Highly selective and AI-predictable Se–N exchange chemistry between benzoselenazolones and boronic acids for programmable, parallel, and DNA-encoded library synthesis

Abstract

Chemical reactions compatible with multiple functionalities are essential for rapid, programmable, and automatable synthesis of functional molecules. However, achieving such reactivity poses significant challenges. Here, we developed a novel multi-orthogonal C(sp²)–Se bond formation reaction between benzoselenazolones and boronic acids via Ag(I)-catalyzed selective selenium(II)–nitrogen exchange. This chemistry is compatible with diverse functionalities, enabling sequential and programmable synthesis. Moreover, it features modular, high-yielding (485 examples, with yields or conversions exceeding 70% in 95% of cases), and switchable reaction systems under mild conditions. Its practical utility was exemplified through late-stage functionalization of natural products, peptide modification and ligation, diversified synthesis, sequential click chemistry, protecting group-free syntheses of sequence-defined oligo selenides (nonamers), on-plate nanomole-scale parallel synthesis (200 nmol, 412 selenides), and DNA-encoded library (DEL) synthesis (10 nmol, 92 examples). Notably, a target-based screening identified SA-16 as a potent CAXII inhibitor with an IC₅₀ value of 72 nM. Furthermore, a machine learning-based model (SeNEx-ML) was established for reaction yield prediction, achieving 80% accuracy in binary classification and 70% balanced accuracy in ternary classification. These results demonstrated that this chemistry serves as a powerful tool to bridge the selenium chemical space with the existing chemical world, offering transformative potential across multidisciplinary fields.