Hierarchical graph representation learning for the prediction of drug-target binding affinity

Computationally predicting drug-target binding affinity (DTA) has attracted increasing attention due to its benefit for accelerating drug discovery. Currently, numerous deep learning-based prediction models have been proposed, often with a biencoder architecture that commonly focuses on how extract expressive representations drugs and targets but overlooks modeling explicit interactions. However, known DTA can provide underlying knowledge about the interact is beneficial predictive accuracy. In this paper, we propose novel hierarchical graph representation learning model prediction, named HGRL-DTA. The main contribution of our establish integrate coarse- fine-level information from an drug/target molecule graphs, respectively, in well-designed coarse-to-fine manner. addition, design similarity-based inference method infer coarse-level when it unavailable new or under cold start scenario. Comprehensive experimental results four scenarios across two benchmark datasets indicate HGRL-DTA outperforms state-of-the-art almost all cases.