MultiGranDTI: an explainable multi-granularity representation framework for drug-target interaction prediction

Drug-target interaction (DTI) prediction is a tough task with critical applications in drug repurposing and design scenarios, as it significantly reduces resource consumption and accelerates the drug discovery process. With the proliferation of experimentally measured pharmaceutical data and increasingly complex drug-target interactions, deep DTI approaches are becoming increasingly competitive due to their ability to utilize large-scale data in an end-to-end manner. It is fascinating to consider how to consolidate drug-target pair representations at different granularities to enhance deep DTI models with limited performance. The employed models typically involve solely single granular information and thus lead to a significant lack of features. Motivated by this concern, this study proposes an explainable multi-granularity representation framework for DTI prediction (MultiGranDTI). First, a hierarchical network with constraint is devised to enable the natural conversion of drug representations with different granularities, effectively integrating atomic and substructural information. Second, the 1st-order and 2nd-order sequence information of the target protein is modeled and then encoded together with the spatial information. Subsequently, several convolution layers further extract various levels of protein features. Finally, the drug and protein features are concatenated and fed into regular dense layers for interaction prediction. Comprehensive experiments reveal that MultiGranDTI achieves competitive performance in two types of tasks on four benchmark datasets. Additionally, a visualization case shows that our method is capable of efficiently identifying particular functional groups and substructures in molecules and providing a plausible explanation for the predicted results.A novel MultiGranDTI model to boost DTI prediction by comprehensively mining the multi-granularity information of compounds and proteins