MCGM: Multi-stage Clustered Global Modeling for Long-range Interactions in Molecules

2025-09-26

AI Summary

Geometric graph neural networks (GNNs) struggle to capture long-range molecular interactions due to their inherent local message-passing mechanism. To address this, we propose a multi-stage dynamic hierarchical clustering framework that constructs multi-resolution atomic clusters, enabling lightweight global contextual modeling and iterative feature refinement. Our approach is the first to jointly integrate learnable clustering, residual connections, and geometric GNNs—without requiring enlarged cutoff radii, intricate physics-based kernels, or Fourier basis functions—ensuring plug-and-play compatibility. On the OE62 dataset, our method reduces energy prediction error by 26.2%; on AQM, it achieves state-of-the-art accuracy of 17.0 meV (energy) and 4.9 meV/Å (force), while reducing model parameters by 20%.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Geometric graph neural networks (GNNs) excel at capturing molecular geometry, yet their locality-biased message passing hampers the modeling of long-range interactions. Current solutions have fundamental limitations: extending cutoff radii causes computational costs to scale cubically with distance; physics-inspired kernels (e.g., Coulomb, dispersion) are often system-specific and lack generality; Fourier-space methods require careful tuning of multiple parameters (e.g., mesh size, k-space cutoff) with added computational overhead. We introduce Multi-stage Clustered Global Modeling (MCGM), a lightweight, plug-and-play module that endows geometric GNNs with hierarchical global context through efficient clustering operations. MCGM builds a multi-resolution hierarchy of atomic clusters, distills global information via dynamic hierarchical clustering, and propagates this context back through learned transformations, ultimately reinforcing atomic features via residual connections. Seamlessly integrated into four diverse backbone architectures, MCGM reduces OE62 energy prediction error by an average of 26.2%. On AQM, MCGM achieves state-of-the-art accuracy (17.0 meV for energy, 4.9 meV/Å for forces) while using 20% fewer parameters than Neural P3M. Code will be made available upon acceptance.

Problem

• Modeling long-range molecular interactions with geometric graph neural networks
• Overcoming computational cost and parameter tuning limitations in existing methods
• Enhancing molecular property prediction accuracy through hierarchical global context

Innovation

• Hierarchical atomic clustering for global context
• Dynamic clustering distills multi-resolution molecular information
• Plug-and-play module enhances geometric GNNs via residuals