DeepStack-DTIs: Predicting Drug-Target Interactions Using LightGBM Feature Selection and Deep-Stacked Ensemble...

Area of Science:

• Computational biology
• Drug discovery and development
• Bioinformatics

Background:

• Accurate prediction of drug-target interactions (DTIs) is crucial for drug discovery and repositioning.
• Existing methods face challenges in effectively integrating diverse molecular features for DTI prediction.

Purpose of the Study:

• To propose DeepStack-DTIs, a novel deep-stacked ensemble method for enhancing DTI prediction accuracy.
• To leverage advanced feature extraction and machine learning techniques for improved DTI prediction.

Main Methods:

• Feature extraction using pseudo-position specific score matrix (PsePSSM), pseudo amino acid composition (PseAAC), SPIDER3 for proteins, and fingerprint features (FP2) for drugs.
• Data balancing with Synthetic Minority Oversampling Technique (SMOTE) and feature selection using Light Gradient Boosting Machine (LightGBM).
• Ensemble classification integrating Gated Recurrent Unit (GRU), Deep Neural Network (DNN), Support Vector Machine (SVM), eXtreme Gradient Boosting (XGBoost), and Logistic Regression (LR).

Main Results:

• DeepStack-DTIs demonstrated higher prediction accuracy compared to existing methods on a gold standard dataset.
• Validation on an additional dataset confirmed the method's excellent predictive ability and potential for predicting drug-target interaction networks.
• The approach effectively fuses diverse molecular data and utilizes advanced ensemble learning for superior DTI prediction.

Conclusions:

• DeepStack-DTIs offers a significant improvement in predicting drug-target interactions.
• The method provides novel insights and a powerful tool for drug discovery and repositioning.
• The fusion of multiple feature types and the deep-stacked ensemble approach contribute to its enhanced predictive performance.