Drug Targets

Drug Targets predicts drug–target interactions using knowledge graph embeddings to infer on-target and off-target effects and elucidate mechanisms of action.

Key Features:

• Knowledge Graph-Based Approach: Formulates DTI prediction as a link prediction problem within knowledge graphs constructed from biomedical knowledge bases that represent entities connected to drugs and targets.
• TriModel Embedding Framework: Applies the TriModel knowledge graph embedding model to learn vector representations for drugs and target proteins and to compute interaction scores from those embeddings.
• Enhanced Predictive Accuracy: Validated through computer simulations and benchmark tests, showing superior performance versus five existing models with higher area under ROC and precision-recall curves.
• Comprehensive Proteome Coverage: Leverages extensive biomedical knowledge bases to extend proteome coverage beyond the limited drug processing capabilities of many existing models.
• Reduced False Positives: Mitigates high false positive prediction rates commonly associated with current approaches to provide more reliable interaction predictions.

Scientific Applications:

• Drug Discovery and Development: Identifies candidate drug–target interactions to support discovery and development of novel therapeutics.
• Mechanism of Action Analysis: Infers on-target and off-target interactions to aid interpretation of drug mechanisms of action.
• Safety Profiling: Predicts potential unintended drug interactions to inform safety assessment and risk of adverse effects.

Methodology:

Integrates biomedical knowledge bases to construct a knowledge graph, applies the TriModel embedding framework to learn vector representations for drugs and targets, computes interaction scores from those embeddings, and evaluates predictions against benchmark datasets using area under ROC and precision-recall curves; validation was performed via computer simulations.