Fig. 3
From: Learning symmetry-aware atom mapping in chemical reactions through deep graph matching
Workflow of AMNet for atom mapping: AMNet utilizes a combination of feature matching, molecular symmetry identification, and correspondence refinement to establish correspondences between atoms in the reactant graph and product graph. The process involves: (i) Transforming molecular structures into graphs and generating node embeddings to capture structure and features. (ii) Pairwise matching scores are computed between reactant and product embedded graphs, (iii) yielding the initial correspondence matrix. (iv) Normalizing this matrix provides matching probabilities. (v) Symmetry identification by the Weisfeiler-Lehman test