Fusion Graph Attention And Knowledge Graph For Drug Target Prediction

Identification of Drug-Target Interactions(DTIs)has now become an important prerequisite for cognitive areas such as polypharmacology,drug repositioning,drug discovery,side effect prediction,and drug resistance.Secondly,experiments against undiscovered drug-target interaction biochemical assays involve substantial cost,time and challenging work,making the experimentation and confirmation of drug-target pairs has been a major challenge for many drug studies.Automated prediction methods for DTIs can not only provide valuable insights into drug mechanism of action,but also save enormous resource costs.However,many existing methods use only a limited number of structured datasets to achieve the prediction task,and in addition to the known drug targets already stored in various databases,a large number of unknown drug-target interactions are hidden in unstructured data.Therefore,this paper proposes a multi-relationship prediction approach fusing attention and medical knowledge graphs(DITGAT).Firstly,an unstructured text dataset centered on the drug dataset is established,and the genetic dataset and biological literature dataset associated with the drug dataset are selected as supplements to obtain medical triads and construct medical knowledge graphs using medical entity and relationship extraction tools.Then,we initialize the embedding representations of entities and relations,entity types and relation types in the knowledge graph using the spatial embedding model and use it as the input of the graph attention model,and train the embedding representations of entities and relations using the graph attention model as the encoder.Finally,the convolutional embedding model is used as a decoder to implement the medical relationship prediction task.Under the guidance of medical experts,the prediction results are screened and drug interaction target relationships and drug action on disease target relationships are obtained.Experimental results show that the method in this paper can effectively discover the relationships between drugs,targets and diseases and provide the mechanism of action of corresponding drugs.