Distance-Geometric Graph Attention Network (DG-GAT) for 3D Molecular Geometry

TL;DR

This paper introduces DG-GAT, a novel 3D graph attention network that leverages distance geometry for improved molecular property prediction, demonstrating significant performance gains over traditional 2D graph methods.

Contribution

The paper proposes DG-GAT, integrating distance-geometric representation with a dynamic attention mechanism for 3D molecular graphs, advancing deep learning in molecular science.

Findings

Major improvement (31% and 38%) over 2D graph methods on ESOL and FreeSolv datasets.

Effective utilization of 3D geometry in molecular property prediction.

Demonstrates the utility of DG-GAT for various molecular datasets.

Abstract

Deep learning for molecular science has so far mainly focused on 2D molecular graphs. Recently, however, there has been work to extend it to 3D molecular geometry, due to its scientific significance and critical importance in real-world applications. The 3D distance-geometric graph representation (DG-GR) adopts a unified scheme (distance) for representing the geometry of 3D graphs. It is invariant to rotation and translation of the graph, and it reflects pair-wise node interactions and their generally local nature, particularly relevant for 3D molecular geometry. To facilitate the incorporation of 3D molecular geometry in deep learning for molecular science, we adopt the new graph attention network with dynamic attention (GATv2) for use with DG-GR and propose the 3D distance-geometric graph attention network (DG-GAT). GATv2 is a great fit for DG-GR since the attention can vary by node and by distance between nodes. Experimental results of DG-GAT for the ESOL and FreeSolv datasets show major improvement (31% and 38%, respectively) over those of the standard graph convolution network based on 2D molecular graphs. The same is true for the QM9 dataset. Our work demonstrates the utility and value of DG-GAT for deep learning based on 3D molecular geometry.