Spherical Message Passing for 3D Graph Networks

2021/02/09 by Yi Liu, Limei Wang, Liu, Yi +9 · 13 citations
Computer Science · Physics and Astronomy · #Advanced Graph Neural Networks #Complex Network Analysis Techniques #FOS: Computer and information sciences #Graph Theory and Algorithms #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.2102.05013

openalex publication_date 2021/02/09 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We consider representation learning of 3D molecular graphs in which each atom is associated with a spatial position in 3D. This is an under-explored area of research, and a principled message passing framework is currently lacking. In this work, we conduct analyses in the spherical coordinate system (SCS) for the complete identification of 3D graph structures. Based on such observations, we propose the spherical message passing (SMP) as a novel and powerful scheme for 3D molecular learning. SMP dramatically reduces training complexity, enabling it to perform efficiently on large-scale molecules. In addition, SMP is capable of distinguishing almost all molecular structures, and the uncovered cases may not exist in practice. Based on meaningful physically-based representations of 3D information, we further propose the SphereNet for 3D molecular learning. Experimental results demonstrate that the use of meaningful 3D information in SphereNet leads to significant performance improvements in prediction tasks. Our results also demonstrate the advantages of SphereNet in terms of capability, efficiency, and scalability. Our code is publicly available as part of the DIG library (https://github.com/divelab/DIG).

Citations

• Neural Message Passing for Quantum Chemistry
• Detecting Functionality-Specific Vulnerabilities via Retrieving Individual Functionality-Equivalent APIs in Open-Source Repositories

Cited by

• Dual-Level Atomic and Coordination Geometry Learning for Crystal Property Prediction Using Graph Neural Networks
• Multi-Scale Protein Structure Modelling with Geometric Graph U-Nets
• VS-Graph: Scalable and Efficient Graph Classification Using Hyperdimensional Computing
• Toward bilipshiz geometric models
• Pre-training Molecular Graph Representation with 3D Geometry
• Universally Invariant Learning in Equivariant GNNs
• MolGA: Molecular Graph Adaptation with Pre-trained 2D Graph Encoder
• Physics-Inspired All-Pair Interaction Learning for 3D Dynamics Modeling
• Shoot from the HIP: Hessian Interatomic Potentials without derivatives
• Learning 3D Representations of Molecular Chirality with Invariance to Bond Rotations
• DEQuify your force field: More efficient simulations using deep equilibrium models
• Challenges in Non-Polymeric Crystal Structure Prediction: Why a Geometric, Permutation-Invariant Loss is Needed
• Data-driven particle dynamics: Structure-preserving coarse-graining for emergent behavior in non-equilibrium systems

Related

• ComENet: Towards Complete and Efficient Message Passing for 3D Molecular Graphs
• 3D Molecular Geometry Analysis with 2D Graphs
• GemNet: Universal Directional Graph Neural Networks for Molecules
• Molecule3D: A Benchmark for Predicting 3D Geometries from Molecular Graphs
• A Hitchhiker's Guide to Geometric GNNs for 3D Atomic Systems