Molecular Graph Representation Learning via Heterogeneous Motif Graph Construction

Zhaoning Yu, Hongyang Gao

Keywords: Molecular Graph Representation, Graph Neural Networks, Heterogeneous

Abstract: We consider feature representation learning of molecular graphs. Graph Neural Networks have been widely used in feature representation learning of molecular graphs. However, most proposed methods focus on the individual molecular graph while neglecting their connections, such as motif-level relationships. We propose a novel molecular graph representation learning method by constructing a Heterogeneous Motif graph (HM-graph) to address this issue. In particular, we build an HM-graph that contains motif nodes and molecular nodes. Each motif node corresponds to a motif extracted from molecules. Then, we propose a Heterogeneous Motif Graph Neural Network (HM-GNN) to learn feature representations for each node in the HM-graph. Our HM-graph also enables effective multi-task learning, especially for small molecular datasets. To address the potential efficiency issue, we propose an edge sampler, which significantly reduces computational resources usage. The experimental results show that our model consistently outperforms previous state-of-the-art models. Under multi-task settings, the promising performances of our methods on combined datasets shed light on a new learning paradigm for small molecular datasets. Finally, we show that our model achieves similar performances with significantly less computational resources by using our edge sampler.

One-sentence Summary: We propose a novel molecular graph representation learning method by constructing a Heterogeneous Motif graph to learn motif-level feature representations.