Improving prediction of drug-target interactions based on fusing multiple features with data balancing and feature...

Area of Science:

• Computational chemistry
• Bioinformatics
• Machine learning in drug discovery

Background:

• Drug discovery necessitates accurate prediction of drug-target interactions (DTI).
• Traditional experimental methods for DTI prediction are costly and time-consuming.
• Computational approaches, particularly machine learning, offer efficient alternatives for DTI prediction.

Purpose of the Study:

• To propose a novel multi-stage computational approach, SRX-DTI, for predicting drug-target interactions.
• To address the challenge of imbalanced datasets common in DTI prediction.
• To enhance the accuracy and efficiency of DTI prediction through optimized feature selection.

Main Methods:

• Feature extraction from protein sequences and drug fingerprints (FP2).
• Implementation of the One-SVM-US technique to handle imbalanced data.
• Utilization of a forward feature selection algorithm (FFS-RF) with a random forest classifier.
• Final DTI prediction using the XGBoost classifier on a balanced dataset with optimal features.

Main Results:

• The proposed SRX-DTI approach demonstrated superior performance compared to existing methods in DTI prediction.
• The multi-stage strategy effectively managed imbalanced data and identified optimal features.
• The combination of feature engineering, data balancing, and advanced classification yielded high predictive accuracy.

Conclusions:

• SRX-DTI offers a robust and accurate computational solution for drug-target interaction prediction.
• The methodology provides a valuable tool for accelerating the drug discovery process.
• The developed approach can significantly reduce the time and cost associated with identifying potential drug candidates.