NPI-GNN

NPI-GNN predicts ncRNA-protein interactions using graph neural networks by integrating network topology and sequence data to study ncRNA roles in gene regulation and disease.

Key Features:

• Graph Neural Network Architecture: Uses GNNs to predict ncRNA-protein interactions and represents the first application of GNNs for this task.
• Benchmarking and Validation: Evaluated on five benchmark datasets using 5-fold cross-validation and achieved accuracy comparable to existing methods.
• Robustness to Sequence Information Variability: Maintains high predictive performance with incomplete or noisy sequence data.
• Prediction of Novel Interactions: Integrates network topology and sequence information to predict previously unknown ncRNA-protein interactions.
• End-to-End Predictive Model: Implements an end-to-end GNN-based predictor for NPIs from data input to interaction prediction.

Scientific Applications:

• ncRNA–protein interaction prediction: Enables accurate prediction of ncRNA-protein interactions to aid studies of gene regulation and ncRNA roles in biological processes and disease.
• Functional characterization and therapeutic target identification: Supports exploration of ncRNA function and identification of therapeutic targets.

Methodology:

Integrates network topology and sequence data using graph neural networks to capture complex patterns in biological networks and enable predictions with incomplete or noisy data.