MMGNN: Multi-level, multi-color graph neural networks for molecular property prediction

2026-06-18

AI Summary

This work addresses the limitations of conventional molecular graph neural networks, which conflate diverse chemical interactions within a single graph and struggle to disentangle specific signals or capture long-range dependencies. To overcome these challenges, the authors propose a multi-level, multi-colored subgraph decomposition framework (MMGNN) that partitions the molecular graph into overlapping subgraphs based on atom-type pairs, enabling separate modeling of covalent topology (MMGNN-2D) or 3D spatial proximity (MMGNN-3D). The approach integrates geometric descriptors—including distances, angles, and dihedral angles—and employs subgraph-level message passing with atomic-level aggregation to enhance representation specificity while preserving atomic resolution. Evaluated on MoleculeNet benchmarks, MMGNN achieves strong performance: MMGNN-2D yields an average AUC-ROC of 0.838 on classification tasks and an RMSE of 0.803 on ESOL regression, while MMGNN-3D attains an AUC-ROC of 0.956 on BBBP and an RMSE of 1.793 on FreeSolv.

Abstract

Molecular message-passing neural networks commonly propagate chemically diverse interactions through a single graph, which may mix interaction-specific signals and require deep propagation to capture long-range effects. We introduce the Multi-level, Multi-color Graph Neural Network (MMGNN), a hierarchical framework that decomposes a molecular graph into overlapping atom-type-pair-specific subgraphs while preserving atom-level resolution. MMGNN-2D constructs chemical-colored subgraphs from covalent connectivity, whereas MMGNN-3D constructs geometric-colored subgraphs from spatial proximity and augments their edges with distance, angular, and torsional descriptors. Both variants apply a shared communicative message-passing backbone to each subgraph and combine the resulting representations through atom-wise aggregation and molecular readout. We evaluated MMGNN on five classification and three regression benchmarks from MoleculeNet using common scaffold splits and five independent runs. MMGNN-2D achieved the highest macro-average AUC-ROC of 0.838 across the classification datasets and the lowest RMSE on ESOL (0.803). MMGNN-3D obtained the highest mean AUC-ROC on BBBP (0.956) and the lowest RMSE on FreeSolv (1.793), indicating complementary strengths of topological and geometric representations. Structural and leave-one-out analyses further illustrate how the subgraph decomposition affects learned representations and atom-type-pair sensitivities. These results support overlapping interaction-specific graph decomposition as a competitive strategy for molecular property prediction.

Problem

• molecular property prediction
• graph neural networks
• message-passing
• interaction-specific signals
• long-range effects

Innovation

• multi-color graph decomposition
• hierarchical message passing
• geometric-aware GNN
• atom-type-pair subgraphs
• molecular property prediction