Prediction of Drug-Target Interaction Using Dual-Network Integrated Logistic Matrix Factorization and Knowledge Graph...

Area of Science:

• Bioinformatics
• Computational Chemistry
• Pharmacology

Background:

• Drug-target interactions (DTIs) prediction is crucial for drug repositioning.
• Existing models often neglect prior knowledge or fail to predict interactions for under-studied drugs/targets.
• There is a need for advanced DTIs prediction methods that incorporate prior knowledge effectively.

Purpose of the Study:

• To propose a novel dual-network integrated logistic matrix factorization (DNILMF) scheme for DTIs prediction, named Ro-DNILMF.
• To integrate prior knowledge into DTIs prediction models using a knowledge graph embedding approach.
• To enhance the prediction of DTIs for under-studied drugs and targets.

Main Methods:

• A knowledge graph embedding model (RotatE) was employed to construct an interaction adjacency matrix and integrate prior knowledge.
• A dual-network integrated logistic matrix factorization (DNILMF) model was utilized for predicting new drug-target interactions.
• The Ro-DNILMF scheme combines knowledge graph embeddings with DNILMF for improved DTIs prediction.

Main Results:

• The Ro-DNILMF method successfully integrates prior knowledge into the DTIs prediction model.
• The proposed model demonstrates the capability to predict interactions for under-studied drugs and targets.
• Experimental results show that Ro-DNILMF outperforms baseline and mainstream methods in terms of efficiency.

Conclusions:

• Ro-DNILMF offers an effective approach for drug-target interaction prediction by leveraging prior knowledge.
• The method addresses limitations of existing models in handling under-studied drugs and targets.
• Ro-DNILMF shows significant promise for advancing drug repositioning and drug discovery efforts.