ReaKE: Contrastive Molecular Representation Learning with Chemical Synthetic Knowledge Graph

Yi Wang, Shuangjia Zheng, Jiahua Rao, Yunan Luo, Yuedong Yang

Abstract: Molecular representation learning has demonstrated great promise in bridging machine learning and chemical science and in supporting novel chemical discoveries. State-of-the-art methods mostly employ graph neural networks (GNNs) with self-supervised learning (SSL) and extra chemical reaction knowledge to empower the learned embeddings. However, prior works ignore three major issues in modeling reaction data, that is abnormal energy flow, ambiguous embeddings, and sparse embedding space problems. To address these problems, we propose ReaKE, a chemical synthetic knowledge graph-driven pre-training framework for molecular representation learning. We first construct a large-scale chemical synthetic knowledge graph comprising reactants, products and reaction rules. We then propose triplet-level and graph-level contrastive learning strategies to jointly optimize the knowledge graph and molecular embeddings. Representations learned by ReaKE can capture intermolecular relationships reflected in the semantic knowledge graph and molecular structures. By comparing with other state-of-the-art methods, we show that ReaKE can achieve competitive performance on the reaction prediction pretext task and the learned representations transfer well to various downstream tasks, including reaction classification, yield prediction, and molecule property prediction. Further visualization shows that the learned representations can capture the fine-grained differences both between reactions and between molecules.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Machine Learning for Sciences (eg biology, physics, health sciences, social sciences, climate/sustainability )