Development of Photochemical Strategies for the Selective Functionalization of π-Extended Systems Using Bayesian Optimization

Visible-light-mediated palladium photocatalysis has emerged as a powerful strategy for radical-mediated C–C bond formation by combining the photochemical and catalytic properties of palladium under mild conditions.Leveraging this strategy, our group developed to the best of our knowledge an unprecedented γ-selective vinylogous Heck-type alkylation of π-extended silyl enol ethers using nucleophilic alkyl radicals generated from readily available alkyl bromides, providing an efficient approach for the construction of C(sp³)–C(sp²) bonds. Despite its synthetic potential, identifying the optimal reaction conditions through the conventional one-factor-at-a-time (OFAT) strategy required approximately 40 experiments. To accelerate this process, Bayesian Optimization (BO) was implemented using the EDBO+ platform. The highest-yielding conditions within the explored chemical space were identified after only five optimization rounds (15 experiments), reducing the experimental effort by more than 60%. These findings demonstrate how machine learning can efficiently accelerate the development of innovative Pd-photocatalyzed methodologies.