Spherical Message Passing for 3D Graph Networks

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).

The paper has been accepted by ICLR 2022. You can also cite the conference version

References in corpus (1)

• Deep Learning of High-Order Interactions for Protein Interface Prediction

Cited by in corpus (9)

• Pre-training Molecular Graph Representation with 3D Geometry
• 3D Infomax improves GNNs for Molecular Property Prediction
• DIG: A Turnkey Library for Diving into Graph Deep Learning Research
• Rotation Invariant Graph Neural Networks using Spin Convolutions
• Learning 3D Representations of Molecular Chirality with Invariance to Bond Rotations
• Geometric Transformers for Protein Interface Contact Prediction
• Geometric and Physical Quantities Improve E(3) Equivariant Message Passing
• Molecule3D: A Benchmark for Predicting 3D Geometries from Molecular Graphs
• Fast Quantum Property Prediction via Deeper 2D and 3D Graph Networks