (HPAE) vectors. By combining a structured experimental design with an interpretable machine learning framework, we reveal that transfection efficiency is primarily governed by polymer features, while solvent properties modulate performance through strong polymer-solvent interactions. Descriptor-level analysis identifies key molecular features associated with high-performing formulations, highlighting the importance of polymer size, charge density, and hydrophobic-polar balance. Selected polymer-solvent combinations enable efficient CFTR protein restoration in cystic fibrosis disease cells and achieve robust in vivo transgene expression following systemic administration in mice. Overall, this work establishes stock solvent as a critical and tunable formulation parameter and provides a data-driven framework for optimizing polymeric gene delivery systems.