GIFDTI

GIFDTI predicts drug-target interactions using deep learning to represent molecular structures and model intermolecular interactions for drug discovery and drug repurposing.

Key Features:

• Sequence Feature Extraction: CNNFormer integrates Convolutional Neural Networks (CNN) with Transformer architectures to capture local chemical patterns and long-distance relationships among atoms in drug molecules or amino acids in protein sequences.
• Global Molecular Feature Extraction (GF): The Global Feature component extracts comprehensive molecular features that encapsulate both local and global chemical environments.
• Intermolecular Interaction Modeling (IIF): The Intermolecular Interaction Feature module models interactions between drugs and proteins to inform DTI prediction.

Scientific Applications:

• Predicting drug-target interactions: Evaluated across six realistic evaluation strategies and reported to outperform state-of-the-art methods in DTI prediction.
• Drug repurposing: Identifying potential new uses for existing drugs by predicting relevant DTIs.
• Novel drug discovery: Supporting the discovery of novel drug candidates through predicted drug-protein interactions.
• Computational biology and pharmacology: Applied to studies in computational biology and pharmacology, including efforts to predict low-cost DTIs.

Methodology:

GIFDTI integrates three core components—CNNFormer, GF, and IIF—combining convolutional and transformer-based approaches to capture molecular details and long-range sequence dependencies, while the GF extractor encodes local and global structural information and the IIF module models drug–protein interactions.