PRID: Prediction Model Using RWR for Interactions between Drugs.

Area of Science:

• Pharmacology
• Computational Biology
• Bioinformatics

Background:

• Drug-drug interactions (DDIs) can cause unexpected pharmacological effects, complicating polypharmacy.
• Current DDI identification relies on clinical experience, lacking standardized databases for safe co-prescriptions.
• Existing computational methods for DDI prediction have limitations due to incomplete feature availability and inability to capture complex pathological mechanisms.

Purpose of the Study:

• To develop a novel deep learning model for predicting drug-drug interactions (DDIs).
• To integrate chemical structure similarity and protein-protein interaction (PPI) data for enhanced DDI prediction.
• To improve patient safety and treatment strategies by accurately identifying potential DDIs.

Main Methods:

• A deep learning model, PRID, was developed utilizing chemical structure similarity and drug-binding protein (CTET) information.
• The random walk with restart (RWR) algorithm was applied to propagate CTET proteins across a PPI network (STRING database).
• This approach incorporated hidden biological mechanisms between CTET proteins and disease-associated genes.

Main Results:

• The PRID model successfully predicted known drug-drug interactions, including those involving epilepsy drugs.
• RWR propagation of CTET proteins effectively captured indirectly co-regulated biological mechanisms relevant to DDIs.
• The study demonstrated PRID's effectiveness in predicting both known and novel drug combinations.

Conclusions:

• PRID offers a robust method for predicting drug-drug interactions by integrating diverse biological and chemical data.
• The model's ability to uncover hidden biological mechanisms enhances the understanding of DDI causation.
• PRID has the potential to identify novel DDIs and guide safer polypharmacy practices.