Automated extraction of potential migraine biomarkers using a semantic graph

Area of Science:

• Biomedical informatics
• Computational biology
• Data science

Background:

• Biomedical literature and databases are rich sources for potential disease biomarkers.
• Integrating information from diverse sources is challenging and resource-intensive.
• Automated methods are needed to efficiently identify biomarkers.

Purpose of the Study:

• To demonstrate the utility of semantically integrated knowledge graphs for automated biomarker identification.
• To specifically identify potential migraine biomarkers using this approach.
• To evaluate the performance of the knowledge graph against manual literature reviews.

Main Methods:

• Constructed a knowledge graph with over 3.5 million biomedical concepts and 68.4 million relationships.
• Filtered and ranked biochemical compounds based on their association with migraine-related concepts within the graph.
• Validated automated results against a systematic literature review conducted by migraine researchers.

Main Results:

• Automated ranking showed high consistency with manual review findings.
• 73% of reference compounds were ranked within the top 2000 by the knowledge graph.
• Achieved a Receiver Operating Characteristic Area Under the Curve (ROC-AUC) of 0.974, indicating strong performance.

Conclusions:

• Semantically integrated knowledge graphs are effective tools for assisting researchers in biomarker discovery.
• This approach streamlines the identification of potential disease biomarkers from complex data.
• The findings support the use of knowledge graphs in accelerating biomedical research.