WGNN-DTA

WGNN-DTA predicts drug–target binding affinities from protein sequences and molecular SMILES using weighted graph neural networks for sequence-based drug–target affinity (DTA) and compound–protein interaction (CPI) prediction.

Key Features:

• Sequence-Based Prediction: Predicts binding affinities directly from protein sequences without requiring experimental protein structural coordinates, enabling analysis of large datasets.
• Graph Neural Network Utilization: Constructs protein and molecular graphs from sequence data and SMILES (Simplified Molecular Input Line Entry System) notation to represent structural and chemical information.
• Feature Extraction and Prediction: Applies graph neural networks to extract intricate features and contact information from the constructed graphs to predict binding affinities.
• No Multiple Sequence Alignment (MSA): Operates without multiple sequence alignment (MSA) or other complex preparatory steps, supporting large-scale virtual screening.
• Versatility: Supports both drug–target affinity (DTA) prediction and compound–protein interaction (CPI) prediction tasks.

Scientific Applications:

• Virtual Screening: Facilitates virtual screening in drug development by predicting binding affinities from sequences for candidate prioritization.
• High-Throughput Screening: Enables large-scale screening of chemical libraries when protein structural data are unavailable.
• Compound–Protein Interaction Prediction: Predicts CPI relationships using graph representations derived from sequences and SMILES.
• Early-Stage Drug Discovery: Assists early-stage compound prioritization by providing sequence-based affinity estimates.

Methodology:

Constructs protein and molecular graphs from sequence data and SMILES, inputs these graphs to weighted graph neural networks that extract features including contact information, and outputs predicted binding affinities.