ChemFusion: A Multimodal Cross-Attention Network for Reaction Yield Prediction
Machine Learning
Summary
The authors developed ChemFusion, a new kind of neural network that combines electronic information with 3D shapes to better predict outcomes of reactions involving transition metals. Their model uses a special attention method to connect overall electronic details with specific 3D arrangements of atoms. Tested on many cross-coupling reactions, ChemFusion predicted results more accurately than older models that looked at only one type of data. The network also learned to recognize and account for parts where atoms are crowded, helping explain its decisions based on physical chemistry.
Authors
Qiwei Han, Chi Zhou
Abstract
Forecasting the outcomes of transition-metal-catalyzed reactions is notoriously complex due to the interplay of diverse physical and chemical variables. A persistent computational bottleneck has been effectively merging broad electronic descriptors with the localized, three-dimensional geometry of the reactive site. To bridge this representation gap, we present ChemFusion, a hybrid neural network that fuses conventional electronic features with explicit 3D atomic coordinates. Using a cross-attention mechanism, the model enables global electronic states to dynamically attend to specific spatial constraints within un-pooled molecular point clouds. When benchmarked against a diverse library of cross-couplings, this approach delivers exceptional predictive performance, decisively surpassing traditional single-modality frameworks. Importantly, extracting the attention matrices reveals that the architecture autonomously learns to identify and penalize restrictive steric hindrances. This provides a physically grounded interpretability, demonstrating that spatially aware networks can navigate complex reaction sterics that standard statistical models typically miss.