Machine Learning-Enabled Polymer Discovery for Enhanced Pulmonary siRNA Delivery

Abstract

Nucleic acid therapeutics are poised to revolutionize the clinical treatment of diseases once considered undruggable. Although these therapeutic approaches hold significant promise, delivering the nucleic acid cargo remains challenging due to susceptibility to nuclease degradation. Among all carrier systems, polymers stand out for their high tunability and cost-effectiveness. However, their flexible structure greatly expands the chemical space, making experimental exploration both costly and time-consuming. Leveraging published data and machine learning methods provides a valuable strategy to address these issues. The present study demonstrates a way to merge data from multiple sources and uses this information to identify new polyesters that effectively deliver siRNA into lung cells. One newly discovered polymer is further examined in ex vivo experiments and tested in a mouse model. The results indicate that a polymer capable of silencing specific genes in vivo can be discovered through machine learning, circumventing an extensive trial-and-error process in the search for novel materials.