A graph-based recovery and decomposition of Swanson's hypothesis using semantic predications

Area of Science:

• Biomedical Informatics
• Computational Biology
• Scientific Literature Analysis

Background:

• Manual hypothesis generation from literature is time-consuming.
• Swanson's Raynaud Syndrome-Fish Oil (RS-DFO) hypothesis was manually derived.
• Existing methods lack the ability to deeply analyze and decompose complex hypotheses from literature.

Purpose of the Study:

• To develop a semi-automatic methodology for recovering and decomposing Swanson's RS-DFO hypothesis.
• To demonstrate that manual hypothesis discovery techniques can be automated.
• To pave the way for fully automatic, semantics-based hypothesis generation.

Main Methods:

• Utilizing semantic predications (assertions) extracted from biomedical literature.
• Constructing labeled directed graphs to represent associations among concepts.
• Employing graph-based algorithms and structured background knowledge to uncover and decompose hypothesis details.

Main Results:

• Successfully recovered the three known associations of Swanson's hypothesis.
• Decomposed these into 16 additional detailed associations, forming chains of semantic predications.
• Retrieved 14 out of 19 total associations attributed to Swanson, a level of detail not previously achieved.

Conclusions:

• A novel methodology for semi-automatic hypothesis recovery and decomposition was presented.
• The approach highlights the importance of expressive representations beyond Swanson's ABC model for Literature-Based Discovery (LBD).
• Effective LBD requires accurate semantic information extraction and the integration of literature with structured knowledge.