RGCN

RGCN implements relational graph convolutional networks in PyTorch to model graph-structured biological data for node classification and link prediction, including protein function and drug-target interaction prediction.

Key Features:

• Relational Data Handling: Manages graphs with multiple relationship types to represent biological networks.
• Node Classification and Link Prediction: Performs node classification and link prediction validated on Knowledge Graph datasets for protein function and drug-target interaction prediction.
• Parameter Efficiency: Provides parameter-efficient configurations that reduce computational overhead for large-scale biological datasets.
• Reproducibility: Empirically validated on benchmark datasets to support reproducible research.
• PyTorch Implementation: Implemented in PyTorch for model development and experimentation.

Scientific Applications:

• Protein-Protein Interaction Networks: Models protein-protein interactions with multiple relationship types.
• Gene Regulatory Networks: Analyzes diverse gene–gene interactions within regulatory networks.
• Drug Discovery: Predicts drug-target interactions using relational graph data.

Methodology:

Applies Graph Convolutional Networks with specialized relational convolutions and is implemented in PyTorch.