Practical Aspects of Using Large Language Models to Screen Abstracts for Cardiovascular Drug Development:...

Area of Science:

• Pharmacology
• Artificial Intelligence
• Biomedical Informatics

Background:

• Cardiovascular drug development relies on comprehensive literature synthesis.
• Efficiently screening scientific literature is crucial for identifying drug indications, mechanisms, biomarkers, and outcomes.
• Current literature screening methods can be time-consuming and resource-intensive.

Purpose of the Study:

• To investigate the performance, cost, and prompt engineering trade-offs of three large language models (LLMs).
• To assess the potential of LLMs in accelerating the literature screening process for cardiovascular drug development.
• To provide insights into optimizing LLM application for biomedical research.

Main Methods:

• Evaluation of three distinct large language models.
• Analysis of model performance metrics relevant to literature screening.
• Assessment of associated costs and prompt engineering strategies.
• Comparative analysis of LLM capabilities in synthesizing cardiovascular drug development literature.

Main Results:

• Demonstrated significant acceleration in literature screening compared to traditional methods.
• Identified varying performance and cost efficiencies across the evaluated LLMs.
• Highlighted the impact of prompt engineering on the accuracy and relevance of LLM outputs.
• Provided a framework for selecting appropriate LLMs based on specific research needs.

Conclusions:

• Large language models show promise in expediting cardiovascular drug development literature synthesis.
• Careful consideration of model choice, cost, and prompt design is essential for effective implementation.
• LLMs can enhance the efficiency of identifying key information for drug discovery and development.