High-throughput screening and machine learning prediction of Rh-catalyzed ortho-C(sp²)–H amidation of arylaldehyde hydrazones

Abstract

Arylaldehyde hydrazones are versatile building blocks for organic transformations, yet their ortho-C(sp²)-H functionalization remains underdeveloped. Herein, we report a general rhodium-catalyzed method for the direct ortho-C(sp²)-H amidation of arylaldehyde hydrazones with dioxazolones, enabled by high-throughput experimentation (HTE) and machine learning (ML). This operationally simple method affords diverse orthoamide-functionalized aryl derivatives (e.g., hydrazones, carboxylic acids, nitriles) in moderate to excellent yields with broad substrate scope (>60 examples), excellent functional group tolerance, and high chemoselectivity under mild conditions, facilitating versatile product derivatization. Moreover, a comprehensive exploration of the reaction space (1000 reactions) facilitated training of an XGBoost model using RDKit descriptors for yield prediction, achieving a mean absolute error (MAE) of 5.2% on an external validation set and demonstrating accurate predictive capability beyond the training set.