Crowdsourcing-Based Knowledge Graph Construction for Drug Side Effects Using Large Language Models with an...

Area of Science:

• Pharmacovigilance and Drug Safety
• Artificial Intelligence in Healthcare
• Social Media Data Mining

Background:

• Social media provides valuable real-world patient experience data for pharmacovigilance.
• Extracting structured information from noisy social media content is challenging.

Purpose of the Study:

• To present a systematic framework using large language models (LLMs) to extract medication side effects from social media.
• To organize extracted data into a knowledge graph (KG) for comprehensive analysis.
• To investigate semaglutide side effects for weight loss using Reddit data.

Main Methods:

• Developed a framework leveraging LLMs for automated extraction of medication side effects from social media.
• Applied the framework to semaglutide data from Reddit, constructing a knowledge graph.
• Analyzed reported side effects across different semaglutide brands and over time.
• Validated findings by comparing with the FAERS database.

Main Results:

• Successfully extracted and structured medication side effects from social media data using LLMs.
• Generated a knowledge graph detailing semaglutide side effects reported on Reddit.
• Identified patient-centered insights into semaglutide's safety profile, complementing FAERS data.

Conclusions:

• LLMs can effectively transform unstructured social media data into structured knowledge graphs for pharmacovigilance.
• This framework provides valuable patient-centered insights into drug side effects, enhancing safety monitoring.
• The approach demonstrates feasibility for broader applications in pharmacovigilance and real-world evidence generation.