An ensemble-based drug-target interaction prediction approach using multiple feature information with data balancing.

Area of Science:

• Computational chemistry and bioinformatics
• Drug discovery and development

Background:

• Drug repositioning is a cost-effective strategy in pharmaceutical development.
• Artificial intelligence (AI) aids in identifying drug profiles, side effects, and targets.
• Increasing drug data presents challenges with imbalanced datasets for prediction models.

Purpose of the Study:

• To propose a novel scheme for predicting drug-target interactions (DTIs).
• To address the issue of imbalanced data in drug-target interaction prediction.
• To leverage drug chemical structures and protein sequences for enhanced prediction.

Main Methods:

• Feature extraction using drug Morgan fingerprint, constitutional descriptors, and protein composition (amino acid and dipeptide).
• Development of a support vector machine one-class classifier for negative sample extraction to handle data imbalance.
• Construction of positive and negative samples for prediction algorithms and evaluation using 10-fold cross-validation.

Main Results:

• The proposed model demonstrated superior performance over existing methods in metrics including AUC, accuracy, precision, recall F-score, MSE, and MCC.
• AdaBoost classifier improved prediction accuracy by 2.74%, precision by 1.98%, AUC by 1.14%, F-score by 3.53%, and MCC by 4.54% compared to existing techniques.

Conclusions:

• The novel scheme effectively predicts drug-target interactions by utilizing chemical and physical features.
• The method successfully tackles imbalanced data issues, leading to improved prediction performance.
• This approach offers a valuable tool for accelerating drug discovery and development through accurate DTI prediction.