Leveraging Bayesian Optimization for Accelerated Ligand Selection in Atroposelective Negishi Coupling.

Area of Science:

• Catalysis
• Organic Chemistry
• Computational Chemistry

Background:

• Ligand selection is crucial for asymmetric catalysis but challenging under sparse-data conditions.
• Traditional methods like multivariate linear regression struggle with complex catalytic landscapes.
• Accelerating the discovery of efficient ligands is vital for synthesizing complex molecules.

Purpose of the Study:

• To develop and demonstrate a descriptor-informed Bayesian optimization strategy for accelerating ligand selection in asymmetric catalysis.
• To optimize yield and diastereoselectivity in a Negishi cross-coupling reaction for KRAS G12C inhibitor synthesis.
• To explore the utility of DFT-derived descriptors in guiding catalytic optimization.

Main Methods:

• Implemented a Bayesian optimization strategy informed by chemical descriptors.
• Utilized forward-step descriptor selection integrated with Density Functional Theory (DFT)-derived descriptors.
• Applied the strategy to an atroposelective Negishi cross-coupling reaction.

Main Results:

• The descriptor-informed Bayesian optimization successfully optimized yield and diastereoselectivity, outperforming multivariate linear regression.
• Identified underexploited reactivity of Phanephos ligands.
• Achieved 84% yield and 91:9 diastereomeric ratio (dr) for the target atropisomer.

Conclusions:

• Descriptor-informed Bayesian optimization is an effective strategy for accelerating ligand discovery in asymmetric catalysis.
• DFT-derived descriptors can efficiently guide optimization of complex catalytic reactions.
• The developed method enabled the synthesis of a KRAS G12C inhibitor precursor with high yield and selectivity.