End-to-end integration of Density Functional Theory simulations and Machine Learning affords broad feasibility model for Nickel metallaphotoredox Csp2-Csp3 cross-coupling reactions

Nickel/photoredox catalysis in cross-coupling reactions has recently emerged as a powerful tool for efficient C–C bond formation as it enables mild operating conditions, thus expanding synthetic scope to molecules of pharmaceutical interest. Successful routine implementation of such reactions is limited, due to the large number of chemical species at play, resulting in complex optimization tasks. The development of predictive models for these novel synthetic methodologies remains challenging due to limited experimental data availability. The integration of quantum-mechanics (QM) calculations with machine learning has proven effective for developing predictive models of complex reactions with sparse experimental data. Here, we present a quantum mechanics-machine learning (QM-ML) approach to predict the feasibility of dual-catalyzed nickel-photoredox cross-coupling reactions across four reaction subtypes: bromide cross-electrophile couplings, chloride cross-electrophile couplings, deoxygenative couplings, and amino radical transfer (ART) couplings. Random-Forest classification models are trained using DFT-computed descriptors and automated simulations outcomes alongside cheminformatics features. This model demonstrates that QM-ML approaches can successfully predict complex reaction feasibility with minimal data requirements, offering a practical solution for rapidly adopting new synthetic methodologies in pharmaceutical discovery campaigns.