Identification of drug-target interactions via multiple kernel-based triple collaborative matrix factorization

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning in drug discovery

Background:

• Drug-target interactions (DTIs) are crucial for targeted cancer therapies, but experimental detection is time-consuming.
• Machine learning (ML) accelerates drug screening, yet few methods effectively fuse multiple data sources for DTI prediction.
• Existing computational methods for DTI prediction often lack comprehensive data integration strategies.

Purpose of the Study:

• To develop an advanced computational method for predicting drug-target interactions (DTIs) by integrating multiple biological and chemical data sources.
• To propose a novel multiple kernel-based triple collaborative matrix factorization (MK-TCMF) model for enhanced DTI prediction.
• To improve the efficiency and accuracy of identifying potential drug candidates for cancer treatment.

Main Methods:

• Developed a multiple kernel-based triple collaborative matrix factorization (MK-TCMF) model.
• Integrated diverse data sources including chemical, biological, and clinical information using a multi-kernel learning (MKL) algorithm.
• Employed matrix decomposition to derive latent feature matrices for drug and target spaces, and a bi-projection matrix.

Main Results:

• The MK-TCMF model demonstrated superior performance in predicting DTIs compared to existing computational methods across four independent test datasets.
• The multi-kernel learning algorithm effectively regulated kernel weights, assigning the highest importance to drug side-effects and target sequence information.
• The proposed method successfully integrated multiple data modalities, leading to more accurate DTI predictions.

Conclusions:

• The MK-TCMF method offers a powerful and accurate approach for predicting drug-target interactions, significantly aiding in the identification of novel therapeutic agents.
• Effective fusion of multiple data types, particularly drug side-effects and target sequences, is critical for improving DTI prediction accuracy.
• This computational strategy can accelerate the drug discovery pipeline for targeted cancer therapies, reducing reliance on laborious experimental methods.