DTiGEMS+: drug-target interaction prediction using graph embedding, graph mining, and similarity-based techniques.

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Accurate in silico prediction of drug-target interactions is crucial for sustainable drug development and drug repositioning.
• Existing computational methods often yield high false-positive rates, necessitating improved prediction accuracy.

Purpose of the Study:

• To introduce DTiGEMS+, a novel computational method for predicting drug-target interactions.
• To address the limitations of current methods by reducing false-positive rates in drug-target interaction prediction.

Main Methods:

• DTiGEMS+ utilizes Graph Embedding, graph Mining, and Similarity-based techniques to predict drug-target interactions.
• It models interactions as a link prediction problem in a heterogeneous network, integrating drug-drug and target-target similarity graphs.
• The method combines graph embeddings, graph mining, and machine learning, incorporating a similarity selection and fusion algorithm.

Main Results:

• DTiGEMS+ demonstrated superior prediction performance across four benchmark datasets.
• Achieved the highest average AUPR (0.92) among state-of-the-art in silico methods.
• Reduced the error rate by 33.3% compared to the second-best performing model.

Conclusions:

• DTiGEMS+ offers a significant advancement in predicting drug-target interactions.
• The method's improved accuracy supports more efficient and reliable drug repositioning and development.
• DTiGEMS+ provides a robust computational tool for identifying novel drug-target relationships.