Leveraging Bayesian Optimization for Accelerated Ligand Selection in Atroposelective Negishi Coupling

We report a descriptor-informed Bayesian optimization strategy for accelerating ligand selection in asymmetric catalysis, demonstrated in a highly atroposelective Negishi cross-coupling reaction en route to a KRAS G12C inhibitor. Whereas multivariate linear regression failed to capture this complex catalytic landscape under sparse-data conditions, integration of forward-step descriptor selection with DFT-derived descriptors enabled efficient optimization of yield and diastereoselectivity, revealing the underexploited reactivity of Phanephos ligands that delivered the target atropisomer in 84% yield and 91:9 dr.