Fig. 6: Results on the case study of Lewis acid-mediated Suzuki-Miyaura cross-coupling (L-SM).
From: Reaction-conditioned generative model for catalyst design and optimization with CatDRX
a Lewis acid-mediated Suzuki-Miyaura cross-coupling reaction used in optimization of ligand of palladium catalyst8. b Predictive performance of the model on the entire dataset. The blue line represents the regression line, and the light blue area represents the 95% confidence interval for that regression. c Distribution of the dataset and predicted percent yields from different generation approaches, visualized using kernel density estimation (KDE). d Examples of ligands generated by different approaches with their predicted percent yields. e Ligand generation using optimization. The starting ligand (gray diamond) and optimized ligand (black star) are mapped onto the latent space of randomly generated ligands, colored by predicted yield (left). The optimization transformation from the starting to optimized ligand is visualized with yield predictions and searching path indicated by numbers on the latent space (right).
