Learn molecular representations from large-scale unlabeled molecules for drug discovery
This work provides a novel approach for drug discovery researchers to overcome the bottleneck of scarce labeled data, enabling more effective AI models for molecular property prediction and drug interactions.
This paper addresses the challenge of creating expressive molecular representations for AI-driven drug discovery, which typically suffer from limited labeled data. The authors propose MPG, a graph-based deep learning framework that pre-trains a MolGNet model on 11 million unlabeled molecules using a self-supervised strategy. The pre-trained MolGNet achieves state-of-the-art performance across 13 benchmark datasets for various drug discovery tasks.
How to produce expressive molecular representations is a fundamental challenge in AI-driven drug discovery. Graph neural network (GNN) has emerged as a powerful technique for modeling molecular data. However, previous supervised approaches usually suffer from the scarcity of labeled data and have poor generalization capability. Here, we proposed a novel Molecular Pre-training Graph-based deep learning framework, named MPG, that leans molecular representations from large-scale unlabeled molecules. In MPG, we proposed a powerful MolGNet model and an effective self-supervised strategy for pre-training the model at both the node and graph-level. After pre-training on 11 million unlabeled molecules, we revealed that MolGNet can capture valuable chemistry insights to produce interpretable representation. The pre-trained MolGNet can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of drug discovery tasks, including molecular properties prediction, drug-drug interaction, and drug-target interaction, involving 13 benchmark datasets. Our work demonstrates that MPG is promising to become a novel approach in the drug discovery pipeline.