Pre-training Molecular Graph Representation with 3D Geometry

TL;DR

This paper introduces GraphMVP, a self-supervised learning framework that leverages 3D geometric information to enhance molecular graph representations, addressing the lack of 3D data in real-world applications.

Contribution

The paper proposes a novel multi-view pre-training method that integrates 2D and 3D information for improved molecular graph encoding, supported by theoretical analysis.

Findings

GraphMVP outperforms existing SSL methods in molecular tasks.

It effectively incorporates 3D geometry to enrich 2D graph representations.

Theoretical insights justify the method's effectiveness.

Abstract

Molecular graph representation learning is a fundamental problem in modern drug and material discovery. Molecular graphs are typically modeled by their 2D topological structures, but it has been recently discovered that 3D geometric information plays a more vital role in predicting molecular functionalities. However, the lack of 3D information in real-world scenarios has significantly impeded the learning of geometric graph representation. To cope with this challenge, we propose the Graph Multi-View Pre-training (GraphMVP) framework where self-supervised learning (SSL) is performed by leveraging the correspondence and consistency between 2D topological structures and 3D geometric views. GraphMVP effectively learns a 2D molecular graph encoder that is enhanced by richer and more discriminative 3D geometry. We further provide theoretical insights to justify the effectiveness of GraphMVP. Finally, comprehensive experiments show that GraphMVP can consistently outperform existing graph SSL methods.