Microllm: a structured information extraction tool using large language models and named entity recognition in...

Area of Science:

• Microbial bioinformatics
• Computational biology
• Genomics

Background:

• Microbial community analysis requires detailed individual microorganism characteristics.
• Current bioinformatics methods face challenges with data scarcity, manual review, and NLP limitations.
• Effective microbial system manipulation is hindered by these data accessibility issues.

Purpose of the Study:

• To develop an automated tool for extracting microbial phenotypic information from unstructured text.
• To enhance the efficiency and accuracy of large-scale microbial data analysis.
• To facilitate the construction of structured microbial knowledge bases.

Main Methods:

• Developed MicroLLM, integrating fine-tuned large language models (LLMs) and BERT.
• Utilized named entity recognition (NER) for complex information extraction.
• Converted unstructured text data into structured JSON objects.

Main Results:

• MicroLLM efficiently extracts complex, multirelational, and multi-entity microbial information.
• The tool transforms unstructured text into organized, accessible JSON data.
• Demonstrated enhanced extraction of microbial phenotypic data compared to existing NLP methods.

Conclusions:

• MicroLLM provides a robust foundation for building large-scale structured microbial knowledge bases.
• The tool streamlines scientific knowledge creation in microbial bioinformatics.
• Advances understanding of microbial ecosystems and their impact on human health.