DeepPurpose

DeepPurpose predicts drug-target interactions using deep learning to model relationships between compounds and proteins for drug discovery.

Key Features:

• Extensive Model Customization: Implements 15 distinct compound and protein encoders enabling varied molecular representation strategies.
• Diverse Neural Architectures: Provides over 50 neural network architectures for constructing models that capture complex biological patterns.
• Multi-encoder Integration: Integrates multiple compound and protein encoders to capture intricate molecular features for prediction.
• Benchmark Performance: Demonstrated state-of-the-art results on several benchmark datasets for drug-target interaction prediction.

Scientific Applications:

• Drug-target interaction (DTI) prediction: Predicts interactions between compounds and protein targets to support identification of therapeutic compounds.
• Protein function prediction: Applies learned protein representations to infer protein functional properties.
• Drug property analysis: Uses compound encodings to analyze drug properties.
• Protein-protein interaction (PPI) analysis: Can be adapted to model interactions between proteins.
• Drug-drug interaction (DDI) analysis: Can be adapted to model interactions between drugs.

Methodology:

Leverages advanced deep learning techniques by implementing 15 compound and protein encoders, integrating multiple encoders, and employing over 50 neural network architectures to encode biological data into representations for interaction prediction.