DNA-Scaffolded Ultrahigh-Throughput Reaction Screening

Discovering and optimizing reactions is central to synthetic chemistry. However, chemical reactions are traditionally screened using relatively low-throughput methods, prohibiting exploration of diverse chemical space, particularly for reactions involving multiple components with millions of potential combinations. State-of-the-art technologies designed to increase screening throughput are limited to thousands of reactions and often require weeks to months for data collection. Here we report a DNA-encoded combinatorial screening platform, enabling large matrices of reactions to be performed in a single test tube followed by simultaneous reaction analysis using DNA sequencing. Using this platform, a single researcher performed 504,000 reactions in < 3 days, encompassing a variety of reaction conditions and times. We combined this platform with data science tools to design a targeted library covering chemical space broadly, then used the resulting dataset to train a machine learning model for both prediction and interpretation tasks, demonstrating utility for reaction development and mechanistic studies. Overall, this technology discloses fundamentally new opportunities in data-driven discovery, optimization, and mechanistic understanding of synergistic catalytic systems.