Hydride-transfer reactions underpin key redox processes in both biology and reduction reactions in synthetic chemistry. However, rationalizing existing and developing novel hydride-transfer reactions in water remains difficult because hydrolysis competes with nucleophilic attack. In this work, we combined experimental kinetic measurements with a data-driven approach to map the reactivity space of boron-based hydride donors. Initially, a library of trihydroborates was synthesized and evaluated. Next, multivariate linear regression (MLR) models based on quantum-chemical descriptors were used to predict both hydride-transfer reactivity and hydrolytic stability. Finally, we used our trained model to predict the properties of hundreds of substrates extracted from a literature database; selected candidates were then synthesized and experimentally evaluated. Overall, this work establishes an experimentally grounded, data-driven framework for rapid virtual screening and rational design of hydride donors for reduction chemistry in water.
Felix Söhngen