Opportunities and Pitfalls with Large Language Models for Biomedical Annotation.

Area of Science:

• Biomedical Informatics
• Artificial Intelligence in Medicine
• Natural Language Processing

Background:

• Biomedical annotations are crucial for training and fine-tuning large language models (LLMs).
• Traditional human curation of these annotations is resource-intensive, involving significant time and cost.
• LLMs offer potential to streamline and accelerate the biomedical annotation process.

Purpose of the Study:

• To review the rise of LLMs in biomedicine.
• To discuss the opportunities presented by LLMs in biomedical annotation.
• To identify and analyze the limitations and risks associated with LLM application in this field.

Main Methods:

• Review of current literature and trends in LLM development and application.
• Analysis of the symbiotic relationship between LLMs and biomedical annotations.
• Discussion of the practical implications for annotation and curation workflows.

Main Results:

• LLMs can significantly enhance the efficiency of biomedical annotation.
• A feedback loop exists where LLMs improve curation, which in turn improves LLMs.
• New challenges and risks emerge with the integration of LLMs into biomedical data processes.

Conclusions:

• The application of LLMs in biomedical annotation presents a dual nature of significant opportunity and inherent risk.
• Understanding and mitigating these risks is essential for realizing the full potential of LLMs in accelerating biomedical research.
• Further investigation into best practices for LLM-assisted biomedical curation is warranted.