Multi-way association extraction and visualization from biological text documents using hyper-graphs: applications to...

Area of Science:

• Bioinformatics
• Computational Biology
• Text Mining

Background:

• Biological research generates vast knowledge on associations between entities like genes, proteins, and diseases.
• Existing text mining methods primarily focus on binary relationships, neglecting complex multi-way associations and their contextual information.
• Understanding multi-way relationships is crucial for a comprehensive grasp of biological systems.

Purpose of the Study:

• To develop and compare methods for extracting multi-way biological associations from literature.
• To represent these multi-way associations using a hyper-graph model.
• To investigate the computational feasibility and effectiveness of these approaches.

Main Methods:

• Developed two hyper-graph extraction approaches: exhaustive enumeration and an extension of the A Priori algorithm.
• Applied the text-based A Priori algorithm to unstructured biological text data.
• Utilized representative graph-based visualization for extracted hyper-graphs.

Main Results:

• The text-based A Priori method efficiently extracts hyper-edges representing multi-way associations.
• This approach achieves comparable results to exhaustive methods but with significantly reduced computational cost.
• Case studies on lung and colorectal cancer demonstrate the method's utility.

Conclusions:

• The text-based A Priori algorithm is a practical and effective tool for extracting hyper-graphs of multi-way biological associations.
• Hyper-graph visualization provides valuable contextual insights into gene-disease relationships.
• This method enhances understanding of complex biological interactions relevant to diseases.