Virtual Variables-Enabled Generation of Datasets for Prediction in Organic Synthesis: Digitalization of Small Molecules and Application to Functional Molecule Syntheses

The generation of large datasets in traditional organic synthesis experiments is extremely challenging. Hence, methods that predict outcomes accurately using a limited amount of data are in demand. We propose a molecular technology based on generative artificial intelligence that generates data from unexplored conditions and establishes the most suitable relationships between different small molecules using virtual variables. Our approach reveals relationships among three structurally different small molecules and represents them as virtual variables, which are then utilized to propose the reaction conditions for synthesizing target molecules in high yields. We demonstrate its utility in small molecule syntheses through its application to the iodination reaction of polyfluoronaphthalenes. By computationally generating inaccessible data through reasonable reaction experiments, we successfully optimized reaction conditions. We introduce a novel application of machine learning as a molecular technology for predicting reaction outcomes based on a small dataset containing less than 100 data points.