Sparse hierarchical representation learning on molecular graphs

TL;DR

This paper introduces a novel hierarchical graph pooling method that incorporates edge features, significantly improving molecular property prediction accuracy and training efficiency in drug discovery benchmarks.

Contribution

It proposes two new pooling layers compatible with edge-feature graph convolutional networks, addressing a gap in hierarchical learning for molecular graphs.

Findings

Outperforms previous benchmarks on three datasets

Achieves state-of-the-art results on one dataset

Pooling improves performance and training speed in most tasks

Abstract

Architectures for sparse hierarchical representation learning have recently been proposed for graph-structured data, but so far assume the absence of edge features in the graph. We close this gap and propose a method to pool graphs with edge features, inspired by the hierarchical nature of chemistry. In particular, we introduce two types of pooling layers compatible with an edge-feature graph-convolutional architecture and investigate their performance for molecules relevant to drug discovery on a set of two classification and two regression benchmark datasets of MoleculeNet. We find that our models significantly outperform previous benchmarks on three of the datasets and reach state-of-the-art results on the fourth benchmark, with pooling improving performance for three out of four tasks, keeping performance stable on the fourth task, and generally speeding up the training process.