Generic Vision and Cross-Attention for Reaction Yield Prediction

2026-08-01

AI Summary

Traditional reaction yield prediction relies on one-dimensional quantum descriptors that lack spatial information, making it difficult to accurately model molecular steric effects. This work proposes a bimodal visual cross-attention architecture that, for the first time, leverages general-purpose computer vision backbone networks to represent two-dimensional molecular topology images. These visual representations are fused with tabular physical organic data, and a cross-attention mechanism dynamically learns chemical hierarchies to focus on critical steric bottlenecks, while residual connections preserve non-spatial electronic parameters. Evaluated on a test set, the method achieves an RMSE of 5.27%, significantly outperforming baseline models based solely on quantum descriptors, and offers both higher predictive accuracy and enhanced interpretability.

Abstract

Traditional reaction yield prediction is constrained by 1D quantum descriptors that lack explicit spatial information. To address this gap, a dual-modal Vision Cross-Attention architecture is proposed, fusing tabular physical-organic data with 2D molecular topologies. Notably, it is demonstrated that a generic computer vision backbone processing simple 2D skeletal structures independently outperforms purely quantum-based baselines. By synergizing both modalities, superior predictive accuracy compared to traditional methodologies is achieved by the optimal cross-attention framework (Test RMSE = 5.27%). Through mechanistic probing, active, descriptor-guided spatial querying is observed, effectively offloading macroscopic steric identification to the visual pathway. Furthermore, a dynamic chemical hierarchy is learned by the network to heavily prioritize critical steric bottlenecks, such as the aryl halide. Concurrently, residual skip connections are utilized to protect non-spatial electronic parameters from destructive attenuation during fusion. Collectively, a scalable and highly interpretable blueprint is provided for augmenting physical chemistry with deep visual learning.

Problem

• reaction yield prediction
• spatial information
• molecular topology
• quantum descriptors
• steric effects

Innovation

• Vision Cross-Attention
• Reaction Yield Prediction
• 2D Molecular Topology
• Dual-Modal Fusion
• Steric Bottleneck Prioritization