QTLMiner: QTL database curation by mining tables in literature

Area of Science:

• Biomedical Informatics
• Genomics
• Data Science

Background:

• Biomedical literature contains extensive experimental results in figures and tables.
• Quantitative trait locus (QTL) information is often presented in tables within scientific papers.
• Existing text-mining methods primarily focus on unstructured text, neglecting valuable tabular data.

Purpose of the Study:

• To develop and evaluate a method for extracting QTL information from both tables and plain text in biomedical literature.
• To address the gap in current text-mining approaches by enabling the mining of structured data within tables.
• To reduce the labor-intensive nature of biological database curation.

Main Methods:

• A novel method was proposed to extract QTL information from heterogeneous and complex tables.
• Tables were converted into a structured database format.
• Information from plain text was integrated with the structured table data.

Main Results:

• The method was applied to curate a soybean QTL database.
• 2278 records were successfully extracted from 228 research papers.
• The method demonstrated a high precision rate of 96.9% and a recall rate of 83.3%, with an F-value of 89.6%.

Conclusions:

• The developed method effectively extracts QTL information from biomedical literature tables and text.
• This approach significantly enhances the efficiency and accuracy of biological database curation.
• The findings highlight the potential for automated knowledge extraction from structured data in scientific publications.