Drug-Target Interaction Prediction Based on Drug Subgraph Fingerprint Extraction Strategy and Subgraph Attention Mechanism

Lizhi Wang, Xiaolong Zhang, Xiaoli Lin, Jing Hu

Abstract

Drug discovery is a major focus of modern research, and predicting drug-target interactions is one of the strategies to facilitate this research process. Traditional laboratory methods have long time cycles and are relatively costly, so the use of high-precision virtual screening is essential. Previous virtual screens have only considered the sequence structure of the drug or the graph structure of the drug, without considering the role of the drug subgraph structure. Therefore, this paper adopts a method for extracting drug subgraphs and designs a strategy for extracting drug subgraph fingerprints. Then, the fingerprint vector data of the drugs are trained by the graph neural network to produce drug information vectors containing the spatial structure of the drug subgraph. Finally, the subgraph attention mechanism is used to update the information on protein targets, facilitating the model to extract more information about the target site. The experimental results show that the model can predict drug-target interactions well, outperforming the state-of-the-art model in all metrics. And it is also effective in both classification prediction and affinity regression prediction.
BibTeX
@inproceedings{wang2023drug,
  author = {Lizhi Wang and Xiaolong Zhang and Xiaoli Lin and Jing Hu},
  title = {{Drug-Target Interaction Prediction Based on Drug Subgraph Fingerprint Extraction Strategy and Subgraph Attention Mechanism}},
  booktitle = {ADMA},
  year = {2023},
  doi = {10.1007/978-3-031-46671-7\_1},
}

Loading related content...

Figure