Self-Supervised Graph Transformer on Large-Scale Molecular Data

• DOI: 10.48550/arxiv.2007.02835
• Publication Date: 2020-01-01
• Authors: Rong, Yu; Bian, Yatao; Xu, Tingyang; Xie, Weiyang; Wei, Ying; Huang, Wenbing; Huang, Junzhou
• Keywords: FOS: Computer and information sciences; Computer Science - Machine Learning; J.3; Quantitative Biology - Biomolecules; FOS: Biological sciences; 0202 electrical engineering, electronic engineering, information engineering; Biomolecules (q-bio.BM); 02 engineering and technology; I.2.0; J.3; I.2.0; Machine Learning (cs.LG)

How to obtain informative representations of molecules is a crucial prerequisite in AI-driven drug design and discovery. Recent researches abstract molecules as graphs and employ Graph Neural Networks (GNNs) for molecular representation learning. Nevertheless, two issues impede the usage of GNNs in real scenarios: (1) insufficient labeled molecules for supervised training; (2) poor generalization capability to new-synthesized molecules. To address them both, we propose a novel framework, GROVER, which stands for Graph Representation frOm self-superVised mEssage passing tRansformer. With carefully designed self-supervised tasks in node-, edge- and graph-level, GROVER can learn rich structural and semantic information of molecules from enormous unlabelled molecular data. Rather, to encode such complex information, GROVER integrates Message Passing Networks into the Transformer-style architecture to deliver a class of more expressive encoders of molecules. The flexibility of GROVER allows it to be trained efficiently on large-scale molecular dataset without requiring any supervision, thus being immunized to the two issues mentioned above. We pre-train GROVER with 100 million parameters on 10 million unlabelled molecules -- the biggest GNN and the largest training dataset in molecular representation learning. We then leverage the pre-trained GROVER for molecular property prediction followed by task-specific fine-tuning, where we observe a huge improvement (more than 6% on average) from current state-of-the-art methods on 11 challenging benchmarks. The insights we gained are that well-designed self-supervision losses and largely-expressive pre-trained models enjoy the significant potential on performance boosting.<br/>17 pages, 7 figures<br/>