Predictive and Personalized Drug Query System

Area of Science:

• Pharmacology
• Bioinformatics
• Data Science

Background:

• Drug selection is complex due to interactions, side effects, and contraindications.
• Patient characteristics like age, gender, and genetics influence drug properties and efficacy.
• Existing drug information systems struggle with noisy, incomplete data and personalized needs.

Purpose of the Study:

• To develop a tool assisting medical professionals and consumers in choosing personalized pharmaceutical drugs.
• To create a query system that tailors drug recommendations based on specific patient profiles.
• To address data limitations by considering both exact and approximate query matches.

Main Methods:

• Representing pharmaceutical drug information as a heterogeneous graph.
• Modeling drug information retrieval as a subgraph matching problem.
• Leveraging graph structure and heterogeneity to quantify edge likelihood and rank answers.

Main Results:

• The network-based approach significantly improves edge likelihood quantification.
• Achieved up to an 18% improvement in the area under the receiver operating characteristic curve compared to baseline methods.
• Demonstrated the system's benefits through a functional prototype and illustrative examples.

Conclusions:

• The proposed graph-based approach effectively handles noisy drug data for personalized recommendations.
• The system enhances drug selection by considering patient profiles and approximate matches.
• This tool has the potential to improve medication safety and efficacy through tailored drug discovery.