MSFDTA: Predicting drug target affinity based on multi-scale features and adaptive fusion

Accurate prediction of drug target affinity (DTA) is crucial for drug development and drug repurposing. However, most deep learning-based methods base their predictions only on the molecular structure information of the drug or target or the network information of its interactions. This may neglect the specificity of different scale features thus reducing the effectiveness of the experiment. In this study, we propose a multiscale approach that combines the molecular structure scale and interaction network scale of drugs and targets to capture feature information, and then obtain composite features through a feature adaptive fusion module. Our experimental results on two benchmark datasets show that the proposed MSFDTA outperforms the state-of-the-art methods, and the significant improvement of DTA prediction performance by combining multiscale feature information is demonstrated by ablation experiments.