A machine learning framework for extracting information from biological pathway images in the literature

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Literature mining advances enable information extraction, but biological pathway images pose challenges for data retrieval.
• Extracting structured data from visual representations of biological pathways remains a significant hurdle in bioinformatics.

Purpose of the Study:

• To develop a machine learning framework, Extraction of Biological Pathway Information (EBPI), for automated extraction of biological pathway data from literature images.
• To enable high-throughput extraction of genes, enzymes, and metabolites from pathway diagrams.
• To generate extracted pathway information in a usable tabular format.

Main Methods:

• Developed the Extraction of Biological Pathway Information (EBPI) machine learning framework.
• Automated literature searches for relevant publications.
• Implemented image analysis to detect and classify text and determine biochemical reaction directions within pathway images.
• Generated extracted pathway data in a tabular format.

Main Results:

• The EBPI framework successfully automates the extraction of biological pathway information from images.
• The system identifies genes, enzymes, and metabolites and determines reaction directions.
• Performance evaluation showed accurate extraction when compared to manually curated pathway maps.

Conclusions:

• The Extraction of Biological Pathway Information (EBPI) framework offers a novel solution for high-throughput extraction of biological pathway data from scientific literature.
• This tool facilitates pathway studies and applications in areas like metabolic engineering.
• EBPI enhances the accessibility and usability of complex biological pathway information embedded in images.