DTI-RME: a robust and multi-kernel ensemble approach for drug-target interaction prediction

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Drug-target interactions (DTIs) are crucial for understanding drug mechanisms.
• Computational methods offer efficient alternatives to experimental DTI prediction.
• Existing DTI prediction methods face challenges with noisy labels, multi-view fusion, and structural modeling.

Purpose of the Study:

• To develop a novel computational method for accurate drug-target interaction prediction.
• To address limitations in current DTI prediction models, including noisy data and ineffective fusion techniques.

Main Methods:

• Proposed DTI-RME method utilizing a novel L2-C loss function for reduced errors and outlier handling.
• Employed multi-kernel learning for effective fusion of multiple data views.
• Applied ensemble learning to model diverse structures, including drug-target pairs, individual drugs and targets, and low-rank representations.

Main Results:

• DTI-RME demonstrated superior performance across five real-world DTI datasets and three key experimental scenarios.
• The method effectively handled noisy interaction labels and integrated multi-view information.
• Ensemble learning contributed to a comprehensive understanding of interaction structures.

Conclusions:

• DTI-RME significantly outperforms existing methods in drug-target interaction prediction.
• A case study validated DTI-RME's capability to accurately identify novel drug-target interactions, with 17 of the top 50 predictions confirmed.
• The findings highlight DTI-RME's potential for accelerating drug discovery.