Precision Drug Repurposing (PDR): Patient-level modeling and prediction combining foundational knowledge graph with...

Area of Science:

• Biomedical Informatics
• Pharmacogenomics
• Computational Biology

Background:

• Drug repurposing accelerates therapeutic development but struggles with individual patient variability.
• Personalized medicine requires integrating patient-specific data for tailored drug discovery.

Purpose of the Study:

• Introduce a Precision Drug Repurposing (PDR) framework for single-patient resolution.
• Enable personalized drug discovery by integrating individual data with a biomedical knowledge graph.

Main Methods:

• Developed a framework integrating UK Biobank data (Polygenic Risk Scores, biomarkers, medical history) with a biomedical knowledge graph.
• Used Alzheimer's Disease as a case study, comparing patient-specific models against a foundational model using link prediction.
• Evaluated candidate drugs via patient medication history and literature review.

Main Results:

• The PDR framework maintained robust prediction capabilities, with Polygenic Risk Scores significantly influencing drug prioritization (Cohen's d = 1.05).
• Ablation studies confirmed the crucial role of Polygenic Risk Scores (PRS).
• Patient-specific models identified novel drug candidates missed by the foundational model, validated by medication history and literature aligned with genetic profiles.

Conclusions:

• Demonstrates a promising approach for precision drug repurposing by integrating patient-specific data with knowledge graphs.
• Highlights the potential of Polygenic Risk Scores in personalizing drug discovery for complex diseases.