Cubane Isomerization: Transformations Overlooked for Half a Century Uncovered by an Algorithmic, First-Principles Catalyst Discovery Framework

The discovery of new chemical reactions remains a central challenge in chemistry, as it still relies heavily on trial-and-error experimentation guided by human intuition. Among such longstanding targets, cubane isomerization has been studied extensively for decades. Here we investigate this reaction using an algorithmic, first-principles strategy for reaction discovery and catalyst optimization. The method integrates automated reaction path exploration, kinetic prediction of product distributions, and virtual ligand-based optimization of catalyst properties to identify catalyst features that favor a target product. Applied to cubane isomerization, the framework revealed overlooked rearrangement pathways and the catalyst characteristics required to realize them. Using these predicted catalyst characteristics as experimental design principles, we confirmed two reactions that had remained hidden for more than five decades without prior empirical optimization.