An Ensemble Learning-Based Method for Inferring Drug-Target Interactions Combining Protein Sequences and Drug...

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Identifying drug-target interactions (DTIs) is crucial for drug discovery and repositioning.
• High-throughput biotechnologies and clinical trials are expensive, laborious, and intricate.
• Development of convenient and reliable computer-aided methods for inferring DTIs is a research focus.

Purpose of the Study:

• To propose a novel computational model for identifying drug-target interactions (DTIs).
• To integrate Position-Specific Scoring Matrix (PSSM), pyramid histogram of oriented gradients (PHOG), and rotation forest (RF) for DTI prediction.
• To provide an efficient and reliable computational tool for large-scale DTI prediction.

Main Methods:

• Protein primary sequences converted into PSSMs to capture biological evolution information.
• PHOG used to extract representative features from PSSMs at multiple pyramid levels.
• Combined molecular substructure fingerprints and PHOG features as complete descriptors for drug-target pairs, classified using RF.

Main Results:

• Achieved mean accuracies of 88.96% (enzyme), 86.37% (ion channel), 82.88% (GPCRs), and 76.92% (nuclear receptor) via 5-fold Cross-Validation.
• Performance validated against state-of-the-art Light Gradient Boosting Machine (LGBM) and Support Vector Machine (SVM).
• Experimental outcomes demonstrate the model's feasibility and reliability in predicting DTIs.

Conclusions:

• The proposed computational model integrating PSSM, PHOG, and RF is effective for DTI prediction.
• The model offers a feasible and reliable alternative to traditional methods, reducing cost and complexity.
• This approach shows excellent prospects for large-scale DTI prediction in drug discovery and repositioning.