The search for synthetic polymers capable of fulfilling multiple desired objectives simultaneously remains challenging. One of the approaches is to use multicomponent copolymers, which use multiple monomers capable of affecting desired objectives into the same material. Due to the interdependent effects of these monomers on material function, the overall properties of produced polymer remain difficult to predict a priori. Discovery of high contrast 19F magnetic resonance imaging (MRI) polymer-based agents is one of such cases. Discovering a composition capable of producing a copolymer that is both water-soluble and contains enough fluorine nuclei to be visualized on clinical MRI scanners remains a challenge. Copolymers with these combined functions result in searching within a multidimensional chemical space with the unique organization of confronting properties; moreover, experimental evaluation is possible for only a small fraction of that space. Recent achievements in continuous-flow copolymer synthesis have resulted in an automated platform for copolymer synthesis that enables data acquisition at a rate suitable for cross-comparison of copolymer compositions. This opens a way for the application of Artificial Intelligence to navigate the search: high throughput polymer synthesis coupled with machine learning models in an Active Learning cycle allows a systematic exploration of chemical space, model development with respect to multiple desired properties, and selection of new putative copolymer composition in an iterative manner. We developed an AI-based platform that allows us to operate the continuous-flow synthesis in a semi-autonomous mode. We evaluated a large compositional space of potential copolymers and performed a multiobjective optimization of copolymer 19F MRI agents. Utilizing only eight active learning steps and sampling under 1% of the space, we developed a new best-in-class polymer with a record 19F signal-to-noise ratio.

Active Learning guided design of automated copolymer synthesis for novel 19F MRI agents.

ACS Fall 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.