CFSBoost: Cumulative feature subspace boosting for drug-target interaction prediction

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Experimental drug target interaction prediction is costly and time-consuming.
• In silico methods are crucial for identifying potential drug-target interactions efficiently.
• Existing computational approaches require improvement in accuracy and cost-effectiveness.

Purpose of the Study:

• To introduce CFSBoost, a novel computational model for drug-target interaction prediction.
• To develop a computationally inexpensive yet high-performing classification model.
• To leverage evolutionary and structural features for enhanced prediction accuracy.

Main Methods:

• CFSBoost employs an ensemble boosting classification strategy.
• It utilizes extra trees as weak learners within a boosting framework.
• A novel feature group selection procedure ensures computational efficiency.

Main Results:

• CFSBoost achieved superior performance on benchmark datasets, outperforming existing methods.
• The model demonstrated high accuracy in predicting drug-target interactions, particularly on imbalanced datasets.
• CFSBoost showed improved area under the precision-recall curve (auPR) on 3 out of 4 datasets.

Conclusions:

• CFSBoost offers a computationally cheap and effective solution for drug-target interaction prediction.
• The model's performance is satisfactory and competitive with state-of-the-art methods.
• CFSBoost provides valuable predictions for potential new drug-target interactions, facilitating further research.