Molecular representation contrastive learning via transformer embedding to graph neural networks

Yunwu Liu, Ruisheng Zhang, Tongfeng Li, Jing Jiang, Jun Ma, Yongna Yuan, Ping Wang

Abstract: Highlights•MolFG integrates Transformer and graph neural networks to learn molecular features.•Several augmentation strategies are used to learn molecular representations.•Virtual nodes improve the interaction of distant nodes in the molecular network.•Visualizing molecular embedding promotes comprehension of MolFG.•The results of the experiment indicate that MolFG improves prediction performance.