A review of machine learning-based methods for predicting drug-target interactions.

Area of Science:

• Computational chemistry
• Bioinformatics
• Drug discovery

Background:

• Drug-target interaction (DTI) prediction is vital for efficient drug discovery, mitigating risks and costs of experimental validation.
• Machine learning (ML) methods are increasingly crucial for accurate DTI prediction.
• Current DTI prediction relies on diverse data representations for drugs and proteins.

Purpose of the Study:

• To review and categorize machine learning-based DTI prediction methods.
• To emphasize the role of data representation in DTI prediction.
• To provide a taxonomy for deep neural network models in DTI prediction and discuss datasets and metrics.

Main Methods:

• Categorization of DTI prediction methods into traditional ML and deep learning (DL) approaches.
• Delineation of five drug and four protein data representations.
• Introduction of a novel taxonomy for deep neural network models in DTI prediction.

Main Results:

• A comprehensive overview of ML-based DTI prediction strategies.
• Analysis of data representation techniques for drugs and proteins.
• Synthesis of commonly used datasets and evaluation metrics for practical application.

Conclusions:

• Machine learning, particularly deep learning, offers powerful tools for DTI prediction.
• Standardized data representation and evaluation metrics are essential for advancing the field.
• Future research should address current challenges and explore novel methodologies in DTI prediction.