PreDTIs: prediction of drug-target interactions based on multiple feature information using gradient boosting...

Area of Science:

• Bioinformatics
• Computational Chemistry
• Drug Discovery

Background:

• Drug-target interactions (DTIs) are crucial for novel drug development.
• Experimental DTI prediction is costly and time-consuming.
• Machine learning methods face challenges with data imbalance and high dimensionality.

Purpose of the Study:

• To develop a novel, accurate, and efficient computational method for predicting DTIs.
• To address data imbalance and feature dimensionality issues in DTI prediction.
• To enhance the discovery of new drug candidates and understand drug mechanisms.

Main Methods:

• Feature extraction using pseudo-position-specific scoring matrix (PsePSSM), dipeptide composition (DC), and pseudo amino acid composition (PseAAC) for proteins.
• Drug encoding using MACCS substructure fingerprinting.
• Application of FastUS algorithm for class imbalance and Minimum Redundancy Maximum Relevance (mRMR) feature selection for optimal feature identification.
• Classification using the LightGBM model with 5-fold cross-validation.

Main Results:

• The proposed PreDTIs method demonstrates superior performance compared to existing DTI prediction methods.
• The FastUS and mRMR algorithms effectively handle data imbalance and reduce feature dimensionality.
• Optimized features and the LightGBM classifier yield high prediction accuracy for DTIs.

Conclusions:

• PreDTIs offers a significant advancement in computational DTI prediction.
• The method can accelerate drug discovery by identifying potential drug-target pairs.
• PreDTIs has potential applications in discovering treatments for diseases like COVID-19.