Directed Graph Attention Neural Network Utilizing 3D Coordinates for Molecular Property Prediction

TL;DR

This paper introduces DGANN, a graph attention neural network that uses chemical bonds and 3D coordinates for molecular property prediction, achieving high accuracy without requiring fully connected graphs or invariance to rotations and translations.

Contribution

The work presents a novel directed graph attention neural network that leverages 3D coordinates and bond information, differing from previous models that rely on pairwise distances and fully connected graphs.

Findings

DGANN matches or outperforms baseline GNNs on QM9 datasets.

Utilizing 3D coordinates without invariance still yields high accuracy.

Model operates effectively with only chemical bonds as edges.

Abstract

The prosperity of computer vision (CV) and natural language procession (NLP) in recent years has spurred the development of deep learning in many other domains. The advancement in machine learning provides us with an alternative option besides the computationally expensive density functional theories (DFT). Kernel method and graph neural networks have been widely studied as two mainstream methods for property prediction. The promising graph neural networks have achieved comparable accuracy to the DFT method for specific objects in the recent study. However, most of the graph neural networks with high precision so far require fully connected graphs with pairwise distance distribution as edge information. In this work, we shed light on the Directed Graph Attention Neural Network (DGANN), which only takes chemical bonds as edges and operates on bonds and atoms of molecules. DGANN distinguishes from previous models with those features: (1) It learns the local chemical environment encoding by graph attention mechanism on chemical bonds. Every initial edge message only flows into every message passing trajectory once. (2) The transformer blocks aggregate the global molecular representation from the local atomic encoding. (3) The position vectors and coordinates are used as inputs instead of distances. Our model has matched or outperformed most baseline graph neural networks on QM9 datasets even without thorough hyper-parameters searching. Moreover, this work suggests that models directly utilizing 3D coordinates can still reach high accuracies for molecule representation even without rotational and translational invariance incorporated.