LAPINE

LAPINE predicts adverse drug reaction (ADR)-related proteins by applying network embedding and a single-target compound concept to identify off-target proteins implicated in ADRs across the proteome.

Key Features:

• Network Embedding Technique: Employs a network embedding approach that integrates the single-target compound concept for representation and analysis of protein–protein interaction networks at large scale.
• Single-Target Compound Concept: Incorporates single-target compound information to enhance specificity and accuracy in identifying off-target interactions that may lead to ADRs.
• Protein–Protein Interaction Network Analysis: Analyzes protein–protein interaction (PPI) networks across the proteome to predict potential ADR-related proteins.
• Predictive Reliability: Demonstrates improved predictive performance with a value-added positive predictive value (PPV) of 0.12 and statistical significance (P < 0.001) compared to previous methodologies.
• Benchmark Dataset Analysis: Validates predictions using benchmark datasets to justify broadening the range of potential ADR-related proteins under investigation.
• Literature-Backed Predictions: Includes case studies showing that most predicted proteins are corroborated by existing literature.

Scientific Applications:

• ADR Mechanism Investigation: Identifies candidate off-target proteins to help elucidate molecular mechanisms underlying adverse drug reactions.
• Drug Safety Assessment: Supports identification of off-target interactions relevant to drug safety and risk mitigation during drug development.
• Proteome-Wide Screening: Enables large-scale, proteome-wide prediction of proteins potentially associated with ADRs.
• Hypothesis Generation and Validation: Produces predictions that can guide experimental follow-up and be corroborated against literature and benchmark datasets.

Methodology:

Uses network embedding that integrates the single-target compound concept to analyze protein–protein interaction networks across the proteome; predictions were evaluated on benchmark datasets with reported PPV = 0.12 (P < 0.001) and supported by literature-backed case studies.