Virtual Forward Synthesis: Designing Novel Polymers from Commercially Available Chemistries

Every year, humanity churns out a staggering 380 million tons of plastic, with millions of these tons escaping into the environment. The modern plastics we rely on possess such remarkable thermodynamic stability that recycling them remains a daunting task, with only a meager 9% successfully recycled, while they persist for decades, if not centuries, in our ecosystems. This dire situation has ignited an ecological crisis as these plastics seep into the environment, fragment into minuscule particles, and accumulate within the bodies of humans, plants, and animals.<br/> <br/>In response to this critical issue, we've developed a computational method known as 'Virtual Forward Synthesis' (VFS) to explore the creation of hypothetical polymers using readily available chemicals. This approach involves the automated retrieval of chemical data and pricing from vendors, followed by systematic cataloging of the substructures of these chemicals. Viable chemicals are then subjected to virtual reaction templates, yielding hypothetical polymers, with machine learning being used to predict the polymers’ properties.<br/> <br/>The most promising polymers, along with comprehensive information on their chemical composition, pricing, vendors, and the reaction procedure used to create them, are then shared with our collaborating chemists who attempt to synthesize these novel polymers. We evaluate a polymer's potential by assessing how well it matches the required properties for common plastic applications like packaging. We assign scores based on how closely its predicted properties align with the desired values. Additionally, this presentation provides an overview of our carefully designed, performance-optimized relational database schema.