NDD

NDD predicts unknown drug-drug interactions (DDIs) by integrating multiple drug similarity measures and applying a two-layer fully connected neural network for DDI prediction.

Key Features:

• Comprehensive Drug Characterization: Represents drugs using drug substructure, target, side effects (standard and off-label), pathways, transporters, and indications.
• Similarity Calculation: Computes multiple drug similarity matrices derived from the listed drug characteristics.
• Heuristic Similarity Selection: Selects relevant similarity measures from the computed set using heuristic methods.
• Non-linear Similarity Fusion (SNF): Integrates selected similarities using Similarity Network Fusion (SNF) to produce fused similarity profiles.
• Neural Network Architecture: Uses a two-layer fully connected neural network to predict DDIs from fused features.
• Benchmarking and Performance Metrics: Evaluated against six machine learning classifiers and six state-of-the-art graph-based methods across three benchmark datasets with reported AUPR 0.830–0.947, AUC 0.954–0.994, and F-measure 0.772–0.902.
• Case Study Validation: Includes case studies on numerous drug pairs that corroborate the method's predictive capability.

Scientific Applications:

• Drug development: Predicts unknown DDIs to inform safety and efficacy assessments during drug development.
• Clinical safety assessment: Identifies potential adverse drug interactions to enhance patient safety and optimize therapeutic outcomes in healthcare settings.

Methodology:

Calculate multiple drug similarities from drug substructure, targets, side effects, pathways, transporters, and indications; select relevant similarities via heuristic methods; integrate selected similarities using Similarity Network Fusion (SNF); input fused features to a two-layer fully connected neural network for DDI prediction; evaluate against six machine learning classifiers and six graph-based methods on three benchmark datasets reporting AUPR, AUC, and F-measure metrics.