Self-driving lab for the data-driven design of single-chain polymer nanoparticles

Careful mapping of protein structure-activity relationships connecting amino acid sequence to its tertiary structure has spurred the rise of rational protein design using computational and experimental methods. Drawing inspiration from proteins for synthetic materials, careful choice of monomers enables the creation of novel single-chain polymer nanoparticles (SCNPs) with self-assembling characteristics. Rationally designed SCNPs permit the necessary structural complexity for use as protein mimics and polymer-protein hybrids, while granting access to a broad chemical design space, straightforward preparation, and tunable properties. Advances in oxygen-tolerant photoinduced polymerization chemistries have greatly facilitated efficient preparation of SCNPs. Here, we describe a high-throughput, autonomous workflow for active learning to discover structure-property relationships and use them to iteratively predict and synthesize novel SCNPs. We developed a control system consisting of a liquid handling robot to mix polymerization reagents, a custom-built lightbox to catalyze PET-RAFT polymerization, a dynamic light scattering (DLS) plate reader to characterize polymer hydrodynamic radius (Rh), and a robotic arm to transport polymer-containing well plates from one instrument to another. By first testing low-feature, rationally designed polymer libraries, we gained experimental context for the development of higher-feature, randomly sampled seed libraries as training sets for Gaussian process regressor (GPR) models. Over multiple generations of Bayesian optimization (BO), additional generations of polymer synthesis were found not only to successfully improve model performance but also to represent the impact of specific monomer content on Rh. This automated polymer discovery platform serves as a useful prototype for SCNP design using more complex design features as well as more advanced optimization targets, including polymers whose structure can be tailored for enzyme stabilization and other biomedical applications.