posted on 2022-01-24, 19:35authored byRyan C. Cammarota, Wenbin Liu, John Bacsa, Huw M. L. Davies, Matthew S. Sigman
Leveraging congested catalyst scaffolds
has emerged as a key strategy
for altering innate substrate site-selectivity profiles in C–H
functionalization reactions. Similar to enzyme active sites, optimal
small molecule catalysts often feature reactive cavities tailored
for controlling substrate approach trajectories. However, relating
three-dimensional catalyst shape to reaction output remains a formidable
challenge, in part due to the lack of molecular features capable of
succinctly describing complex reactive site topologies in terms of
numerical inputs for machine learning applications. Herein, we present
a new set of descriptors, “Spatial Molding for Approachable
Rigid Targets” (SMART), which we have applied to quantify reactive
site spatial constraints for an expansive library of dirhodium catalysts
and to predict site-selectivity for C–H functionalization of
1-bromo-4-pentylbenzene via donor/acceptor carbene intermediates.
Optimal site-selectivity for the terminal methylene position was obtained
with Rh2(S-2-Cl-5-MesTPCP)4 (30.9:1 rr, 14:1 dr, 87% ee), while C–H functionalization
at the electronically activated benzylic site was increasingly favored
for Rh2(TPCP)4 catalysts lacking an ortho-Cl, Rh2(S-PTAD)4, and Rh2(S-TCPTAD)4, respectively.
Intuitive global site-selectivity models for 25 disparate dirhodium
catalysts were developed via multivariate linear regression to explicitly
assess the contributing roles of steric congestion and dirhodium-carbene
electrophilicity in controlling the site of C–H functionalization.
The workflow utilizes spatial classification to extract descriptors
only for reactive catalyst conformers, a nuance that may be widely
applicable for establishing close correspondence between ground-state
model systems and transition states. Broader still, SMART descriptors
are amenable for delineating salient reactive site features to predict
reactivity in other chemical and biological contexts.