Spherical Message Passing for 3D Graph Networks

TL;DR

This paper introduces SphereNet, a novel 3D molecular graph learning framework that leverages spherical message passing to improve efficiency and accuracy in representing complex molecular structures.

Contribution

It proposes the spherical message passing (SMP) scheme and SphereNet, enabling efficient, scalable, and highly discriminative 3D molecular learning.

Findings

SMP reduces training complexity significantly.

SphereNet achieves superior prediction performance.

The approach scales well to large molecules.

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).