Data-driven construction of an imine reductase library capable of broad-scope reductive amination at equimolar...

Reductive amination catalysed by imine reductases and reductive aminases has emerged as a powerful method for the asymmetric synthesis of chiral amines, yet the generality and practical limitations of these enzymes remain poorly understood. Here, we combine large-scale, iterative activity screening, sequence analysis and machine-learning-guided exploration to systematically profile reductive amination performance across 175 enzymes and structurally diverse substrate panels. We show that asymmetric reductive amination at equimolar substrate concentrations is widespread within this enzyme family and assemble catalysts that display broad scope, high specific activity and excellent stereoselectivity. Sequence-level analysis identifies characteristic features associated with broad-scope reductive aminases and pinpoints key residues governing stereoselectivity, while machine-learning models trained on the screening data enable successful prediction of enzyme activity for previously untested substrate combinations. Selected reactions were scaled to a preparative batch size, affording gram quantities of optically pure amine products.