BRWCP

BRWCP predicts drug-protein interactions by applying a bidirectional random walk with pruning on heterogeneous networks to correct for incomplete information in molecular and clinical datasets.

Key Features:

• Correction for Incomplete Information: Distinguishes interaction versus non-interaction (or correlation versus non-correlation) within datasets and corrects predictions using a complete-information network.
• Network-Based Approach: Constructs an initial heterogeneous network from known DPIs and feature similarities of drugs and proteins.
• Bidirectional Random Walk with Pruning Algorithm: Employs a bidirectional random walk with pruning to traverse the heterogeneous network and integrate chemical fingerprint similarities of drugs and amino acid sequence similarities of proteins for prediction scoring.
• Iterative Refinement: Reconstructs a complete-information network using additional similarity measures and reapplies the algorithm iteratively until convergence.

Scientific Applications:

• Drug discovery and development: Prioritizes potential drug-protein interactions to inform target identification and lead selection.
• Drug repurposing and candidate identification: Identifies novel or previously unrecognized interactions to support repurposing or discovery of promising drug candidates.

Methodology:

Integrate heterogeneous information sources to derive feature similarities among drugs and proteins; combine these similarities with known DPIs to construct an initial incomplete-information network; apply a bidirectional random walk with pruning to predict interactions; build a complete-information network using chemical fingerprint similarities for drugs and amino acid sequence similarities for proteins; iteratively reapply the algorithm until convergence.